{ "meta": { "generated_at": "2026-06-17T20:43:32.053504+00:00", "window_count": 1161, "feature_dim": 8546, "object_label_rows": 1161, "object_vocab_count": 34, "timeline_prediction_rows": 2079, "source_policy": "Window-level labels, features, predictions, and diagnostics are embedded here. Official raw MP4/HDF5/RRD files are linked from the Raw Sample Browser, with compact browser-preview clips for immediate MP4/audio playback.", "annotation_hash_recorded": true, "summary": { "num_windows": 1161, "feature_dim": 8546, "window_frames": 20, "stride_frames": 5 } }, "tasks": { "timeline_action": "Action Recognition", "timeline_subtask": "Procedure Step Recognition", "transition_detection": "Action Boundary Detection", "next_action": "Next-Action Prediction", "contact_prediction": "Contact State Prediction", "object_relevance": "Object Relevance Prediction" }, "task_display_names": { "timeline_action": "Action Recognition", "timeline_subtask": "Procedure Step Recognition", "transition_detection": "Action Boundary Detection", "next_action": "Next-Action Prediction", "hand_trajectory_forecast": "Hand Trajectory Forecasting", "contact_prediction": "Contact State Prediction", "object_relevance": "Object Relevance Prediction", "caption_grounding": "Language Grounding", "cross_modal_retrieval": "Cross-Modal Retrieval", "modality_reconstruction": "Cross-Modal Reconstruction", "temporal_order": "Temporal Order Verification", "misalignment_detection": "Multimodal Synchronization Detection" }, "feature_blocks": [ { "name": "hand_left_joints", "display": "Left Hand", "modality": "motion_capture", "start": 0, "end": 441, "dim": 441 }, { "name": "hand_right_joints", "display": "Right Hand", "modality": "motion_capture", "start": 441, "end": 882, "dim": 441 }, { "name": "body_joints", "display": "Body Joints", "modality": "motion_capture", "start": 882, "end": 1974, "dim": 1092 }, { "name": "body_contacts", "display": "Body Contacts", "modality": "motion_capture", "start": 1974, "end": 2121, "dim": 147 }, { "name": "camera_translation", "display": "Camera Translation", "modality": "pose_slam", "start": 2121, "end": 2142, "dim": 21 }, { "name": "camera_rotation_matrix", "display": "Camera Rotation", "modality": "pose_slam", "start": 2142, "end": 2205, "dim": 63 }, { "name": "imu_accel_gyro", "display": "IMU Accel/Gyro", "modality": "inertial", "start": 2205, "end": 2247, "dim": 42 }, { "name": "depth_confidence", "display": "Depth + Confidence", "modality": "depth", "start": 2247, "end": 3227, "dim": 980 }, { "name": "video_fisheye_cam0", "display": "Video Fisheye Cam0", "modality": "video", "start": 3227, "end": 3913, "dim": 686 }, { "name": "video_fisheye_cam1", "display": "Video Fisheye Cam1", "modality": "video", "start": 3913, "end": 4599, "dim": 686 }, { "name": "video_fisheye_cam2", "display": "Video Fisheye Cam2", "modality": "video", "start": 4599, "end": 5285, "dim": 686 }, { "name": "video_fisheye_cam3", "display": "Video Fisheye Cam3", "modality": "video", "start": 5285, "end": 5971, "dim": 686 }, { "name": "video_stereo_left", "display": "Video Stereo Left", "modality": "video", "start": 5971, "end": 6657, "dim": 686 }, { "name": "video_stereo_right", "display": "Video Stereo Right", "modality": "video", "start": 6657, "end": 7343, "dim": 686 }, { "name": "audio_fisheye_cam0_aac", "display": "Audio", "modality": "audio", "start": 7343, "end": 7511, "dim": 168 }, { "name": "caption_objects_interaction_text", "display": "Language Text", "modality": "language", "start": 7511, "end": 8407, "dim": 896 }, { "name": "slam_point_cloud", "display": "SLAM Point Cloud", "modality": "pose_slam", "start": 8407, "end": 8429, "dim": 22 }, { "name": "calibration", "display": "Calibration", "modality": "pose_slam", "start": 8429, "end": 8546, "dim": 117 } ], "segments": [ { "action": "Pick up kettle", "start_frame": 0, "end_frame": 164, "start_window": 0, "end_window": 29 }, { "action": "", "start_frame": 150, "end_frame": 169, "start_window": 30, "end_window": 30 }, { "action": "Position kettle to pour", "start_frame": 155, "end_frame": 804, "start_window": 31, "end_window": 157 }, { "action": "", "start_frame": 790, "end_frame": 809, "start_window": 158, "end_window": 158 }, { "action": "Move kettle", "start_frame": 795, "end_frame": 1004, "start_window": 159, "end_window": 197 }, { "action": "", "start_frame": 990, "end_frame": 1009, "start_window": 198, "end_window": 198 }, { "action": "Hold coffee carafe", "start_frame": 995, "end_frame": 1244, "start_window": 199, "end_window": 245 }, { "action": "", "start_frame": 1230, "end_frame": 1249, "start_window": 246, "end_window": 246 }, { "action": "Grasp coffee scoop", "start_frame": 1235, "end_frame": 1564, "start_window": 247, "end_window": 309 }, { "action": "", "start_frame": 1550, "end_frame": 1569, "start_window": 310, "end_window": 310 }, { "action": "Transfer coffee to dripper", "start_frame": 1555, "end_frame": 2004, "start_window": 311, "end_window": 397 }, { "action": "", "start_frame": 1990, "end_frame": 2009, "start_window": 398, "end_window": 398 }, { "action": "Hold gooseneck kettle", "start_frame": 1995, "end_frame": 2804, "start_window": 399, "end_window": 557 }, { "action": "", "start_frame": 2790, "end_frame": 2809, "start_window": 558, "end_window": 558 }, { "action": "Grasp gooseneck kettle", "start_frame": 2795, "end_frame": 2924, "start_window": 559, "end_window": 581 }, { "action": "", "start_frame": 2910, "end_frame": 2929, "start_window": 582, "end_window": 582 }, { "action": "Lift gooseneck kettle", "start_frame": 2915, "end_frame": 3404, "start_window": 583, "end_window": 677 }, { "action": "", "start_frame": 3390, "end_frame": 3409, "start_window": 678, "end_window": 678 }, { "action": "Move kettle away", "start_frame": 3395, "end_frame": 3564, "start_window": 679, "end_window": 709 }, { "action": "", "start_frame": 3550, "end_frame": 3569, "start_window": 710, "end_window": 710 }, { "action": "Place kettle on table", "start_frame": 3555, "end_frame": 3684, "start_window": 711, "end_window": 733 }, { "action": "", "start_frame": 3670, "end_frame": 3689, "start_window": 734, "end_window": 734 }, { "action": "Pick up white bottle", "start_frame": 3675, "end_frame": 3804, "start_window": 735, "end_window": 757 }, { "action": "", "start_frame": 3790, "end_frame": 3809, "start_window": 758, "end_window": 758 }, { "action": "Pour liquid from white bottle", "start_frame": 3795, "end_frame": 3924, "start_window": 759, "end_window": 781 }, { "action": "", "start_frame": 3910, "end_frame": 3929, "start_window": 782, "end_window": 782 }, { "action": "Close bottle cap", "start_frame": 3915, "end_frame": 4404, "start_window": 783, "end_window": 877 }, { "action": "", "start_frame": 4390, "end_frame": 4409, "start_window": 878, "end_window": 878 }, { "action": "Place item on table", "start_frame": 4395, "end_frame": 4524, "start_window": 879, "end_window": 901 }, { "action": "", "start_frame": 4510, "end_frame": 4529, "start_window": 902, "end_window": 902 }, { "action": "Wait/Prepare for pouring", "start_frame": 4515, "end_frame": 5004, "start_window": 903, "end_window": 997 }, { "action": "", "start_frame": 4990, "end_frame": 5009, "start_window": 998, "end_window": 998 }, { "action": "Pour coffee", "start_frame": 4995, "end_frame": 5804, "start_window": 999, "end_window": 1157 }, { "action": "", "start_frame": 5790, "end_frame": 5809, "start_window": 1158, "end_window": 1158 }, { "action": "Pour milk into coffee", "start_frame": 5795, "end_frame": 5819, "start_window": 1159, "end_window": 1160 } ], "windows": [ { "window_index": 0, "start_frame": 0, "end_frame": 19, "center_frame": 9, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.071905, "mean_abs": 0.20783, "relative": 0.849608 }, { "name": "hand_right_joints", "l2": 6.215975, "mean_abs": 0.164366, "relative": 0.489985 }, { "name": "body_joints", "l2": 5.27678, "mean_abs": 0.082969, "relative": 0.71273 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.080139, "mean_abs": 0.011786, "relative": 0.189116 }, { "name": "camera_rotation_matrix", "l2": 3.005863, "mean_abs": 0.212246, "relative": 0.923052 }, { "name": "imu_accel_gyro", "l2": 17.096228, "mean_abs": 1.086235, "relative": 0.9224 }, { "name": "depth_confidence", "l2": 10.91011, "mean_abs": 0.202228, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.26361, "mean_abs": 0.109948, "relative": 0.953127 }, { "name": "video_fisheye_cam1", "l2": 5.302784, "mean_abs": 0.112948, "relative": 0.87624 }, { "name": "video_fisheye_cam2", "l2": 5.595868, "mean_abs": 0.115647, "relative": 0.893297 }, { "name": "video_fisheye_cam3", "l2": 5.62084, "mean_abs": 0.114794, "relative": 0.941142 }, { "name": "video_stereo_left", "l2": 5.377779, "mean_abs": 0.11919, "relative": 0.8739 }, { "name": "video_stereo_right", "l2": 6.344293, "mean_abs": 0.142023, "relative": 0.894412 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.141961, "mean_abs": 0.089134, "relative": 0.766576 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007487, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 1, "start_frame": 5, "end_frame": 24, "center_frame": 14, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.150985, "mean_abs": 0.209775, "relative": 0.856279 }, { "name": "hand_right_joints", "l2": 9.110023, "mean_abs": 0.268842, "relative": 0.718113 }, { "name": "body_joints", "l2": 5.48176, "mean_abs": 0.087528, "relative": 0.740417 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.087905, "mean_abs": 0.012773, "relative": 0.207444 }, { "name": "camera_rotation_matrix", "l2": 3.00595, "mean_abs": 0.211683, "relative": 0.923078 }, { "name": "imu_accel_gyro", "l2": 17.154438, "mean_abs": 1.076036, "relative": 0.925541 }, { "name": "depth_confidence", "l2": 10.811137, "mean_abs": 0.201346, "relative": 0.000359 }, { "name": "video_fisheye_cam0", "l2": 5.263857, "mean_abs": 0.110381, "relative": 0.953172 }, { "name": "video_fisheye_cam1", "l2": 5.271797, "mean_abs": 0.113519, "relative": 0.87112 }, { "name": "video_fisheye_cam2", "l2": 5.586165, "mean_abs": 0.116697, "relative": 0.891748 }, { "name": "video_fisheye_cam3", "l2": 5.641662, "mean_abs": 0.114555, "relative": 0.944628 }, { "name": "video_stereo_left", "l2": 5.364504, "mean_abs": 0.121178, "relative": 0.871743 }, { "name": "video_stereo_right", "l2": 6.310376, "mean_abs": 0.143553, "relative": 0.889631 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.11729, "mean_abs": 0.088273, "relative": 0.757746 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007487, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 2, "start_frame": 10, "end_frame": 29, "center_frame": 19, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.232128, "mean_abs": 0.212748, "relative": 0.863124 }, { "name": "hand_right_joints", "l2": 9.907442, "mean_abs": 0.29529, "relative": 0.780971 }, { "name": "body_joints", "l2": 5.426262, "mean_abs": 0.085505, "relative": 0.732921 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.106138, "mean_abs": 0.014795, "relative": 0.250471 }, { "name": "camera_rotation_matrix", "l2": 3.006879, "mean_abs": 0.209828, "relative": 0.923364 }, { "name": "imu_accel_gyro", "l2": 17.039762, "mean_abs": 1.054595, "relative": 0.919354 }, { "name": "depth_confidence", "l2": 10.760702, "mean_abs": 0.202276, "relative": 0.000357 }, { "name": "video_fisheye_cam0", "l2": 5.265715, "mean_abs": 0.110543, "relative": 0.953508 }, { "name": "video_fisheye_cam1", "l2": 5.266138, "mean_abs": 0.113667, "relative": 0.870185 }, { "name": "video_fisheye_cam2", "l2": 5.554532, "mean_abs": 0.116091, "relative": 0.886698 }, { "name": "video_fisheye_cam3", "l2": 5.628293, "mean_abs": 0.114513, "relative": 0.94239 }, { "name": "video_stereo_left", "l2": 5.334852, "mean_abs": 0.121102, "relative": 0.866924 }, { "name": "video_stereo_right", "l2": 6.267776, "mean_abs": 0.143558, "relative": 0.883625 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.157628, "mean_abs": 0.089035, "relative": 0.772183 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007487, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 3, "start_frame": 15, "end_frame": 34, "center_frame": 24, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.335126, "mean_abs": 0.214525, "relative": 0.871812 }, { "name": "hand_right_joints", "l2": 10.837162, "mean_abs": 0.319864, "relative": 0.854258 }, { "name": "body_joints", "l2": 5.428369, "mean_abs": 0.084186, "relative": 0.733205 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.096323, "mean_abs": 0.013755, "relative": 0.227309 }, { "name": "camera_rotation_matrix", "l2": 3.002172, "mean_abs": 0.205735, "relative": 0.921918 }, { "name": "imu_accel_gyro", "l2": 17.153273, "mean_abs": 1.104169, "relative": 0.925478 }, { "name": "depth_confidence", "l2": 10.664346, "mean_abs": 0.198642, "relative": 0.000354 }, { "name": "video_fisheye_cam0", "l2": 5.253023, "mean_abs": 0.109543, "relative": 0.95121 }, { "name": "video_fisheye_cam1", "l2": 5.245481, "mean_abs": 0.112316, "relative": 0.866771 }, { "name": "video_fisheye_cam2", "l2": 5.521056, "mean_abs": 0.114102, "relative": 0.881354 }, { "name": "video_fisheye_cam3", "l2": 5.60478, "mean_abs": 0.113389, "relative": 0.938453 }, { "name": "video_stereo_left", "l2": 5.290122, "mean_abs": 0.117978, "relative": 0.859655 }, { "name": "video_stereo_right", "l2": 6.235144, "mean_abs": 0.140581, "relative": 0.879025 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.21277, "mean_abs": 0.094391, "relative": 0.791917 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007487, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 4, "start_frame": 20, "end_frame": 39, "center_frame": 29, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.380361, "mean_abs": 0.211917, "relative": 0.875628 }, { "name": "hand_right_joints", "l2": 11.587594, "mean_abs": 0.330609, "relative": 0.913412 }, { "name": "body_joints", "l2": 5.549197, "mean_abs": 0.085959, "relative": 0.749525 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.085688, "mean_abs": 0.013232, "relative": 0.202212 }, { "name": "camera_rotation_matrix", "l2": 2.999711, "mean_abs": 0.202318, "relative": 0.921162 }, { "name": "imu_accel_gyro", "l2": 17.154985, "mean_abs": 1.124466, "relative": 0.925571 }, { "name": "depth_confidence", "l2": 10.628193, "mean_abs": 0.197535, "relative": 0.000353 }, { "name": "video_fisheye_cam0", "l2": 5.243904, "mean_abs": 0.109017, "relative": 0.949559 }, { "name": "video_fisheye_cam1", "l2": 5.222509, "mean_abs": 0.110865, "relative": 0.862976 }, { "name": "video_fisheye_cam2", "l2": 5.496797, "mean_abs": 0.112323, "relative": 0.877482 }, { "name": "video_fisheye_cam3", "l2": 5.585249, "mean_abs": 0.11306, "relative": 0.935183 }, { "name": "video_stereo_left", "l2": 5.272299, "mean_abs": 0.116378, "relative": 0.856759 }, { "name": "video_stereo_right", "l2": 6.230491, "mean_abs": 0.139646, "relative": 0.878368 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.254999, "mean_abs": 0.095293, "relative": 0.80703 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007487, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 5, "start_frame": 25, "end_frame": 44, "center_frame": 34, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.386657, "mean_abs": 0.210641, "relative": 0.876159 }, { "name": "hand_right_joints", "l2": 11.304488, "mean_abs": 0.309954, "relative": 0.891096 }, { "name": "body_joints", "l2": 5.525513, "mean_abs": 0.085232, "relative": 0.746327 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.063946, "mean_abs": 0.010158, "relative": 0.150904 }, { "name": "camera_rotation_matrix", "l2": 3.005166, "mean_abs": 0.20641, "relative": 0.922838 }, { "name": "imu_accel_gyro", "l2": 17.191223, "mean_abs": 1.135743, "relative": 0.927526 }, { "name": "depth_confidence", "l2": 10.606854, "mean_abs": 0.198583, "relative": 0.000352 }, { "name": "video_fisheye_cam0", "l2": 5.240405, "mean_abs": 0.109353, "relative": 0.948925 }, { "name": "video_fisheye_cam1", "l2": 5.238441, "mean_abs": 0.112441, "relative": 0.865608 }, { "name": "video_fisheye_cam2", "l2": 5.503064, "mean_abs": 0.113407, "relative": 0.878482 }, { "name": "video_fisheye_cam3", "l2": 5.56439, "mean_abs": 0.11322, "relative": 0.93169 }, { "name": "video_stereo_left", "l2": 5.299284, "mean_abs": 0.11924, "relative": 0.861144 }, { "name": "video_stereo_right", "l2": 6.28964, "mean_abs": 0.142694, "relative": 0.886707 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.17389, "mean_abs": 0.089363, "relative": 0.778003 }, { "name": "caption_objects_interaction_text", "l2": 1.881092, "mean_abs": 0.012769, "relative": 0.863752 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 6, "start_frame": 30, "end_frame": 49, "center_frame": 39, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.362794, "mean_abs": 0.211256, "relative": 0.874146 }, { "name": "hand_right_joints", "l2": 11.120339, "mean_abs": 0.322395, "relative": 0.87658 }, { "name": "body_joints", "l2": 5.425999, "mean_abs": 0.083845, "relative": 0.732885 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030628, "mean_abs": 0.004644, "relative": 0.072278 }, { "name": "camera_rotation_matrix", "l2": 3.004729, "mean_abs": 0.207067, "relative": 0.922703 }, { "name": "imu_accel_gyro", "l2": 17.110546, "mean_abs": 1.123113, "relative": 0.923173 }, { "name": "depth_confidence", "l2": 10.587531, "mean_abs": 0.197642, "relative": 0.000351 }, { "name": "video_fisheye_cam0", "l2": 5.245223, "mean_abs": 0.109577, "relative": 0.949798 }, { "name": "video_fisheye_cam1", "l2": 5.266428, "mean_abs": 0.113481, "relative": 0.870233 }, { "name": "video_fisheye_cam2", "l2": 5.520807, "mean_abs": 0.114066, "relative": 0.881315 }, { "name": "video_fisheye_cam3", "l2": 5.550081, "mean_abs": 0.112733, "relative": 0.929294 }, { "name": "video_stereo_left", "l2": 5.316826, "mean_abs": 0.11947, "relative": 0.863995 }, { "name": "video_stereo_right", "l2": 6.36061, "mean_abs": 0.144019, "relative": 0.896713 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.181581, "mean_abs": 0.091763, "relative": 0.780755 }, { "name": "caption_objects_interaction_text", "l2": 1.881092, "mean_abs": 0.013159, "relative": 0.863752 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 7, "start_frame": 35, "end_frame": 54, "center_frame": 44, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.318597, "mean_abs": 0.213652, "relative": 0.870418 }, { "name": "hand_right_joints", "l2": 10.857965, "mean_abs": 0.32316, "relative": 0.855898 }, { "name": "body_joints", "l2": 5.438507, "mean_abs": 0.08426, "relative": 0.734575 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030329, "mean_abs": 0.004599, "relative": 0.071573 }, { "name": "camera_rotation_matrix", "l2": 3.00237, "mean_abs": 0.204569, "relative": 0.921979 }, { "name": "imu_accel_gyro", "l2": 17.245008, "mean_abs": 1.120363, "relative": 0.930428 }, { "name": "depth_confidence", "l2": 10.58062, "mean_abs": 0.195433, "relative": 0.000351 }, { "name": "video_fisheye_cam0", "l2": 5.250477, "mean_abs": 0.109312, "relative": 0.950749 }, { "name": "video_fisheye_cam1", "l2": 5.279491, "mean_abs": 0.113222, "relative": 0.872391 }, { "name": "video_fisheye_cam2", "l2": 5.547081, "mean_abs": 0.11377, "relative": 0.885509 }, { "name": "video_fisheye_cam3", "l2": 5.529569, "mean_abs": 0.111465, "relative": 0.92586 }, { "name": "video_stereo_left", "l2": 5.324302, "mean_abs": 0.118001, "relative": 0.86521 }, { "name": "video_stereo_right", "l2": 6.41603, "mean_abs": 0.143437, "relative": 0.904526 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.155805, "mean_abs": 0.090026, "relative": 0.77153 }, { "name": "caption_objects_interaction_text", "l2": 1.881092, "mean_abs": 0.013224, "relative": 0.863752 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 8, "start_frame": 40, "end_frame": 59, "center_frame": 49, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.292138, "mean_abs": 0.214444, "relative": 0.868186 }, { "name": "hand_right_joints", "l2": 9.803386, "mean_abs": 0.268256, "relative": 0.772769 }, { "name": "body_joints", "l2": 5.56978, "mean_abs": 0.087542, "relative": 0.752306 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.061862, "mean_abs": 0.009056, "relative": 0.145985 }, { "name": "camera_rotation_matrix", "l2": 3.001326, "mean_abs": 0.20467, "relative": 0.921658 }, { "name": "imu_accel_gyro", "l2": 17.198103, "mean_abs": 1.116269, "relative": 0.927897 }, { "name": "depth_confidence", "l2": 10.577054, "mean_abs": 0.194953, "relative": 0.000351 }, { "name": "video_fisheye_cam0", "l2": 5.254185, "mean_abs": 0.108933, "relative": 0.951421 }, { "name": "video_fisheye_cam1", "l2": 5.317717, "mean_abs": 0.113245, "relative": 0.878708 }, { "name": "video_fisheye_cam2", "l2": 5.560911, "mean_abs": 0.113528, "relative": 0.887717 }, { "name": "video_fisheye_cam3", "l2": 5.565385, "mean_abs": 0.111649, "relative": 0.931857 }, { "name": "video_stereo_left", "l2": 5.33448, "mean_abs": 0.116944, "relative": 0.866864 }, { "name": "video_stereo_right", "l2": 6.457295, "mean_abs": 0.143679, "relative": 0.910343 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.017378, "mean_abs": 0.07844, "relative": 0.721989 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 9, "start_frame": 45, "end_frame": 64, "center_frame": 54, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.215645, "mean_abs": 0.213739, "relative": 0.861733 }, { "name": "hand_right_joints", "l2": 9.854802, "mean_abs": 0.295652, "relative": 0.776822 }, { "name": "body_joints", "l2": 5.757971, "mean_abs": 0.093029, "relative": 0.777725 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.109081, "mean_abs": 0.016241, "relative": 0.257416 }, { "name": "camera_rotation_matrix", "l2": 3.00393, "mean_abs": 0.208557, "relative": 0.922458 }, { "name": "imu_accel_gyro", "l2": 17.167713, "mean_abs": 1.128009, "relative": 0.926257 }, { "name": "depth_confidence", "l2": 10.637951, "mean_abs": 0.198032, "relative": 0.000353 }, { "name": "video_fisheye_cam0", "l2": 5.254636, "mean_abs": 0.109326, "relative": 0.951502 }, { "name": "video_fisheye_cam1", "l2": 5.371398, "mean_abs": 0.11436, "relative": 0.887578 }, { "name": "video_fisheye_cam2", "l2": 5.579643, "mean_abs": 0.114207, "relative": 0.890707 }, { "name": "video_fisheye_cam3", "l2": 5.607028, "mean_abs": 0.113512, "relative": 0.938829 }, { "name": "video_stereo_left", "l2": 5.342422, "mean_abs": 0.118479, "relative": 0.868154 }, { "name": "video_stereo_right", "l2": 6.466339, "mean_abs": 0.146114, "relative": 0.911618 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.031218, "mean_abs": 0.081196, "relative": 0.726943 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 10, "start_frame": 50, "end_frame": 69, "center_frame": 59, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.174905, "mean_abs": 0.215787, "relative": 0.858297 }, { "name": "hand_right_joints", "l2": 10.280642, "mean_abs": 0.309242, "relative": 0.81039 }, { "name": "body_joints", "l2": 5.917358, "mean_abs": 0.096587, "relative": 0.799253 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.129067, "mean_abs": 0.019002, "relative": 0.30458 }, { "name": "camera_rotation_matrix", "l2": 3.005853, "mean_abs": 0.211968, "relative": 0.923049 }, { "name": "imu_accel_gyro", "l2": 17.227596, "mean_abs": 1.146174, "relative": 0.929488 }, { "name": "depth_confidence", "l2": 10.6776, "mean_abs": 0.199912, "relative": 0.000354 }, { "name": "video_fisheye_cam0", "l2": 5.261408, "mean_abs": 0.109927, "relative": 0.952729 }, { "name": "video_fisheye_cam1", "l2": 5.421916, "mean_abs": 0.116022, "relative": 0.895926 }, { "name": "video_fisheye_cam2", "l2": 5.593826, "mean_abs": 0.115425, "relative": 0.892971 }, { "name": "video_fisheye_cam3", "l2": 5.62968, "mean_abs": 0.114908, "relative": 0.942622 }, { "name": "video_stereo_left", "l2": 5.380738, "mean_abs": 0.120502, "relative": 0.874381 }, { "name": "video_stereo_right", "l2": 6.492383, "mean_abs": 0.147792, "relative": 0.91529 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.000808, "mean_abs": 0.079156, "relative": 0.716059 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 11, "start_frame": 55, "end_frame": 74, "center_frame": 64, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.162643, "mean_abs": 0.217606, "relative": 0.857262 }, { "name": "hand_right_joints", "l2": 10.728663, "mean_abs": 0.316603, "relative": 0.845706 }, { "name": "body_joints", "l2": 6.002609, "mean_abs": 0.096058, "relative": 0.810767 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.11658, "mean_abs": 0.017442, "relative": 0.275113 }, { "name": "camera_rotation_matrix", "l2": 3.00093, "mean_abs": 0.209128, "relative": 0.921537 }, { "name": "imu_accel_gyro", "l2": 17.125793, "mean_abs": 1.108538, "relative": 0.923996 }, { "name": "depth_confidence", "l2": 10.69073, "mean_abs": 0.198563, "relative": 0.000355 }, { "name": "video_fisheye_cam0", "l2": 5.262859, "mean_abs": 0.10948, "relative": 0.952991 }, { "name": "video_fisheye_cam1", "l2": 5.449652, "mean_abs": 0.116147, "relative": 0.900509 }, { "name": "video_fisheye_cam2", "l2": 5.606254, "mean_abs": 0.115522, "relative": 0.894955 }, { "name": "video_fisheye_cam3", "l2": 5.638626, "mean_abs": 0.114489, "relative": 0.94412 }, { "name": "video_stereo_left", "l2": 5.402489, "mean_abs": 0.119952, "relative": 0.877915 }, { "name": "video_stereo_right", "l2": 6.550119, "mean_abs": 0.148358, "relative": 0.923429 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.007804, "mean_abs": 0.078478, "relative": 0.718563 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 12, "start_frame": 60, "end_frame": 79, "center_frame": 69, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.171799, "mean_abs": 0.220691, "relative": 0.858035 }, { "name": "hand_right_joints", "l2": 11.663902, "mean_abs": 0.300238, "relative": 0.919428 }, { "name": "body_joints", "l2": 6.115053, "mean_abs": 0.092645, "relative": 0.825955 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.073812, "mean_abs": 0.011021, "relative": 0.174185 }, { "name": "camera_rotation_matrix", "l2": 3.000126, "mean_abs": 0.208905, "relative": 0.92129 }, { "name": "imu_accel_gyro", "l2": 17.168068, "mean_abs": 1.1211, "relative": 0.926276 }, { "name": "depth_confidence", "l2": 10.770989, "mean_abs": 0.200015, "relative": 0.000357 }, { "name": "video_fisheye_cam0", "l2": 5.266608, "mean_abs": 0.109747, "relative": 0.95367 }, { "name": "video_fisheye_cam1", "l2": 5.490323, "mean_abs": 0.116665, "relative": 0.90723 }, { "name": "video_fisheye_cam2", "l2": 5.62387, "mean_abs": 0.115633, "relative": 0.897767 }, { "name": "video_fisheye_cam3", "l2": 5.642755, "mean_abs": 0.113948, "relative": 0.944811 }, { "name": "video_stereo_left", "l2": 5.405042, "mean_abs": 0.12158, "relative": 0.87833 }, { "name": "video_stereo_right", "l2": 6.560255, "mean_abs": 0.149747, "relative": 0.924858 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.200298, "mean_abs": 0.09683, "relative": 0.787454 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 13, "start_frame": 65, "end_frame": 84, "center_frame": 74, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.216249, "mean_abs": 0.224649, "relative": 0.861784 }, { "name": "hand_right_joints", "l2": 11.474133, "mean_abs": 0.290184, "relative": 0.904469 }, { "name": "body_joints", "l2": 6.129619, "mean_abs": 0.090278, "relative": 0.827923 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.039839, "mean_abs": 0.006337, "relative": 0.094015 }, { "name": "camera_rotation_matrix", "l2": 3.009538, "mean_abs": 0.216826, "relative": 0.92418 }, { "name": "imu_accel_gyro", "l2": 17.180666, "mean_abs": 1.143023, "relative": 0.926956 }, { "name": "depth_confidence", "l2": 10.849826, "mean_abs": 0.204107, "relative": 0.00036 }, { "name": "video_fisheye_cam0", "l2": 5.262784, "mean_abs": 0.111085, "relative": 0.952978 }, { "name": "video_fisheye_cam1", "l2": 5.523172, "mean_abs": 0.119688, "relative": 0.912658 }, { "name": "video_fisheye_cam2", "l2": 5.618403, "mean_abs": 0.117395, "relative": 0.896895 }, { "name": "video_fisheye_cam3", "l2": 5.624982, "mean_abs": 0.115793, "relative": 0.941835 }, { "name": "video_stereo_left", "l2": 5.418926, "mean_abs": 0.124633, "relative": 0.880586 }, { "name": "video_stereo_right", "l2": 6.565949, "mean_abs": 0.152534, "relative": 0.925661 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.270465, "mean_abs": 0.099076, "relative": 0.812565 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 14, "start_frame": 70, "end_frame": 89, "center_frame": 79, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.292585, "mean_abs": 0.226096, "relative": 0.868223 }, { "name": "hand_right_joints", "l2": 11.201578, "mean_abs": 0.289531, "relative": 0.882984 }, { "name": "body_joints", "l2": 6.170469, "mean_abs": 0.097478, "relative": 0.83344 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.083136, "mean_abs": 0.013068, "relative": 0.196189 }, { "name": "camera_rotation_matrix", "l2": 3.027257, "mean_abs": 0.224428, "relative": 0.929622 }, { "name": "imu_accel_gyro", "l2": 17.181458, "mean_abs": 1.160231, "relative": 0.926999 }, { "name": "depth_confidence", "l2": 11.123414, "mean_abs": 0.21093, "relative": 0.000369 }, { "name": "video_fisheye_cam0", "l2": 5.262486, "mean_abs": 0.112108, "relative": 0.952924 }, { "name": "video_fisheye_cam1", "l2": 5.579896, "mean_abs": 0.122717, "relative": 0.922031 }, { "name": "video_fisheye_cam2", "l2": 5.616151, "mean_abs": 0.119043, "relative": 0.896535 }, { "name": "video_fisheye_cam3", "l2": 5.618724, "mean_abs": 0.117627, "relative": 0.940788 }, { "name": "video_stereo_left", "l2": 5.480419, "mean_abs": 0.128334, "relative": 0.890579 }, { "name": "video_stereo_right", "l2": 6.660361, "mean_abs": 0.158548, "relative": 0.938971 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.300109, "mean_abs": 0.101354, "relative": 0.823174 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 15, "start_frame": 75, "end_frame": 94, "center_frame": 84, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.263964, "mean_abs": 0.224544, "relative": 0.865809 }, { "name": "hand_right_joints", "l2": 10.978519, "mean_abs": 0.285249, "relative": 0.865401 }, { "name": "body_joints", "l2": 6.209739, "mean_abs": 0.101335, "relative": 0.838744 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.118916, "mean_abs": 0.018641, "relative": 0.280624 }, { "name": "camera_rotation_matrix", "l2": 3.039634, "mean_abs": 0.229549, "relative": 0.933422 }, { "name": "imu_accel_gyro", "l2": 17.492487, "mean_abs": 1.208528, "relative": 0.94378 }, { "name": "depth_confidence", "l2": 11.567143, "mean_abs": 0.216835, "relative": 0.000384 }, { "name": "video_fisheye_cam0", "l2": 5.265371, "mean_abs": 0.112306, "relative": 0.953446 }, { "name": "video_fisheye_cam1", "l2": 5.634363, "mean_abs": 0.124825, "relative": 0.931031 }, { "name": "video_fisheye_cam2", "l2": 5.607871, "mean_abs": 0.119625, "relative": 0.895213 }, { "name": "video_fisheye_cam3", "l2": 5.607872, "mean_abs": 0.118063, "relative": 0.938971 }, { "name": "video_stereo_left", "l2": 5.449302, "mean_abs": 0.128864, "relative": 0.885522 }, { "name": "video_stereo_right", "l2": 6.69346, "mean_abs": 0.159987, "relative": 0.943637 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.247306, "mean_abs": 0.094654, "relative": 0.804277 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 16, "start_frame": 80, "end_frame": 99, "center_frame": 89, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.251857, "mean_abs": 0.224929, "relative": 0.864788 }, { "name": "hand_right_joints", "l2": 10.68336, "mean_abs": 0.2723, "relative": 0.842135 }, { "name": "body_joints", "l2": 6.118504, "mean_abs": 0.098261, "relative": 0.826421 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.138675, "mean_abs": 0.020158, "relative": 0.327253 }, { "name": "camera_rotation_matrix", "l2": 3.028611, "mean_abs": 0.2246, "relative": 0.930037 }, { "name": "imu_accel_gyro", "l2": 17.437927, "mean_abs": 1.164648, "relative": 0.940836 }, { "name": "depth_confidence", "l2": 11.703682, "mean_abs": 0.216973, "relative": 0.000388 }, { "name": "video_fisheye_cam0", "l2": 5.230608, "mean_abs": 0.111402, "relative": 0.947151 }, { "name": "video_fisheye_cam1", "l2": 5.653731, "mean_abs": 0.12417, "relative": 0.934231 }, { "name": "video_fisheye_cam2", "l2": 5.56728, "mean_abs": 0.118331, "relative": 0.888733 }, { "name": "video_fisheye_cam3", "l2": 5.590134, "mean_abs": 0.117326, "relative": 0.936 }, { "name": "video_stereo_left", "l2": 5.360247, "mean_abs": 0.126528, "relative": 0.871051 }, { "name": "video_stereo_right", "l2": 6.615124, "mean_abs": 0.156991, "relative": 0.932594 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.328459, "mean_abs": 0.10103, "relative": 0.83332 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 17, "start_frame": 85, "end_frame": 104, "center_frame": 94, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.246542, "mean_abs": 0.228328, "relative": 0.86434 }, { "name": "hand_right_joints", "l2": 10.33279, "mean_abs": 0.252755, "relative": 0.8145 }, { "name": "body_joints", "l2": 5.915665, "mean_abs": 0.093843, "relative": 0.799024 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.116231, "mean_abs": 0.016092, "relative": 0.274289 }, { "name": "camera_rotation_matrix", "l2": 3.009881, "mean_abs": 0.215632, "relative": 0.924285 }, { "name": "imu_accel_gyro", "l2": 17.208269, "mean_abs": 1.069773, "relative": 0.928445 }, { "name": "depth_confidence", "l2": 12.248342, "mean_abs": 0.219227, "relative": 0.000407 }, { "name": "video_fisheye_cam0", "l2": 5.192027, "mean_abs": 0.109321, "relative": 0.940165 }, { "name": "video_fisheye_cam1", "l2": 5.64866, "mean_abs": 0.122136, "relative": 0.933393 }, { "name": "video_fisheye_cam2", "l2": 5.534801, "mean_abs": 0.11687, "relative": 0.883549 }, { "name": "video_fisheye_cam3", "l2": 5.526922, "mean_abs": 0.114645, "relative": 0.925416 }, { "name": "video_stereo_left", "l2": 5.332229, "mean_abs": 0.123938, "relative": 0.866498 }, { "name": "video_stereo_right", "l2": 6.551733, "mean_abs": 0.152981, "relative": 0.923657 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.436067, "mean_abs": 0.103767, "relative": 0.871832 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 18, "start_frame": 90, "end_frame": 109, "center_frame": 99, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.19907, "mean_abs": 0.225752, "relative": 0.860335 }, { "name": "hand_right_joints", "l2": 10.117474, "mean_abs": 0.241585, "relative": 0.797528 }, { "name": "body_joints", "l2": 5.776315, "mean_abs": 0.090729, "relative": 0.780202 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.089657, "mean_abs": 0.012087, "relative": 0.211578 }, { "name": "camera_rotation_matrix", "l2": 3.001094, "mean_abs": 0.21355, "relative": 0.921587 }, { "name": "imu_accel_gyro", "l2": 17.087904, "mean_abs": 1.005967, "relative": 0.921951 }, { "name": "depth_confidence", "l2": 12.370868, "mean_abs": 0.220678, "relative": 0.000411 }, { "name": "video_fisheye_cam0", "l2": 5.177875, "mean_abs": 0.108345, "relative": 0.937602 }, { "name": "video_fisheye_cam1", "l2": 5.667918, "mean_abs": 0.12151, "relative": 0.936576 }, { "name": "video_fisheye_cam2", "l2": 5.543074, "mean_abs": 0.116372, "relative": 0.884869 }, { "name": "video_fisheye_cam3", "l2": 5.510427, "mean_abs": 0.113201, "relative": 0.922655 }, { "name": "video_stereo_left", "l2": 5.34458, "mean_abs": 0.123895, "relative": 0.868505 }, { "name": "video_stereo_right", "l2": 6.556045, "mean_abs": 0.152533, "relative": 0.924265 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.434047, "mean_abs": 0.099926, "relative": 0.871109 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 19, "start_frame": 95, "end_frame": 114, "center_frame": 104, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.200437, "mean_abs": 0.223273, "relative": 0.86045 }, { "name": "hand_right_joints", "l2": 10.137251, "mean_abs": 0.241978, "relative": 0.799087 }, { "name": "body_joints", "l2": 5.695421, "mean_abs": 0.088457, "relative": 0.769276 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.057558, "mean_abs": 0.008538, "relative": 0.135829 }, { "name": "camera_rotation_matrix", "l2": 3.003187, "mean_abs": 0.214934, "relative": 0.92223 }, { "name": "imu_accel_gyro", "l2": 17.034746, "mean_abs": 0.977757, "relative": 0.919083 }, { "name": "depth_confidence", "l2": 12.462245, "mean_abs": 0.222349, "relative": 0.000414 }, { "name": "video_fisheye_cam0", "l2": 5.178857, "mean_abs": 0.108397, "relative": 0.93778 }, { "name": "video_fisheye_cam1", "l2": 5.667509, "mean_abs": 0.121351, "relative": 0.936508 }, { "name": "video_fisheye_cam2", "l2": 5.544228, "mean_abs": 0.115894, "relative": 0.885053 }, { "name": "video_fisheye_cam3", "l2": 5.508679, "mean_abs": 0.113108, "relative": 0.922362 }, { "name": "video_stereo_left", "l2": 5.351415, "mean_abs": 0.122709, "relative": 0.869616 }, { "name": "video_stereo_right", "l2": 6.53336, "mean_abs": 0.150957, "relative": 0.921067 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.310646, "mean_abs": 0.102637, "relative": 0.826946 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 20, "start_frame": 100, "end_frame": 119, "center_frame": 109, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.209168, "mean_abs": 0.220231, "relative": 0.861187 }, { "name": "hand_right_joints", "l2": 10.159209, "mean_abs": 0.238729, "relative": 0.800817 }, { "name": "body_joints", "l2": 5.638342, "mean_abs": 0.086468, "relative": 0.761566 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.046234, "mean_abs": 0.007298, "relative": 0.109105 }, { "name": "camera_rotation_matrix", "l2": 3.009762, "mean_abs": 0.215684, "relative": 0.924249 }, { "name": "imu_accel_gyro", "l2": 17.147964, "mean_abs": 1.03584, "relative": 0.925192 }, { "name": "depth_confidence", "l2": 12.549499, "mean_abs": 0.224518, "relative": 0.000416 }, { "name": "video_fisheye_cam0", "l2": 5.158047, "mean_abs": 0.108888, "relative": 0.934012 }, { "name": "video_fisheye_cam1", "l2": 5.651519, "mean_abs": 0.122994, "relative": 0.933866 }, { "name": "video_fisheye_cam2", "l2": 5.509894, "mean_abs": 0.116275, "relative": 0.879573 }, { "name": "video_fisheye_cam3", "l2": 5.5021, "mean_abs": 0.113413, "relative": 0.92126 }, { "name": "video_stereo_left", "l2": 5.382106, "mean_abs": 0.125199, "relative": 0.874603 }, { "name": "video_stereo_right", "l2": 6.509176, "mean_abs": 0.15229, "relative": 0.917657 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.415505, "mean_abs": 0.109754, "relative": 0.864473 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010215, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 21, "start_frame": 105, "end_frame": 124, "center_frame": 114, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.138266, "mean_abs": 0.219684, "relative": 0.855206 }, { "name": "hand_right_joints", "l2": 10.068336, "mean_abs": 0.234517, "relative": 0.793654 }, { "name": "body_joints", "l2": 5.650421, "mean_abs": 0.086689, "relative": 0.763198 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.117721, "mean_abs": 0.014944, "relative": 0.277805 }, { "name": "camera_rotation_matrix", "l2": 3.081764, "mean_abs": 0.236145, "relative": 0.94636 }, { "name": "imu_accel_gyro", "l2": 17.324966, "mean_abs": 1.068182, "relative": 0.934742 }, { "name": "depth_confidence", "l2": 12.684739, "mean_abs": 0.233993, "relative": 0.000421 }, { "name": "video_fisheye_cam0", "l2": 5.24115, "mean_abs": 0.114284, "relative": 0.94906 }, { "name": "video_fisheye_cam1", "l2": 5.836955, "mean_abs": 0.129962, "relative": 0.964507 }, { "name": "video_fisheye_cam2", "l2": 5.450885, "mean_abs": 0.119555, "relative": 0.870153 }, { "name": "video_fisheye_cam3", "l2": 5.582735, "mean_abs": 0.119574, "relative": 0.934762 }, { "name": "video_stereo_left", "l2": 5.391425, "mean_abs": 0.129368, "relative": 0.876117 }, { "name": "video_stereo_right", "l2": 6.56217, "mean_abs": 0.158936, "relative": 0.925128 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.508574, "mean_abs": 0.110679, "relative": 0.897781 }, { "name": "caption_objects_interaction_text", "l2": 1.95185, "mean_abs": 0.014731, "relative": 0.896242 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 22, "start_frame": 110, "end_frame": 129, "center_frame": 119, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.063232, "mean_abs": 0.234264, "relative": 0.848877 }, { "name": "hand_right_joints", "l2": 9.87281, "mean_abs": 0.236293, "relative": 0.778242 }, { "name": "body_joints", "l2": 5.88347, "mean_abs": 0.09812, "relative": 0.794675 }, { "name": "body_contacts", "l2": 3.463979, "mean_abs": 0.081735, "relative": 0.999589 }, { "name": "camera_translation", "l2": 0.226286, "mean_abs": 0.026839, "relative": 0.534003 }, { "name": "camera_rotation_matrix", "l2": 3.228876, "mean_abs": 0.259221, "relative": 0.991535 }, { "name": "imu_accel_gyro", "l2": 17.425911, "mean_abs": 1.107702, "relative": 0.940188 }, { "name": "depth_confidence", "l2": 12.806503, "mean_abs": 0.236482, "relative": 0.000425 }, { "name": "video_fisheye_cam0", "l2": 5.293462, "mean_abs": 0.117936, "relative": 0.958533 }, { "name": "video_fisheye_cam1", "l2": 6.007075, "mean_abs": 0.135714, "relative": 0.992619 }, { "name": "video_fisheye_cam2", "l2": 5.400272, "mean_abs": 0.122941, "relative": 0.862073 }, { "name": "video_fisheye_cam3", "l2": 5.679943, "mean_abs": 0.125355, "relative": 0.951038 }, { "name": "video_stereo_left", "l2": 5.479419, "mean_abs": 0.134097, "relative": 0.890416 }, { "name": "video_stereo_right", "l2": 6.742331, "mean_abs": 0.166498, "relative": 0.950527 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.525335, "mean_abs": 0.110267, "relative": 0.90378 }, { "name": "caption_objects_interaction_text", "l2": 1.95185, "mean_abs": 0.014784, "relative": 0.896242 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 23, "start_frame": 115, "end_frame": 134, "center_frame": 124, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.995994, "mean_abs": 0.243142, "relative": 0.843205 }, { "name": "hand_right_joints", "l2": 9.882475, "mean_abs": 0.243904, "relative": 0.779003 }, { "name": "body_joints", "l2": 6.382695, "mean_abs": 0.109536, "relative": 0.862105 }, { "name": "body_contacts", "l2": 3.463779, "mean_abs": 0.081749, "relative": 0.999532 }, { "name": "camera_translation", "l2": 0.309958, "mean_abs": 0.036722, "relative": 0.731457 }, { "name": "camera_rotation_matrix", "l2": 3.25644, "mean_abs": 0.270478, "relative": 1.0 }, { "name": "imu_accel_gyro", "l2": 17.534552, "mean_abs": 1.102238, "relative": 0.946049 }, { "name": "depth_confidence", "l2": 12.773782, "mean_abs": 0.235795, "relative": 0.000424 }, { "name": "video_fisheye_cam0", "l2": 5.229254, "mean_abs": 0.116725, "relative": 0.946906 }, { "name": "video_fisheye_cam1", "l2": 6.047579, "mean_abs": 0.136409, "relative": 0.999311 }, { "name": "video_fisheye_cam2", "l2": 5.306391, "mean_abs": 0.121688, "relative": 0.847086 }, { "name": "video_fisheye_cam3", "l2": 5.657567, "mean_abs": 0.124842, "relative": 0.947291 }, { "name": "video_stereo_left", "l2": 5.472551, "mean_abs": 0.135458, "relative": 0.8893 }, { "name": "video_stereo_right", "l2": 6.797644, "mean_abs": 0.166678, "relative": 0.958325 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.686971, "mean_abs": 0.113657, "relative": 0.961626 }, { "name": "caption_objects_interaction_text", "l2": 1.951849, "mean_abs": 0.014421, "relative": 0.896242 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 24, "start_frame": 120, "end_frame": 139, "center_frame": 129, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.901276, "mean_abs": 0.241353, "relative": 0.835215 }, { "name": "hand_right_joints", "l2": 10.125581, "mean_abs": 0.268218, "relative": 0.798167 }, { "name": "body_joints", "l2": 6.934552, "mean_abs": 0.118289, "relative": 0.936644 }, { "name": "body_contacts", "l2": 3.463718, "mean_abs": 0.081749, "relative": 0.999514 }, { "name": "camera_translation", "l2": 0.369316, "mean_abs": 0.042524, "relative": 0.871533 }, { "name": "camera_rotation_matrix", "l2": 3.158904, "mean_abs": 0.260463, "relative": 0.970048 }, { "name": "imu_accel_gyro", "l2": 17.354315, "mean_abs": 1.013329, "relative": 0.936325 }, { "name": "depth_confidence", "l2": 12.618315, "mean_abs": 0.233672, "relative": 0.000419 }, { "name": "video_fisheye_cam0", "l2": 5.0763, "mean_abs": 0.11169, "relative": 0.919209 }, { "name": "video_fisheye_cam1", "l2": 6.051746, "mean_abs": 0.134724, "relative": 1.0 }, { "name": "video_fisheye_cam2", "l2": 5.094668, "mean_abs": 0.116155, "relative": 0.813288 }, { "name": "video_fisheye_cam3", "l2": 5.663908, "mean_abs": 0.122639, "relative": 0.948353 }, { "name": "video_stereo_left", "l2": 5.401968, "mean_abs": 0.134078, "relative": 0.87783 }, { "name": "video_stereo_right", "l2": 6.72901, "mean_abs": 0.165667, "relative": 0.948649 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.473328, "mean_abs": 0.101952, "relative": 0.885167 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 25, "start_frame": 125, "end_frame": 144, "center_frame": 134, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.64481, "mean_abs": 0.225454, "relative": 0.813581 }, { "name": "hand_right_joints", "l2": 10.681629, "mean_abs": 0.302452, "relative": 0.841998 }, { "name": "body_joints", "l2": 7.21771, "mean_abs": 0.120524, "relative": 0.97489 }, { "name": "body_contacts", "l2": 3.4637, "mean_abs": 0.081713, "relative": 0.999509 }, { "name": "camera_translation", "l2": 0.384145, "mean_abs": 0.043295, "relative": 0.906527 }, { "name": "camera_rotation_matrix", "l2": 3.056166, "mean_abs": 0.251821, "relative": 0.938499 }, { "name": "imu_accel_gyro", "l2": 17.365778, "mean_abs": 0.95282, "relative": 0.936944 }, { "name": "depth_confidence", "l2": 12.5885, "mean_abs": 0.229218, "relative": 0.000418 }, { "name": "video_fisheye_cam0", "l2": 4.881556, "mean_abs": 0.105714, "relative": 0.883945 }, { "name": "video_fisheye_cam1", "l2": 5.982062, "mean_abs": 0.130236, "relative": 0.988485 }, { "name": "video_fisheye_cam2", "l2": 4.944398, "mean_abs": 0.111947, "relative": 0.7893 }, { "name": "video_fisheye_cam3", "l2": 5.611806, "mean_abs": 0.118546, "relative": 0.939629 }, { "name": "video_stereo_left", "l2": 5.386546, "mean_abs": 0.133887, "relative": 0.875324 }, { "name": "video_stereo_right", "l2": 6.500545, "mean_abs": 0.155954, "relative": 0.91644 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.228328, "mean_abs": 0.088849, "relative": 0.797485 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 26, "start_frame": 130, "end_frame": 149, "center_frame": 139, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.512287, "mean_abs": 0.232626, "relative": 0.802402 }, { "name": "hand_right_joints", "l2": 10.910798, "mean_abs": 0.312075, "relative": 0.860063 }, { "name": "body_joints", "l2": 7.149855, "mean_abs": 0.119952, "relative": 0.965725 }, { "name": "body_contacts", "l2": 3.463485, "mean_abs": 0.081695, "relative": 0.999447 }, { "name": "camera_translation", "l2": 0.39431, "mean_abs": 0.044019, "relative": 0.930517 }, { "name": "camera_rotation_matrix", "l2": 3.015839, "mean_abs": 0.244579, "relative": 0.926115 }, { "name": "imu_accel_gyro", "l2": 17.284691, "mean_abs": 0.909775, "relative": 0.932569 }, { "name": "depth_confidence", "l2": 12.077991, "mean_abs": 0.216397, "relative": 0.000401 }, { "name": "video_fisheye_cam0", "l2": 4.800015, "mean_abs": 0.102746, "relative": 0.86918 }, { "name": "video_fisheye_cam1", "l2": 5.870772, "mean_abs": 0.12815, "relative": 0.970096 }, { "name": "video_fisheye_cam2", "l2": 4.970387, "mean_abs": 0.113226, "relative": 0.793448 }, { "name": "video_fisheye_cam3", "l2": 5.587501, "mean_abs": 0.117395, "relative": 0.93556 }, { "name": "video_stereo_left", "l2": 5.537921, "mean_abs": 0.135908, "relative": 0.899923 }, { "name": "video_stereo_right", "l2": 6.428738, "mean_abs": 0.154718, "relative": 0.906317 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.111016, "mean_abs": 0.082153, "relative": 0.755501 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 27, "start_frame": 135, "end_frame": 154, "center_frame": 144, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.493708, "mean_abs": 0.236067, "relative": 0.800835 }, { "name": "hand_right_joints", "l2": 11.204198, "mean_abs": 0.316647, "relative": 0.883191 }, { "name": "body_joints", "l2": 7.217119, "mean_abs": 0.120476, "relative": 0.97481 }, { "name": "body_contacts", "l2": 3.463652, "mean_abs": 0.081744, "relative": 0.999495 }, { "name": "camera_translation", "l2": 0.383699, "mean_abs": 0.043053, "relative": 0.905476 }, { "name": "camera_rotation_matrix", "l2": 3.011836, "mean_abs": 0.243639, "relative": 0.924886 }, { "name": "imu_accel_gyro", "l2": 17.2288, "mean_abs": 0.876344, "relative": 0.929553 }, { "name": "depth_confidence", "l2": 11.912004, "mean_abs": 0.210915, "relative": 0.000395 }, { "name": "video_fisheye_cam0", "l2": 4.826288, "mean_abs": 0.104496, "relative": 0.873938 }, { "name": "video_fisheye_cam1", "l2": 5.717965, "mean_abs": 0.124541, "relative": 0.944845 }, { "name": "video_fisheye_cam2", "l2": 5.03844, "mean_abs": 0.114763, "relative": 0.804312 }, { "name": "video_fisheye_cam3", "l2": 5.569592, "mean_abs": 0.117015, "relative": 0.932561 }, { "name": "video_stereo_left", "l2": 5.603723, "mean_abs": 0.13747, "relative": 0.910616 }, { "name": "video_stereo_right", "l2": 6.266013, "mean_abs": 0.150879, "relative": 0.883376 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.174993, "mean_abs": 0.085283, "relative": 0.778397 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 28, "start_frame": 140, "end_frame": 159, "center_frame": 149, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.522758, "mean_abs": 0.241969, "relative": 0.803285 }, { "name": "hand_right_joints", "l2": 11.647598, "mean_abs": 0.322597, "relative": 0.918142 }, { "name": "body_joints", "l2": 7.362207, "mean_abs": 0.11875, "relative": 0.994407 }, { "name": "body_contacts", "l2": 3.463742, "mean_abs": 0.081747, "relative": 0.999521 }, { "name": "camera_translation", "l2": 0.3291, "mean_abs": 0.036851, "relative": 0.776629 }, { "name": "camera_rotation_matrix", "l2": 3.006932, "mean_abs": 0.240914, "relative": 0.92338 }, { "name": "imu_accel_gyro", "l2": 17.262323, "mean_abs": 0.899236, "relative": 0.931362 }, { "name": "depth_confidence", "l2": 11.755287, "mean_abs": 0.206656, "relative": 0.00039 }, { "name": "video_fisheye_cam0", "l2": 4.888353, "mean_abs": 0.106385, "relative": 0.885176 }, { "name": "video_fisheye_cam1", "l2": 5.515637, "mean_abs": 0.119162, "relative": 0.911412 }, { "name": "video_fisheye_cam2", "l2": 5.146887, "mean_abs": 0.115885, "relative": 0.821624 }, { "name": "video_fisheye_cam3", "l2": 5.521013, "mean_abs": 0.115413, "relative": 0.924427 }, { "name": "video_stereo_left", "l2": 5.685675, "mean_abs": 0.136532, "relative": 0.923934 }, { "name": "video_stereo_right", "l2": 6.06217, "mean_abs": 0.146085, "relative": 0.854639 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.141195, "mean_abs": 0.085677, "relative": 0.766302 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 29, "start_frame": 145, "end_frame": 164, "center_frame": 154, "action": "Pick up kettle", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.517316, "mean_abs": 0.24374, "relative": 0.802826 }, { "name": "hand_right_joints", "l2": 11.914192, "mean_abs": 0.325336, "relative": 0.939157 }, { "name": "body_joints", "l2": 7.403613, "mean_abs": 0.117001, "relative": 1.0 }, { "name": "body_contacts", "l2": 3.463938, "mean_abs": 0.081742, "relative": 0.999578 }, { "name": "camera_translation", "l2": 0.239958, "mean_abs": 0.025919, "relative": 0.566267 }, { "name": "camera_rotation_matrix", "l2": 3.000775, "mean_abs": 0.236226, "relative": 0.921489 }, { "name": "imu_accel_gyro", "l2": 17.282314, "mean_abs": 0.905882, "relative": 0.93244 }, { "name": "depth_confidence", "l2": 11.529029, "mean_abs": 0.198948, "relative": 0.000383 }, { "name": "video_fisheye_cam0", "l2": 4.966702, "mean_abs": 0.107663, "relative": 0.899364 }, { "name": "video_fisheye_cam1", "l2": 5.426962, "mean_abs": 0.115706, "relative": 0.89676 }, { "name": "video_fisheye_cam2", "l2": 5.212072, "mean_abs": 0.114983, "relative": 0.83203 }, { "name": "video_fisheye_cam3", "l2": 5.510688, "mean_abs": 0.114499, "relative": 0.922698 }, { "name": "video_stereo_left", "l2": 5.707816, "mean_abs": 0.134387, "relative": 0.927531 }, { "name": "video_stereo_right", "l2": 5.91888, "mean_abs": 0.139411, "relative": 0.834438 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.035991, "mean_abs": 0.080639, "relative": 0.728651 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 30, "start_frame": 150, "end_frame": 169, "center_frame": 159, "action": "", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.545423, "mean_abs": 0.244162, "relative": 0.805197 }, { "name": "hand_right_joints", "l2": 12.025564, "mean_abs": 0.327915, "relative": 0.947936 }, { "name": "body_joints", "l2": 7.31803, "mean_abs": 0.11324, "relative": 0.98844 }, { "name": "body_contacts", "l2": 3.464068, "mean_abs": 0.081726, "relative": 0.999615 }, { "name": "camera_translation", "l2": 0.134387, "mean_abs": 0.01593, "relative": 0.317135 }, { "name": "camera_rotation_matrix", "l2": 3.004349, "mean_abs": 0.238054, "relative": 0.922587 }, { "name": "imu_accel_gyro", "l2": 17.115383, "mean_abs": 0.862468, "relative": 0.923434 }, { "name": "depth_confidence", "l2": 11.51034, "mean_abs": 0.198648, "relative": 0.000382 }, { "name": "video_fisheye_cam0", "l2": 5.060183, "mean_abs": 0.108939, "relative": 0.916291 }, { "name": "video_fisheye_cam1", "l2": 5.381006, "mean_abs": 0.115177, "relative": 0.889166 }, { "name": "video_fisheye_cam2", "l2": 5.344055, "mean_abs": 0.11719, "relative": 0.853099 }, { "name": "video_fisheye_cam3", "l2": 5.467731, "mean_abs": 0.113737, "relative": 0.915506 }, { "name": "video_stereo_left", "l2": 5.763088, "mean_abs": 0.135285, "relative": 0.936513 }, { "name": "video_stereo_right", "l2": 5.847844, "mean_abs": 0.138136, "relative": 0.824423 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.14379, "mean_abs": 0.085574, "relative": 0.76723 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 31, "start_frame": 155, "end_frame": 174, "center_frame": 164, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.637073, "mean_abs": 0.244995, "relative": 0.812928 }, { "name": "hand_right_joints", "l2": 12.150162, "mean_abs": 0.332064, "relative": 0.957758 }, { "name": "body_joints", "l2": 7.265843, "mean_abs": 0.110985, "relative": 0.981391 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.048034, "mean_abs": 0.006955, "relative": 0.113354 }, { "name": "camera_rotation_matrix", "l2": 3.012181, "mean_abs": 0.243775, "relative": 0.924992 }, { "name": "imu_accel_gyro", "l2": 17.107533, "mean_abs": 0.855885, "relative": 0.92301 }, { "name": "depth_confidence", "l2": 11.650881, "mean_abs": 0.205241, "relative": 0.000387 }, { "name": "video_fisheye_cam0", "l2": 5.172779, "mean_abs": 0.11109, "relative": 0.93668 }, { "name": "video_fisheye_cam1", "l2": 5.396724, "mean_abs": 0.117173, "relative": 0.891763 }, { "name": "video_fisheye_cam2", "l2": 5.463796, "mean_abs": 0.120765, "relative": 0.872214 }, { "name": "video_fisheye_cam3", "l2": 5.41239, "mean_abs": 0.11323, "relative": 0.906239 }, { "name": "video_stereo_left", "l2": 5.811784, "mean_abs": 0.139276, "relative": 0.944426 }, { "name": "video_stereo_right", "l2": 5.879874, "mean_abs": 0.141518, "relative": 0.828939 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.396557, "mean_abs": 0.098377, "relative": 0.857692 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 32, "start_frame": 160, "end_frame": 179, "center_frame": 169, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.712787, "mean_abs": 0.246425, "relative": 0.819315 }, { "name": "hand_right_joints", "l2": 12.200691, "mean_abs": 0.331677, "relative": 0.961741 }, { "name": "body_joints", "l2": 7.226985, "mean_abs": 0.110362, "relative": 0.976143 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.045143, "mean_abs": 0.006368, "relative": 0.106532 }, { "name": "camera_rotation_matrix", "l2": 3.01401, "mean_abs": 0.243771, "relative": 0.925553 }, { "name": "imu_accel_gyro", "l2": 17.174154, "mean_abs": 0.851577, "relative": 0.926605 }, { "name": "depth_confidence", "l2": 12.671931, "mean_abs": 0.212265, "relative": 0.000421 }, { "name": "video_fisheye_cam0", "l2": 5.222461, "mean_abs": 0.112369, "relative": 0.945676 }, { "name": "video_fisheye_cam1", "l2": 5.478796, "mean_abs": 0.120172, "relative": 0.905325 }, { "name": "video_fisheye_cam2", "l2": 5.537065, "mean_abs": 0.122627, "relative": 0.88391 }, { "name": "video_fisheye_cam3", "l2": 5.366605, "mean_abs": 0.112184, "relative": 0.898573 }, { "name": "video_stereo_left", "l2": 5.868729, "mean_abs": 0.141357, "relative": 0.95368 }, { "name": "video_stereo_right", "l2": 5.972583, "mean_abs": 0.145445, "relative": 0.842009 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.315637, "mean_abs": 0.099613, "relative": 0.828732 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 33, "start_frame": 165, "end_frame": 184, "center_frame": 174, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.764066, "mean_abs": 0.245574, "relative": 0.823641 }, { "name": "hand_right_joints", "l2": 12.172253, "mean_abs": 0.327336, "relative": 0.959499 }, { "name": "body_joints", "l2": 7.186648, "mean_abs": 0.110488, "relative": 0.970695 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.072977, "mean_abs": 0.008837, "relative": 0.172216 }, { "name": "camera_rotation_matrix", "l2": 3.012599, "mean_abs": 0.241709, "relative": 0.92512 }, { "name": "imu_accel_gyro", "l2": 17.146059, "mean_abs": 0.834663, "relative": 0.925089 }, { "name": "depth_confidence", "l2": 12.786574, "mean_abs": 0.214011, "relative": 0.000424 }, { "name": "video_fisheye_cam0", "l2": 5.25335, "mean_abs": 0.112923, "relative": 0.951269 }, { "name": "video_fisheye_cam1", "l2": 5.581403, "mean_abs": 0.122243, "relative": 0.92228 }, { "name": "video_fisheye_cam2", "l2": 5.594659, "mean_abs": 0.123156, "relative": 0.893104 }, { "name": "video_fisheye_cam3", "l2": 5.331789, "mean_abs": 0.110678, "relative": 0.892744 }, { "name": "video_stereo_left", "l2": 5.912751, "mean_abs": 0.142185, "relative": 0.960834 }, { "name": "video_stereo_right", "l2": 6.114258, "mean_abs": 0.14952, "relative": 0.861982 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.262646, "mean_abs": 0.093963, "relative": 0.809767 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 34, "start_frame": 170, "end_frame": 189, "center_frame": 179, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.774893, "mean_abs": 0.245535, "relative": 0.824554 }, { "name": "hand_right_joints", "l2": 12.044235, "mean_abs": 0.321418, "relative": 0.949408 }, { "name": "body_joints", "l2": 7.138551, "mean_abs": 0.111305, "relative": 0.964198 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.096606, "mean_abs": 0.010415, "relative": 0.227977 }, { "name": "camera_rotation_matrix", "l2": 3.013024, "mean_abs": 0.23907, "relative": 0.925251 }, { "name": "imu_accel_gyro", "l2": 17.15144, "mean_abs": 0.838575, "relative": 0.925379 }, { "name": "depth_confidence", "l2": 12.955893, "mean_abs": 0.217732, "relative": 0.00043 }, { "name": "video_fisheye_cam0", "l2": 5.254576, "mean_abs": 0.113133, "relative": 0.951491 }, { "name": "video_fisheye_cam1", "l2": 5.65586, "mean_abs": 0.122923, "relative": 0.934583 }, { "name": "video_fisheye_cam2", "l2": 5.635848, "mean_abs": 0.123564, "relative": 0.899679 }, { "name": "video_fisheye_cam3", "l2": 5.312554, "mean_abs": 0.110034, "relative": 0.889523 }, { "name": "video_stereo_left", "l2": 5.918139, "mean_abs": 0.142065, "relative": 0.961709 }, { "name": "video_stereo_right", "l2": 6.237882, "mean_abs": 0.152254, "relative": 0.879411 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.350467, "mean_abs": 0.095996, "relative": 0.841197 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 35, "start_frame": 175, "end_frame": 194, "center_frame": 184, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.741954, "mean_abs": 0.246838, "relative": 0.821775 }, { "name": "hand_right_joints", "l2": 11.85274, "mean_abs": 0.323248, "relative": 0.934313 }, { "name": "body_joints", "l2": 7.140101, "mean_abs": 0.115636, "relative": 0.964408 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.142992, "mean_abs": 0.015243, "relative": 0.337441 }, { "name": "camera_rotation_matrix", "l2": 3.026781, "mean_abs": 0.24057, "relative": 0.929475 }, { "name": "imu_accel_gyro", "l2": 17.214653, "mean_abs": 0.884773, "relative": 0.92879 }, { "name": "depth_confidence", "l2": 16.608189, "mean_abs": 0.236125, "relative": 0.000551 }, { "name": "video_fisheye_cam0", "l2": 5.246274, "mean_abs": 0.113632, "relative": 0.949988 }, { "name": "video_fisheye_cam1", "l2": 5.727883, "mean_abs": 0.125212, "relative": 0.946484 }, { "name": "video_fisheye_cam2", "l2": 5.723358, "mean_abs": 0.127021, "relative": 0.913649 }, { "name": "video_fisheye_cam3", "l2": 5.310738, "mean_abs": 0.11063, "relative": 0.889219 }, { "name": "video_stereo_left", "l2": 5.972185, "mean_abs": 0.145939, "relative": 0.970492 }, { "name": "video_stereo_right", "l2": 6.370792, "mean_abs": 0.157047, "relative": 0.898148 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.268737, "mean_abs": 0.095243, "relative": 0.811947 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 36, "start_frame": 180, "end_frame": 199, "center_frame": 189, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.789222, "mean_abs": 0.247546, "relative": 0.825763 }, { "name": "hand_right_joints", "l2": 11.563707, "mean_abs": 0.322412, "relative": 0.911529 }, { "name": "body_joints", "l2": 7.171342, "mean_abs": 0.12134, "relative": 0.968627 }, { "name": "body_contacts", "l2": 3.464426, "mean_abs": 0.081737, "relative": 0.999718 }, { "name": "camera_translation", "l2": 0.238651, "mean_abs": 0.02763, "relative": 0.563182 }, { "name": "camera_rotation_matrix", "l2": 3.103569, "mean_abs": 0.248098, "relative": 0.953056 }, { "name": "imu_accel_gyro", "l2": 17.236109, "mean_abs": 0.961881, "relative": 0.929947 }, { "name": "depth_confidence", "l2": 16.622345, "mean_abs": 0.242309, "relative": 0.000552 }, { "name": "video_fisheye_cam0", "l2": 5.355547, "mean_abs": 0.118251, "relative": 0.969775 }, { "name": "video_fisheye_cam1", "l2": 5.799086, "mean_abs": 0.129852, "relative": 0.95825 }, { "name": "video_fisheye_cam2", "l2": 5.854869, "mean_abs": 0.13231, "relative": 0.934643 }, { "name": "video_fisheye_cam3", "l2": 5.348046, "mean_abs": 0.113601, "relative": 0.895466 }, { "name": "video_stereo_left", "l2": 6.068006, "mean_abs": 0.150976, "relative": 0.986063 }, { "name": "video_stereo_right", "l2": 6.516778, "mean_abs": 0.163107, "relative": 0.918729 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.154824, "mean_abs": 0.084956, "relative": 0.771179 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 37, "start_frame": 185, "end_frame": 204, "center_frame": 194, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.867594, "mean_abs": 0.251883, "relative": 0.832374 }, { "name": "hand_right_joints", "l2": 11.236994, "mean_abs": 0.317297, "relative": 0.885776 }, { "name": "body_joints", "l2": 7.171211, "mean_abs": 0.124821, "relative": 0.96861 }, { "name": "body_contacts", "l2": 3.464347, "mean_abs": 0.081736, "relative": 0.999696 }, { "name": "camera_translation", "l2": 0.320045, "mean_abs": 0.039443, "relative": 0.755262 }, { "name": "camera_rotation_matrix", "l2": 3.173927, "mean_abs": 0.259181, "relative": 0.974661 }, { "name": "imu_accel_gyro", "l2": 17.33959, "mean_abs": 1.00408, "relative": 0.935531 }, { "name": "depth_confidence", "l2": 16.51564, "mean_abs": 0.242128, "relative": 0.000548 }, { "name": "video_fisheye_cam0", "l2": 5.398913, "mean_abs": 0.121084, "relative": 0.977628 }, { "name": "video_fisheye_cam1", "l2": 5.819796, "mean_abs": 0.13177, "relative": 0.961672 }, { "name": "video_fisheye_cam2", "l2": 5.954878, "mean_abs": 0.13536, "relative": 0.950608 }, { "name": "video_fisheye_cam3", "l2": 5.407622, "mean_abs": 0.116763, "relative": 0.905441 }, { "name": "video_stereo_left", "l2": 6.153771, "mean_abs": 0.153549, "relative": 1.0 }, { "name": "video_stereo_right", "l2": 6.679194, "mean_abs": 0.167644, "relative": 0.941626 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.177972, "mean_abs": 0.087448, "relative": 0.779463 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 38, "start_frame": 190, "end_frame": 209, "center_frame": 199, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.869316, "mean_abs": 0.247745, "relative": 0.832519 }, { "name": "hand_right_joints", "l2": 10.855374, "mean_abs": 0.306629, "relative": 0.855694 }, { "name": "body_joints", "l2": 7.062306, "mean_abs": 0.125206, "relative": 0.9539 }, { "name": "body_contacts", "l2": 3.46439, "mean_abs": 0.081727, "relative": 0.999708 }, { "name": "camera_translation", "l2": 0.367069, "mean_abs": 0.04648, "relative": 0.86623 }, { "name": "camera_rotation_matrix", "l2": 3.18218, "mean_abs": 0.253028, "relative": 0.977196 }, { "name": "imu_accel_gyro", "l2": 17.368946, "mean_abs": 1.015441, "relative": 0.937114 }, { "name": "depth_confidence", "l2": 16.467949, "mean_abs": 0.242512, "relative": 0.000547 }, { "name": "video_fisheye_cam0", "l2": 5.399897, "mean_abs": 0.120549, "relative": 0.977806 }, { "name": "video_fisheye_cam1", "l2": 5.817954, "mean_abs": 0.1315, "relative": 0.961368 }, { "name": "video_fisheye_cam2", "l2": 6.003902, "mean_abs": 0.136009, "relative": 0.958434 }, { "name": "video_fisheye_cam3", "l2": 5.496236, "mean_abs": 0.119583, "relative": 0.920278 }, { "name": "video_stereo_left", "l2": 6.135659, "mean_abs": 0.152674, "relative": 0.997057 }, { "name": "video_stereo_right", "l2": 6.726882, "mean_abs": 0.167264, "relative": 0.948349 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.138183, "mean_abs": 0.085104, "relative": 0.765223 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 39, "start_frame": 195, "end_frame": 214, "center_frame": 204, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.873807, "mean_abs": 0.240777, "relative": 0.832898 }, { "name": "hand_right_joints", "l2": 10.516901, "mean_abs": 0.29691, "relative": 0.829013 }, { "name": "body_joints", "l2": 6.952831, "mean_abs": 0.122963, "relative": 0.939113 }, { "name": "body_contacts", "l2": 3.464668, "mean_abs": 0.08172, "relative": 0.999788 }, { "name": "camera_translation", "l2": 0.377328, "mean_abs": 0.047824, "relative": 0.89044 }, { "name": "camera_rotation_matrix", "l2": 3.153633, "mean_abs": 0.247547, "relative": 0.968429 }, { "name": "imu_accel_gyro", "l2": 17.393431, "mean_abs": 1.058184, "relative": 0.938435 }, { "name": "depth_confidence", "l2": 13.059101, "mean_abs": 0.231933, "relative": 0.000433 }, { "name": "video_fisheye_cam0", "l2": 5.397734, "mean_abs": 0.118855, "relative": 0.977414 }, { "name": "video_fisheye_cam1", "l2": 5.801985, "mean_abs": 0.130901, "relative": 0.958729 }, { "name": "video_fisheye_cam2", "l2": 6.08741, "mean_abs": 0.136223, "relative": 0.971764 }, { "name": "video_fisheye_cam3", "l2": 5.611084, "mean_abs": 0.121921, "relative": 0.939508 }, { "name": "video_stereo_left", "l2": 6.079814, "mean_abs": 0.150025, "relative": 0.987982 }, { "name": "video_stereo_right", "l2": 6.787455, "mean_abs": 0.16714, "relative": 0.956889 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.19683, "mean_abs": 0.084131, "relative": 0.786213 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 40, "start_frame": 200, "end_frame": 219, "center_frame": 209, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.990927, "mean_abs": 0.23578, "relative": 0.842777 }, { "name": "hand_right_joints", "l2": 10.313119, "mean_abs": 0.291888, "relative": 0.81295 }, { "name": "body_joints", "l2": 6.798811, "mean_abs": 0.117685, "relative": 0.91831 }, { "name": "body_contacts", "l2": 3.464265, "mean_abs": 0.081731, "relative": 0.999672 }, { "name": "camera_translation", "l2": 0.324361, "mean_abs": 0.040639, "relative": 0.765446 }, { "name": "camera_rotation_matrix", "l2": 3.053323, "mean_abs": 0.228383, "relative": 0.937626 }, { "name": "imu_accel_gyro", "l2": 17.2754, "mean_abs": 1.041892, "relative": 0.932067 }, { "name": "depth_confidence", "l2": 12.688605, "mean_abs": 0.227991, "relative": 0.000421 }, { "name": "video_fisheye_cam0", "l2": 5.327301, "mean_abs": 0.116343, "relative": 0.96466 }, { "name": "video_fisheye_cam1", "l2": 5.732761, "mean_abs": 0.128348, "relative": 0.94729 }, { "name": "video_fisheye_cam2", "l2": 6.134239, "mean_abs": 0.134851, "relative": 0.97924 }, { "name": "video_fisheye_cam3", "l2": 5.667938, "mean_abs": 0.122258, "relative": 0.949028 }, { "name": "video_stereo_left", "l2": 5.994567, "mean_abs": 0.145932, "relative": 0.974129 }, { "name": "video_stereo_right", "l2": 6.752292, "mean_abs": 0.164124, "relative": 0.951931 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.209091, "mean_abs": 0.085813, "relative": 0.790601 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 41, "start_frame": 205, "end_frame": 224, "center_frame": 214, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.084253, "mean_abs": 0.237227, "relative": 0.85065 }, { "name": "hand_right_joints", "l2": 10.229318, "mean_abs": 0.2886, "relative": 0.806344 }, { "name": "body_joints", "l2": 6.64996, "mean_abs": 0.112906, "relative": 0.898205 }, { "name": "body_contacts", "l2": 3.464381, "mean_abs": 0.081722, "relative": 0.999705 }, { "name": "camera_translation", "l2": 0.282743, "mean_abs": 0.035275, "relative": 0.667234 }, { "name": "camera_rotation_matrix", "l2": 3.027715, "mean_abs": 0.227982, "relative": 0.929762 }, { "name": "imu_accel_gyro", "l2": 17.15423, "mean_abs": 0.996437, "relative": 0.92553 }, { "name": "depth_confidence", "l2": 11.964319, "mean_abs": 0.221133, "relative": 0.000397 }, { "name": "video_fisheye_cam0", "l2": 5.334301, "mean_abs": 0.11529, "relative": 0.965928 }, { "name": "video_fisheye_cam1", "l2": 5.723415, "mean_abs": 0.12719, "relative": 0.945746 }, { "name": "video_fisheye_cam2", "l2": 6.09985, "mean_abs": 0.131853, "relative": 0.97375 }, { "name": "video_fisheye_cam3", "l2": 5.743546, "mean_abs": 0.121877, "relative": 0.961688 }, { "name": "video_stereo_left", "l2": 5.938808, "mean_abs": 0.141858, "relative": 0.965068 }, { "name": "video_stereo_right", "l2": 6.772438, "mean_abs": 0.162487, "relative": 0.954772 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.188595, "mean_abs": 0.084531, "relative": 0.783265 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 42, "start_frame": 210, "end_frame": 229, "center_frame": 219, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.142739, "mean_abs": 0.235909, "relative": 0.855583 }, { "name": "hand_right_joints", "l2": 10.139879, "mean_abs": 0.279967, "relative": 0.799294 }, { "name": "body_joints", "l2": 6.507735, "mean_abs": 0.108013, "relative": 0.878994 }, { "name": "body_contacts", "l2": 3.464405, "mean_abs": 0.081715, "relative": 0.999712 }, { "name": "camera_translation", "l2": 0.241236, "mean_abs": 0.030307, "relative": 0.569284 }, { "name": "camera_rotation_matrix", "l2": 3.019768, "mean_abs": 0.227405, "relative": 0.927322 }, { "name": "imu_accel_gyro", "l2": 17.093269, "mean_abs": 0.999788, "relative": 0.922241 }, { "name": "depth_confidence", "l2": 11.892011, "mean_abs": 0.217836, "relative": 0.000395 }, { "name": "video_fisheye_cam0", "l2": 5.338962, "mean_abs": 0.113889, "relative": 0.966772 }, { "name": "video_fisheye_cam1", "l2": 5.706186, "mean_abs": 0.125487, "relative": 0.942899 }, { "name": "video_fisheye_cam2", "l2": 6.019257, "mean_abs": 0.128393, "relative": 0.960885 }, { "name": "video_fisheye_cam3", "l2": 5.801113, "mean_abs": 0.121908, "relative": 0.971326 }, { "name": "video_stereo_left", "l2": 5.894741, "mean_abs": 0.139887, "relative": 0.957907 }, { "name": "video_stereo_right", "l2": 6.717358, "mean_abs": 0.158043, "relative": 0.947006 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.267968, "mean_abs": 0.089364, "relative": 0.811672 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 43, "start_frame": 215, "end_frame": 234, "center_frame": 224, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.214542, "mean_abs": 0.236497, "relative": 0.86164 }, { "name": "hand_right_joints", "l2": 10.061892, "mean_abs": 0.273443, "relative": 0.793146 }, { "name": "body_joints", "l2": 6.39587, "mean_abs": 0.103589, "relative": 0.863885 }, { "name": "body_contacts", "l2": 3.464437, "mean_abs": 0.081709, "relative": 0.999722 }, { "name": "camera_translation", "l2": 0.178601, "mean_abs": 0.022543, "relative": 0.421474 }, { "name": "camera_rotation_matrix", "l2": 3.008799, "mean_abs": 0.224886, "relative": 0.923953 }, { "name": "imu_accel_gyro", "l2": 17.16931, "mean_abs": 1.035504, "relative": 0.926343 }, { "name": "depth_confidence", "l2": 11.789493, "mean_abs": 0.21323, "relative": 0.000391 }, { "name": "video_fisheye_cam0", "l2": 5.354587, "mean_abs": 0.11331, "relative": 0.969601 }, { "name": "video_fisheye_cam1", "l2": 5.64959, "mean_abs": 0.122631, "relative": 0.933547 }, { "name": "video_fisheye_cam2", "l2": 5.978529, "mean_abs": 0.126519, "relative": 0.954383 }, { "name": "video_fisheye_cam3", "l2": 5.816206, "mean_abs": 0.120777, "relative": 0.973853 }, { "name": "video_stereo_left", "l2": 5.825469, "mean_abs": 0.136558, "relative": 0.94665 }, { "name": "video_stereo_right", "l2": 6.682828, "mean_abs": 0.154974, "relative": 0.942139 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.599753, "mean_abs": 0.115005, "relative": 0.930413 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 44, "start_frame": 220, "end_frame": 239, "center_frame": 229, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.298186, "mean_abs": 0.237431, "relative": 0.868696 }, { "name": "hand_right_joints", "l2": 10.070559, "mean_abs": 0.268912, "relative": 0.793829 }, { "name": "body_joints", "l2": 6.331341, "mean_abs": 0.100008, "relative": 0.855169 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.126241, "mean_abs": 0.016138, "relative": 0.29791 }, { "name": "camera_rotation_matrix", "l2": 3.005438, "mean_abs": 0.222893, "relative": 0.922921 }, { "name": "imu_accel_gyro", "l2": 17.080551, "mean_abs": 1.019723, "relative": 0.921555 }, { "name": "depth_confidence", "l2": 11.67424, "mean_abs": 0.208767, "relative": 0.000387 }, { "name": "video_fisheye_cam0", "l2": 5.37089, "mean_abs": 0.113042, "relative": 0.972553 }, { "name": "video_fisheye_cam1", "l2": 5.610649, "mean_abs": 0.120414, "relative": 0.927112 }, { "name": "video_fisheye_cam2", "l2": 5.916366, "mean_abs": 0.124232, "relative": 0.94446 }, { "name": "video_fisheye_cam3", "l2": 5.805039, "mean_abs": 0.119974, "relative": 0.971984 }, { "name": "video_stereo_left", "l2": 5.745355, "mean_abs": 0.131793, "relative": 0.933632 }, { "name": "video_stereo_right", "l2": 6.636735, "mean_abs": 0.15133, "relative": 0.93564 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.443834, "mean_abs": 0.102286, "relative": 0.874611 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 45, "start_frame": 225, "end_frame": 244, "center_frame": 234, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.396669, "mean_abs": 0.23868, "relative": 0.877003 }, { "name": "hand_right_joints", "l2": 10.120067, "mean_abs": 0.266608, "relative": 0.797732 }, { "name": "body_joints", "l2": 6.28361, "mean_abs": 0.097276, "relative": 0.848722 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.090908, "mean_abs": 0.011933, "relative": 0.21453 }, { "name": "camera_rotation_matrix", "l2": 3.004107, "mean_abs": 0.221273, "relative": 0.922513 }, { "name": "imu_accel_gyro", "l2": 17.076468, "mean_abs": 1.030576, "relative": 0.921334 }, { "name": "depth_confidence", "l2": 11.533897, "mean_abs": 0.207085, "relative": 0.000383 }, { "name": "video_fisheye_cam0", "l2": 5.395923, "mean_abs": 0.113062, "relative": 0.977086 }, { "name": "video_fisheye_cam1", "l2": 5.592254, "mean_abs": 0.119142, "relative": 0.924073 }, { "name": "video_fisheye_cam2", "l2": 5.843744, "mean_abs": 0.122115, "relative": 0.932867 }, { "name": "video_fisheye_cam3", "l2": 5.803596, "mean_abs": 0.119127, "relative": 0.971742 }, { "name": "video_stereo_left", "l2": 5.701601, "mean_abs": 0.128917, "relative": 0.926521 }, { "name": "video_stereo_right", "l2": 6.613397, "mean_abs": 0.149199, "relative": 0.93235 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.436388, "mean_abs": 0.10183, "relative": 0.871947 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 46, "start_frame": 230, "end_frame": 249, "center_frame": 239, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.435653, "mean_abs": 0.239415, "relative": 0.880292 }, { "name": "hand_right_joints", "l2": 10.145331, "mean_abs": 0.263739, "relative": 0.799724 }, { "name": "body_joints", "l2": 6.2354, "mean_abs": 0.094952, "relative": 0.84221 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.061663, "mean_abs": 0.008534, "relative": 0.145516 }, { "name": "camera_rotation_matrix", "l2": 3.00273, "mean_abs": 0.219348, "relative": 0.92209 }, { "name": "imu_accel_gyro", "l2": 17.068069, "mean_abs": 1.036714, "relative": 0.920881 }, { "name": "depth_confidence", "l2": 11.481098, "mean_abs": 0.205476, "relative": 0.000381 }, { "name": "video_fisheye_cam0", "l2": 5.379795, "mean_abs": 0.112482, "relative": 0.974166 }, { "name": "video_fisheye_cam1", "l2": 5.571832, "mean_abs": 0.117948, "relative": 0.920698 }, { "name": "video_fisheye_cam2", "l2": 5.770789, "mean_abs": 0.120352, "relative": 0.921221 }, { "name": "video_fisheye_cam3", "l2": 5.774413, "mean_abs": 0.117732, "relative": 0.966856 }, { "name": "video_stereo_left", "l2": 5.653144, "mean_abs": 0.126505, "relative": 0.918647 }, { "name": "video_stereo_right", "l2": 6.605373, "mean_abs": 0.14748, "relative": 0.931219 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.341632, "mean_abs": 0.097884, "relative": 0.838035 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 47, "start_frame": 235, "end_frame": 254, "center_frame": 244, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.47239, "mean_abs": 0.240173, "relative": 0.883391 }, { "name": "hand_right_joints", "l2": 10.152325, "mean_abs": 0.261571, "relative": 0.800275 }, { "name": "body_joints", "l2": 6.208699, "mean_abs": 0.093712, "relative": 0.838604 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.041134, "mean_abs": 0.006119, "relative": 0.09707 }, { "name": "camera_rotation_matrix", "l2": 3.001349, "mean_abs": 0.217147, "relative": 0.921665 }, { "name": "imu_accel_gyro", "l2": 17.073933, "mean_abs": 1.033144, "relative": 0.921197 }, { "name": "depth_confidence", "l2": 11.33748, "mean_abs": 0.202531, "relative": 0.000376 }, { "name": "video_fisheye_cam0", "l2": 5.377246, "mean_abs": 0.112267, "relative": 0.973704 }, { "name": "video_fisheye_cam1", "l2": 5.563314, "mean_abs": 0.117309, "relative": 0.919291 }, { "name": "video_fisheye_cam2", "l2": 5.720545, "mean_abs": 0.118975, "relative": 0.9132 }, { "name": "video_fisheye_cam3", "l2": 5.766847, "mean_abs": 0.117503, "relative": 0.965589 }, { "name": "video_stereo_left", "l2": 5.628926, "mean_abs": 0.125091, "relative": 0.914712 }, { "name": "video_stereo_right", "l2": 6.610169, "mean_abs": 0.146295, "relative": 0.931895 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.167618, "mean_abs": 0.091207, "relative": 0.775758 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 48, "start_frame": 240, "end_frame": 259, "center_frame": 249, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.481889, "mean_abs": 0.239694, "relative": 0.884192 }, { "name": "hand_right_joints", "l2": 10.169417, "mean_abs": 0.260669, "relative": 0.801622 }, { "name": "body_joints", "l2": 6.188647, "mean_abs": 0.092428, "relative": 0.835895 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028806, "mean_abs": 0.004349, "relative": 0.067978 }, { "name": "camera_rotation_matrix", "l2": 3.000629, "mean_abs": 0.215309, "relative": 0.921445 }, { "name": "imu_accel_gyro", "l2": 17.053488, "mean_abs": 1.032524, "relative": 0.920094 }, { "name": "depth_confidence", "l2": 11.239372, "mean_abs": 0.200995, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.372251, "mean_abs": 0.111894, "relative": 0.9728 }, { "name": "video_fisheye_cam1", "l2": 5.568218, "mean_abs": 0.11691, "relative": 0.920101 }, { "name": "video_fisheye_cam2", "l2": 5.691351, "mean_abs": 0.117842, "relative": 0.90854 }, { "name": "video_fisheye_cam3", "l2": 5.739689, "mean_abs": 0.116145, "relative": 0.961042 }, { "name": "video_stereo_left", "l2": 5.617539, "mean_abs": 0.124234, "relative": 0.912861 }, { "name": "video_stereo_right", "l2": 6.627856, "mean_abs": 0.145574, "relative": 0.934389 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.113488, "mean_abs": 0.084419, "relative": 0.756386 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 49, "start_frame": 245, "end_frame": 264, "center_frame": 254, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.484982, "mean_abs": 0.239961, "relative": 0.884453 }, { "name": "hand_right_joints", "l2": 10.171444, "mean_abs": 0.260325, "relative": 0.801782 }, { "name": "body_joints", "l2": 6.195982, "mean_abs": 0.092624, "relative": 0.836886 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017401, "mean_abs": 0.002655, "relative": 0.041064 }, { "name": "camera_rotation_matrix", "l2": 3.00021, "mean_abs": 0.213496, "relative": 0.921316 }, { "name": "imu_accel_gyro", "l2": 17.082481, "mean_abs": 1.033319, "relative": 0.921659 }, { "name": "depth_confidence", "l2": 11.224269, "mean_abs": 0.199212, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.34387, "mean_abs": 0.110515, "relative": 0.96766 }, { "name": "video_fisheye_cam1", "l2": 5.562467, "mean_abs": 0.116483, "relative": 0.919151 }, { "name": "video_fisheye_cam2", "l2": 5.666872, "mean_abs": 0.116681, "relative": 0.904632 }, { "name": "video_fisheye_cam3", "l2": 5.720781, "mean_abs": 0.115308, "relative": 0.957876 }, { "name": "video_stereo_left", "l2": 5.604129, "mean_abs": 0.12294, "relative": 0.910682 }, { "name": "video_stereo_right", "l2": 6.628047, "mean_abs": 0.144211, "relative": 0.934416 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.119998, "mean_abs": 0.084464, "relative": 0.758715 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 50, "start_frame": 250, "end_frame": 269, "center_frame": 259, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.505229, "mean_abs": 0.242631, "relative": 0.886161 }, { "name": "hand_right_joints", "l2": 10.183433, "mean_abs": 0.262373, "relative": 0.802727 }, { "name": "body_joints", "l2": 6.216435, "mean_abs": 0.093391, "relative": 0.839649 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007802, "mean_abs": 0.001126, "relative": 0.018411 }, { "name": "camera_rotation_matrix", "l2": 3.000327, "mean_abs": 0.211986, "relative": 0.921352 }, { "name": "imu_accel_gyro", "l2": 17.057438, "mean_abs": 1.026299, "relative": 0.920308 }, { "name": "depth_confidence", "l2": 11.198586, "mean_abs": 0.198645, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.342137, "mean_abs": 0.110273, "relative": 0.967347 }, { "name": "video_fisheye_cam1", "l2": 5.543687, "mean_abs": 0.115937, "relative": 0.916047 }, { "name": "video_fisheye_cam2", "l2": 5.6566, "mean_abs": 0.116133, "relative": 0.902992 }, { "name": "video_fisheye_cam3", "l2": 5.703066, "mean_abs": 0.114932, "relative": 0.95491 }, { "name": "video_stereo_left", "l2": 5.606512, "mean_abs": 0.12229, "relative": 0.911069 }, { "name": "video_stereo_right", "l2": 6.612843, "mean_abs": 0.143081, "relative": 0.932272 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.179594, "mean_abs": 0.086167, "relative": 0.780044 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 51, "start_frame": 255, "end_frame": 274, "center_frame": 264, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.509935, "mean_abs": 0.242133, "relative": 0.886558 }, { "name": "hand_right_joints", "l2": 10.17206, "mean_abs": 0.263009, "relative": 0.80183 }, { "name": "body_joints", "l2": 6.238784, "mean_abs": 0.093917, "relative": 0.842667 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005044, "mean_abs": 0.000707, "relative": 0.011902 }, { "name": "camera_rotation_matrix", "l2": 3.000342, "mean_abs": 0.211752, "relative": 0.921356 }, { "name": "imu_accel_gyro", "l2": 17.036463, "mean_abs": 1.021664, "relative": 0.919176 }, { "name": "depth_confidence", "l2": 11.196754, "mean_abs": 0.197959, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.346163, "mean_abs": 0.110245, "relative": 0.968076 }, { "name": "video_fisheye_cam1", "l2": 5.53931, "mean_abs": 0.115837, "relative": 0.915324 }, { "name": "video_fisheye_cam2", "l2": 5.664237, "mean_abs": 0.115974, "relative": 0.904211 }, { "name": "video_fisheye_cam3", "l2": 5.700902, "mean_abs": 0.11469, "relative": 0.954547 }, { "name": "video_stereo_left", "l2": 5.622946, "mean_abs": 0.122045, "relative": 0.91374 }, { "name": "video_stereo_right", "l2": 6.609012, "mean_abs": 0.142933, "relative": 0.931732 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.085315, "mean_abs": 0.08242, "relative": 0.746303 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 52, "start_frame": 260, "end_frame": 279, "center_frame": 269, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.510965, "mean_abs": 0.243358, "relative": 0.886645 }, { "name": "hand_right_joints", "l2": 10.177813, "mean_abs": 0.265359, "relative": 0.802284 }, { "name": "body_joints", "l2": 6.285186, "mean_abs": 0.094675, "relative": 0.848935 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005305, "mean_abs": 0.000694, "relative": 0.01252 }, { "name": "camera_rotation_matrix", "l2": 3.000151, "mean_abs": 0.2121, "relative": 0.921298 }, { "name": "imu_accel_gyro", "l2": 17.042133, "mean_abs": 1.026156, "relative": 0.919482 }, { "name": "depth_confidence", "l2": 11.167001, "mean_abs": 0.198418, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.349638, "mean_abs": 0.110476, "relative": 0.968705 }, { "name": "video_fisheye_cam1", "l2": 5.53016, "mean_abs": 0.115697, "relative": 0.913812 }, { "name": "video_fisheye_cam2", "l2": 5.673349, "mean_abs": 0.116196, "relative": 0.905666 }, { "name": "video_fisheye_cam3", "l2": 5.69156, "mean_abs": 0.114862, "relative": 0.952983 }, { "name": "video_stereo_left", "l2": 5.628578, "mean_abs": 0.122432, "relative": 0.914655 }, { "name": "video_stereo_right", "l2": 6.600937, "mean_abs": 0.143008, "relative": 0.930594 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.150891, "mean_abs": 0.086081, "relative": 0.769772 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 53, "start_frame": 265, "end_frame": 284, "center_frame": 274, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.514005, "mean_abs": 0.243748, "relative": 0.886901 }, { "name": "hand_right_joints", "l2": 10.179921, "mean_abs": 0.266494, "relative": 0.80245 }, { "name": "body_joints", "l2": 6.311246, "mean_abs": 0.094606, "relative": 0.852455 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01131, "mean_abs": 0.001178, "relative": 0.026689 }, { "name": "camera_rotation_matrix", "l2": 3.000406, "mean_abs": 0.212474, "relative": 0.921376 }, { "name": "imu_accel_gyro", "l2": 17.040142, "mean_abs": 1.018604, "relative": 0.919374 }, { "name": "depth_confidence", "l2": 11.153764, "mean_abs": 0.197876, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.350572, "mean_abs": 0.110467, "relative": 0.968874 }, { "name": "video_fisheye_cam1", "l2": 5.518338, "mean_abs": 0.115545, "relative": 0.911859 }, { "name": "video_fisheye_cam2", "l2": 5.677025, "mean_abs": 0.116388, "relative": 0.906253 }, { "name": "video_fisheye_cam3", "l2": 5.687018, "mean_abs": 0.114858, "relative": 0.952223 }, { "name": "video_stereo_left", "l2": 5.632075, "mean_abs": 0.122825, "relative": 0.915223 }, { "name": "video_stereo_right", "l2": 6.584453, "mean_abs": 0.142958, "relative": 0.92827 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.125075, "mean_abs": 0.085389, "relative": 0.760532 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 54, "start_frame": 270, "end_frame": 289, "center_frame": 279, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.501392, "mean_abs": 0.242988, "relative": 0.885837 }, { "name": "hand_right_joints", "l2": 10.170703, "mean_abs": 0.265795, "relative": 0.801723 }, { "name": "body_joints", "l2": 6.332572, "mean_abs": 0.09449, "relative": 0.855335 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010621, "mean_abs": 0.001204, "relative": 0.025064 }, { "name": "camera_rotation_matrix", "l2": 3.000304, "mean_abs": 0.212399, "relative": 0.921345 }, { "name": "imu_accel_gyro", "l2": 17.045153, "mean_abs": 1.020415, "relative": 0.919645 }, { "name": "depth_confidence", "l2": 11.150211, "mean_abs": 0.197771, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.352726, "mean_abs": 0.110528, "relative": 0.969264 }, { "name": "video_fisheye_cam1", "l2": 5.523856, "mean_abs": 0.115579, "relative": 0.912771 }, { "name": "video_fisheye_cam2", "l2": 5.678971, "mean_abs": 0.116416, "relative": 0.906563 }, { "name": "video_fisheye_cam3", "l2": 5.675601, "mean_abs": 0.114837, "relative": 0.950311 }, { "name": "video_stereo_left", "l2": 5.633018, "mean_abs": 0.122789, "relative": 0.915377 }, { "name": "video_stereo_right", "l2": 6.589821, "mean_abs": 0.142866, "relative": 0.929027 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.085322, "mean_abs": 0.08347, "relative": 0.746305 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 55, "start_frame": 275, "end_frame": 294, "center_frame": 284, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.50104, "mean_abs": 0.243075, "relative": 0.885808 }, { "name": "hand_right_joints", "l2": 10.179086, "mean_abs": 0.265519, "relative": 0.802384 }, { "name": "body_joints", "l2": 6.343948, "mean_abs": 0.094219, "relative": 0.856872 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011264, "mean_abs": 0.001312, "relative": 0.026581 }, { "name": "camera_rotation_matrix", "l2": 2.999931, "mean_abs": 0.212025, "relative": 0.92123 }, { "name": "imu_accel_gyro", "l2": 17.071894, "mean_abs": 1.017247, "relative": 0.921087 }, { "name": "depth_confidence", "l2": 11.208524, "mean_abs": 0.198405, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.352168, "mean_abs": 0.110451, "relative": 0.969163 }, { "name": "video_fisheye_cam1", "l2": 5.523419, "mean_abs": 0.115466, "relative": 0.912698 }, { "name": "video_fisheye_cam2", "l2": 5.683427, "mean_abs": 0.116318, "relative": 0.907275 }, { "name": "video_fisheye_cam3", "l2": 5.663579, "mean_abs": 0.114682, "relative": 0.948298 }, { "name": "video_stereo_left", "l2": 5.629932, "mean_abs": 0.122277, "relative": 0.914875 }, { "name": "video_stereo_right", "l2": 6.5862, "mean_abs": 0.14229, "relative": 0.928516 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.136258, "mean_abs": 0.08691, "relative": 0.764535 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 56, "start_frame": 280, "end_frame": 299, "center_frame": 289, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.510241, "mean_abs": 0.243671, "relative": 0.886584 }, { "name": "hand_right_joints", "l2": 10.186814, "mean_abs": 0.265624, "relative": 0.802993 }, { "name": "body_joints", "l2": 6.352023, "mean_abs": 0.094187, "relative": 0.857963 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010446, "mean_abs": 0.001268, "relative": 0.024652 }, { "name": "camera_rotation_matrix", "l2": 3.000148, "mean_abs": 0.21146, "relative": 0.921297 }, { "name": "imu_accel_gyro", "l2": 17.054495, "mean_abs": 1.010151, "relative": 0.920149 }, { "name": "depth_confidence", "l2": 11.222915, "mean_abs": 0.198581, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.352305, "mean_abs": 0.110438, "relative": 0.969188 }, { "name": "video_fisheye_cam1", "l2": 5.524591, "mean_abs": 0.115363, "relative": 0.912892 }, { "name": "video_fisheye_cam2", "l2": 5.687957, "mean_abs": 0.116326, "relative": 0.907998 }, { "name": "video_fisheye_cam3", "l2": 5.641115, "mean_abs": 0.113751, "relative": 0.944537 }, { "name": "video_stereo_left", "l2": 5.632831, "mean_abs": 0.12216, "relative": 0.915346 }, { "name": "video_stereo_right", "l2": 6.588137, "mean_abs": 0.142195, "relative": 0.928789 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.100228, "mean_abs": 0.080891, "relative": 0.75164 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 57, "start_frame": 285, "end_frame": 304, "center_frame": 294, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.535254, "mean_abs": 0.244059, "relative": 0.888694 }, { "name": "hand_right_joints", "l2": 10.185987, "mean_abs": 0.265517, "relative": 0.802928 }, { "name": "body_joints", "l2": 6.344319, "mean_abs": 0.093956, "relative": 0.856922 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003694, "mean_abs": 0.000624, "relative": 0.008717 }, { "name": "camera_rotation_matrix", "l2": 3.000045, "mean_abs": 0.210864, "relative": 0.921265 }, { "name": "imu_accel_gyro", "l2": 17.064127, "mean_abs": 1.012412, "relative": 0.920668 }, { "name": "depth_confidence", "l2": 11.251786, "mean_abs": 0.198376, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.349633, "mean_abs": 0.110219, "relative": 0.968704 }, { "name": "video_fisheye_cam1", "l2": 5.520924, "mean_abs": 0.115076, "relative": 0.912286 }, { "name": "video_fisheye_cam2", "l2": 5.687249, "mean_abs": 0.116165, "relative": 0.907885 }, { "name": "video_fisheye_cam3", "l2": 5.629753, "mean_abs": 0.113251, "relative": 0.942634 }, { "name": "video_stereo_left", "l2": 5.628438, "mean_abs": 0.12183, "relative": 0.914632 }, { "name": "video_stereo_right", "l2": 6.57941, "mean_abs": 0.14172, "relative": 0.927559 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.092575, "mean_abs": 0.083535, "relative": 0.748901 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 58, "start_frame": 290, "end_frame": 309, "center_frame": 299, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.539229, "mean_abs": 0.243825, "relative": 0.889029 }, { "name": "hand_right_joints", "l2": 10.174846, "mean_abs": 0.265212, "relative": 0.80205 }, { "name": "body_joints", "l2": 6.336211, "mean_abs": 0.093926, "relative": 0.855827 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003752, "mean_abs": 0.000661, "relative": 0.008853 }, { "name": "camera_rotation_matrix", "l2": 2.999995, "mean_abs": 0.210762, "relative": 0.92125 }, { "name": "imu_accel_gyro", "l2": 17.059292, "mean_abs": 1.012648, "relative": 0.920408 }, { "name": "depth_confidence", "l2": 11.277675, "mean_abs": 0.198397, "relative": 0.000374 }, { "name": "video_fisheye_cam0", "l2": 5.35019, "mean_abs": 0.110118, "relative": 0.968805 }, { "name": "video_fisheye_cam1", "l2": 5.518728, "mean_abs": 0.114971, "relative": 0.911923 }, { "name": "video_fisheye_cam2", "l2": 5.689644, "mean_abs": 0.116115, "relative": 0.908267 }, { "name": "video_fisheye_cam3", "l2": 5.624683, "mean_abs": 0.112914, "relative": 0.941785 }, { "name": "video_stereo_left", "l2": 5.628178, "mean_abs": 0.1217, "relative": 0.91459 }, { "name": "video_stereo_right", "l2": 6.575875, "mean_abs": 0.141496, "relative": 0.92706 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.06513, "mean_abs": 0.083919, "relative": 0.739079 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 59, "start_frame": 295, "end_frame": 314, "center_frame": 304, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.533066, "mean_abs": 0.243042, "relative": 0.888509 }, { "name": "hand_right_joints", "l2": 10.164195, "mean_abs": 0.265057, "relative": 0.80121 }, { "name": "body_joints", "l2": 6.326706, "mean_abs": 0.093508, "relative": 0.854543 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005024, "mean_abs": 0.000864, "relative": 0.011856 }, { "name": "camera_rotation_matrix", "l2": 2.999912, "mean_abs": 0.210882, "relative": 0.921224 }, { "name": "imu_accel_gyro", "l2": 17.05864, "mean_abs": 1.01429, "relative": 0.920372 }, { "name": "depth_confidence", "l2": 11.284719, "mean_abs": 0.198454, "relative": 0.000375 }, { "name": "video_fisheye_cam0", "l2": 5.351151, "mean_abs": 0.110218, "relative": 0.968979 }, { "name": "video_fisheye_cam1", "l2": 5.51499, "mean_abs": 0.114962, "relative": 0.911306 }, { "name": "video_fisheye_cam2", "l2": 5.68908, "mean_abs": 0.116155, "relative": 0.908177 }, { "name": "video_fisheye_cam3", "l2": 5.61867, "mean_abs": 0.112961, "relative": 0.940779 }, { "name": "video_stereo_left", "l2": 5.626699, "mean_abs": 0.121825, "relative": 0.91435 }, { "name": "video_stereo_right", "l2": 6.571859, "mean_abs": 0.141647, "relative": 0.926494 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.142382, "mean_abs": 0.086303, "relative": 0.766726 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 60, "start_frame": 300, "end_frame": 319, "center_frame": 309, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.535686, "mean_abs": 0.243069, "relative": 0.88873 }, { "name": "hand_right_joints", "l2": 10.147537, "mean_abs": 0.264214, "relative": 0.799897 }, { "name": "body_joints", "l2": 6.319286, "mean_abs": 0.093369, "relative": 0.853541 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005627, "mean_abs": 0.000826, "relative": 0.013278 }, { "name": "camera_rotation_matrix", "l2": 2.999996, "mean_abs": 0.210701, "relative": 0.92125 }, { "name": "imu_accel_gyro", "l2": 17.072662, "mean_abs": 1.015073, "relative": 0.921129 }, { "name": "depth_confidence", "l2": 11.271672, "mean_abs": 0.197965, "relative": 0.000374 }, { "name": "video_fisheye_cam0", "l2": 5.352024, "mean_abs": 0.110268, "relative": 0.969137 }, { "name": "video_fisheye_cam1", "l2": 5.5121, "mean_abs": 0.114882, "relative": 0.910828 }, { "name": "video_fisheye_cam2", "l2": 5.688869, "mean_abs": 0.116146, "relative": 0.908143 }, { "name": "video_fisheye_cam3", "l2": 5.613499, "mean_abs": 0.11296, "relative": 0.939913 }, { "name": "video_stereo_left", "l2": 5.627257, "mean_abs": 0.121737, "relative": 0.91444 }, { "name": "video_stereo_right", "l2": 6.566622, "mean_abs": 0.141352, "relative": 0.925756 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.145015, "mean_abs": 0.086701, "relative": 0.767669 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008358, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 61, "start_frame": 305, "end_frame": 324, "center_frame": 314, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.538043, "mean_abs": 0.242435, "relative": 0.888929 }, { "name": "hand_right_joints", "l2": 10.142227, "mean_abs": 0.263898, "relative": 0.799479 }, { "name": "body_joints", "l2": 6.31417, "mean_abs": 0.093212, "relative": 0.85285 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.002689, "mean_abs": 0.000448, "relative": 0.006346 }, { "name": "camera_rotation_matrix", "l2": 3.000023, "mean_abs": 0.210631, "relative": 0.921258 }, { "name": "imu_accel_gyro", "l2": 17.060497, "mean_abs": 1.014933, "relative": 0.920473 }, { "name": "depth_confidence", "l2": 11.264976, "mean_abs": 0.198196, "relative": 0.000374 }, { "name": "video_fisheye_cam0", "l2": 5.353688, "mean_abs": 0.110414, "relative": 0.969438 }, { "name": "video_fisheye_cam1", "l2": 5.517889, "mean_abs": 0.115089, "relative": 0.911785 }, { "name": "video_fisheye_cam2", "l2": 5.680873, "mean_abs": 0.116249, "relative": 0.906867 }, { "name": "video_fisheye_cam3", "l2": 5.614627, "mean_abs": 0.113293, "relative": 0.940102 }, { "name": "video_stereo_left", "l2": 5.618299, "mean_abs": 0.12201, "relative": 0.912985 }, { "name": "video_stereo_right", "l2": 6.573225, "mean_abs": 0.141783, "relative": 0.926687 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.073499, "mean_abs": 0.081254, "relative": 0.742074 }, { "name": "caption_objects_interaction_text", "l2": 1.896748, "mean_abs": 0.013106, "relative": 0.870941 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 62, "start_frame": 310, "end_frame": 329, "center_frame": 319, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.541457, "mean_abs": 0.24289, "relative": 0.889217 }, { "name": "hand_right_joints", "l2": 10.155529, "mean_abs": 0.264792, "relative": 0.800527 }, { "name": "body_joints", "l2": 6.308276, "mean_abs": 0.093519, "relative": 0.852054 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.002654, "mean_abs": 0.000418, "relative": 0.006262 }, { "name": "camera_rotation_matrix", "l2": 3.000039, "mean_abs": 0.210621, "relative": 0.921263 }, { "name": "imu_accel_gyro", "l2": 17.059395, "mean_abs": 1.014416, "relative": 0.920413 }, { "name": "depth_confidence", "l2": 11.253683, "mean_abs": 0.198135, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.358119, "mean_abs": 0.110554, "relative": 0.970241 }, { "name": "video_fisheye_cam1", "l2": 5.518296, "mean_abs": 0.115301, "relative": 0.911852 }, { "name": "video_fisheye_cam2", "l2": 5.67453, "mean_abs": 0.11624, "relative": 0.905854 }, { "name": "video_fisheye_cam3", "l2": 5.613483, "mean_abs": 0.113624, "relative": 0.93991 }, { "name": "video_stereo_left", "l2": 5.610917, "mean_abs": 0.121982, "relative": 0.911785 }, { "name": "video_stereo_right", "l2": 6.574263, "mean_abs": 0.142153, "relative": 0.926833 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.09961, "mean_abs": 0.082291, "relative": 0.751419 }, { "name": "caption_objects_interaction_text", "l2": 1.896748, "mean_abs": 0.013369, "relative": 0.870941 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 63, "start_frame": 315, "end_frame": 334, "center_frame": 324, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.5613, "mean_abs": 0.243433, "relative": 0.890891 }, { "name": "hand_right_joints", "l2": 10.16126, "mean_abs": 0.265412, "relative": 0.800979 }, { "name": "body_joints", "l2": 6.299488, "mean_abs": 0.093695, "relative": 0.850867 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.002407, "mean_abs": 0.000384, "relative": 0.005681 }, { "name": "camera_rotation_matrix", "l2": 2.999967, "mean_abs": 0.210447, "relative": 0.921241 }, { "name": "imu_accel_gyro", "l2": 17.061378, "mean_abs": 1.012873, "relative": 0.92052 }, { "name": "depth_confidence", "l2": 11.188826, "mean_abs": 0.197485, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.362854, "mean_abs": 0.110885, "relative": 0.971098 }, { "name": "video_fisheye_cam1", "l2": 5.523861, "mean_abs": 0.115613, "relative": 0.912771 }, { "name": "video_fisheye_cam2", "l2": 5.670418, "mean_abs": 0.116173, "relative": 0.905198 }, { "name": "video_fisheye_cam3", "l2": 5.620752, "mean_abs": 0.114061, "relative": 0.941127 }, { "name": "video_stereo_left", "l2": 5.607663, "mean_abs": 0.121809, "relative": 0.911256 }, { "name": "video_stereo_right", "l2": 6.580559, "mean_abs": 0.142467, "relative": 0.927721 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.11176, "mean_abs": 0.082328, "relative": 0.755767 }, { "name": "caption_objects_interaction_text", "l2": 1.896748, "mean_abs": 0.013298, "relative": 0.870941 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 64, "start_frame": 320, "end_frame": 339, "center_frame": 329, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.566855, "mean_abs": 0.243346, "relative": 0.891359 }, { "name": "hand_right_joints", "l2": 10.173046, "mean_abs": 0.266054, "relative": 0.801908 }, { "name": "body_joints", "l2": 6.278884, "mean_abs": 0.093997, "relative": 0.848084 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003423, "mean_abs": 0.000517, "relative": 0.008077 }, { "name": "camera_rotation_matrix", "l2": 3.000003, "mean_abs": 0.21051, "relative": 0.921252 }, { "name": "imu_accel_gyro", "l2": 17.076361, "mean_abs": 1.015445, "relative": 0.921328 }, { "name": "depth_confidence", "l2": 11.221189, "mean_abs": 0.196877, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.365507, "mean_abs": 0.110907, "relative": 0.971579 }, { "name": "video_fisheye_cam1", "l2": 5.523958, "mean_abs": 0.115615, "relative": 0.912787 }, { "name": "video_fisheye_cam2", "l2": 5.665871, "mean_abs": 0.116095, "relative": 0.904472 }, { "name": "video_fisheye_cam3", "l2": 5.617879, "mean_abs": 0.113893, "relative": 0.940646 }, { "name": "video_stereo_left", "l2": 5.603578, "mean_abs": 0.121837, "relative": 0.910593 }, { "name": "video_stereo_right", "l2": 6.582276, "mean_abs": 0.142603, "relative": 0.927963 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.192585, "mean_abs": 0.087194, "relative": 0.784693 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 65, "start_frame": 325, "end_frame": 344, "center_frame": 334, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.583292, "mean_abs": 0.244987, "relative": 0.892746 }, { "name": "hand_right_joints", "l2": 10.185215, "mean_abs": 0.267063, "relative": 0.802867 }, { "name": "body_joints", "l2": 6.21536, "mean_abs": 0.094235, "relative": 0.839504 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00535, "mean_abs": 0.000839, "relative": 0.012626 }, { "name": "camera_rotation_matrix", "l2": 3.000053, "mean_abs": 0.21055, "relative": 0.921268 }, { "name": "imu_accel_gyro", "l2": 17.065771, "mean_abs": 1.019963, "relative": 0.920757 }, { "name": "depth_confidence", "l2": 11.228629, "mean_abs": 0.197039, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.372324, "mean_abs": 0.111503, "relative": 0.972813 }, { "name": "video_fisheye_cam1", "l2": 5.516524, "mean_abs": 0.115901, "relative": 0.911559 }, { "name": "video_fisheye_cam2", "l2": 5.658613, "mean_abs": 0.11613, "relative": 0.903313 }, { "name": "video_fisheye_cam3", "l2": 5.627267, "mean_abs": 0.115271, "relative": 0.942218 }, { "name": "video_stereo_left", "l2": 5.599951, "mean_abs": 0.122336, "relative": 0.910003 }, { "name": "video_stereo_right", "l2": 6.571826, "mean_abs": 0.143061, "relative": 0.92649 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.248141, "mean_abs": 0.085718, "relative": 0.804576 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 66, "start_frame": 330, "end_frame": 349, "center_frame": 339, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.62168, "mean_abs": 0.245615, "relative": 0.895984 }, { "name": "hand_right_joints", "l2": 10.195558, "mean_abs": 0.266943, "relative": 0.803683 }, { "name": "body_joints", "l2": 6.170398, "mean_abs": 0.094205, "relative": 0.833431 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004718, "mean_abs": 0.000741, "relative": 0.011133 }, { "name": "camera_rotation_matrix", "l2": 3.000156, "mean_abs": 0.210429, "relative": 0.921299 }, { "name": "imu_accel_gyro", "l2": 17.036724, "mean_abs": 1.01966, "relative": 0.91919 }, { "name": "depth_confidence", "l2": 11.199795, "mean_abs": 0.196897, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.37413, "mean_abs": 0.111424, "relative": 0.97314 }, { "name": "video_fisheye_cam1", "l2": 5.506195, "mean_abs": 0.11631, "relative": 0.909852 }, { "name": "video_fisheye_cam2", "l2": 5.661299, "mean_abs": 0.116056, "relative": 0.903742 }, { "name": "video_fisheye_cam3", "l2": 5.621819, "mean_abs": 0.115149, "relative": 0.941306 }, { "name": "video_stereo_left", "l2": 5.603442, "mean_abs": 0.122261, "relative": 0.91057 }, { "name": "video_stereo_right", "l2": 6.559236, "mean_abs": 0.143582, "relative": 0.924715 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.174122, "mean_abs": 0.082015, "relative": 0.778085 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 67, "start_frame": 335, "end_frame": 354, "center_frame": 344, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.626592, "mean_abs": 0.244796, "relative": 0.896398 }, { "name": "hand_right_joints", "l2": 10.204858, "mean_abs": 0.265826, "relative": 0.804416 }, { "name": "body_joints", "l2": 6.135717, "mean_abs": 0.093543, "relative": 0.828746 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00509, "mean_abs": 0.000819, "relative": 0.012013 }, { "name": "camera_rotation_matrix", "l2": 3.000093, "mean_abs": 0.210518, "relative": 0.92128 }, { "name": "imu_accel_gyro", "l2": 17.038498, "mean_abs": 1.020441, "relative": 0.919286 }, { "name": "depth_confidence", "l2": 11.187368, "mean_abs": 0.197134, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.380547, "mean_abs": 0.111443, "relative": 0.974302 }, { "name": "video_fisheye_cam1", "l2": 5.494634, "mean_abs": 0.115681, "relative": 0.907942 }, { "name": "video_fisheye_cam2", "l2": 5.667057, "mean_abs": 0.116151, "relative": 0.904661 }, { "name": "video_fisheye_cam3", "l2": 5.619343, "mean_abs": 0.114483, "relative": 0.940891 }, { "name": "video_stereo_left", "l2": 5.60632, "mean_abs": 0.122116, "relative": 0.911038 }, { "name": "video_stereo_right", "l2": 6.537932, "mean_abs": 0.142612, "relative": 0.921711 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.210959, "mean_abs": 0.085017, "relative": 0.791269 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 68, "start_frame": 340, "end_frame": 359, "center_frame": 349, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.63557, "mean_abs": 0.244062, "relative": 0.897156 }, { "name": "hand_right_joints", "l2": 10.216199, "mean_abs": 0.264959, "relative": 0.80531 }, { "name": "body_joints", "l2": 6.081012, "mean_abs": 0.091953, "relative": 0.821357 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013156, "mean_abs": 0.001865, "relative": 0.031047 }, { "name": "camera_rotation_matrix", "l2": 3.000182, "mean_abs": 0.211474, "relative": 0.921307 }, { "name": "imu_accel_gyro", "l2": 17.033783, "mean_abs": 1.023907, "relative": 0.919031 }, { "name": "depth_confidence", "l2": 11.177936, "mean_abs": 0.197493, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.376217, "mean_abs": 0.111757, "relative": 0.973518 }, { "name": "video_fisheye_cam1", "l2": 5.488466, "mean_abs": 0.115853, "relative": 0.906923 }, { "name": "video_fisheye_cam2", "l2": 5.675303, "mean_abs": 0.116788, "relative": 0.905978 }, { "name": "video_fisheye_cam3", "l2": 5.618766, "mean_abs": 0.114809, "relative": 0.940795 }, { "name": "video_stereo_left", "l2": 5.601129, "mean_abs": 0.123058, "relative": 0.910195 }, { "name": "video_stereo_right", "l2": 6.515047, "mean_abs": 0.143013, "relative": 0.918485 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.284588, "mean_abs": 0.088811, "relative": 0.81762 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 69, "start_frame": 345, "end_frame": 364, "center_frame": 354, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.623002, "mean_abs": 0.240702, "relative": 0.896096 }, { "name": "hand_right_joints", "l2": 10.213144, "mean_abs": 0.262854, "relative": 0.805069 }, { "name": "body_joints", "l2": 6.004619, "mean_abs": 0.09013, "relative": 0.811039 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.02678, "mean_abs": 0.003654, "relative": 0.063197 }, { "name": "camera_rotation_matrix", "l2": 3.001189, "mean_abs": 0.212608, "relative": 0.921616 }, { "name": "imu_accel_gyro", "l2": 17.081009, "mean_abs": 1.032228, "relative": 0.921579 }, { "name": "depth_confidence", "l2": 11.170732, "mean_abs": 0.198523, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.373235, "mean_abs": 0.111877, "relative": 0.972978 }, { "name": "video_fisheye_cam1", "l2": 5.468278, "mean_abs": 0.116051, "relative": 0.903587 }, { "name": "video_fisheye_cam2", "l2": 5.683589, "mean_abs": 0.11734, "relative": 0.9073 }, { "name": "video_fisheye_cam3", "l2": 5.613098, "mean_abs": 0.115028, "relative": 0.939846 }, { "name": "video_stereo_left", "l2": 5.592976, "mean_abs": 0.123965, "relative": 0.90887 }, { "name": "video_stereo_right", "l2": 6.474507, "mean_abs": 0.143255, "relative": 0.91277 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.150919, "mean_abs": 0.084416, "relative": 0.769782 }, { "name": "caption_objects_interaction_text", "l2": 1.770626, "mean_abs": 0.011532, "relative": 0.813029 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 70, "start_frame": 350, "end_frame": 369, "center_frame": 359, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.591638, "mean_abs": 0.239375, "relative": 0.89345 }, { "name": "hand_right_joints", "l2": 10.220386, "mean_abs": 0.261146, "relative": 0.80564 }, { "name": "body_joints", "l2": 5.975263, "mean_abs": 0.090237, "relative": 0.807074 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.046195, "mean_abs": 0.006423, "relative": 0.109013 }, { "name": "camera_rotation_matrix", "l2": 3.002176, "mean_abs": 0.214327, "relative": 0.921919 }, { "name": "imu_accel_gyro", "l2": 17.081757, "mean_abs": 1.043056, "relative": 0.92162 }, { "name": "depth_confidence", "l2": 11.157333, "mean_abs": 0.199659, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.374912, "mean_abs": 0.112783, "relative": 0.973282 }, { "name": "video_fisheye_cam1", "l2": 5.47812, "mean_abs": 0.117105, "relative": 0.905213 }, { "name": "video_fisheye_cam2", "l2": 5.698023, "mean_abs": 0.118521, "relative": 0.909605 }, { "name": "video_fisheye_cam3", "l2": 5.613207, "mean_abs": 0.115969, "relative": 0.939864 }, { "name": "video_stereo_left", "l2": 5.596223, "mean_abs": 0.12581, "relative": 0.909397 }, { "name": "video_stereo_right", "l2": 6.470383, "mean_abs": 0.146038, "relative": 0.912188 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.156116, "mean_abs": 0.085386, "relative": 0.771641 }, { "name": "caption_objects_interaction_text", "l2": 1.770626, "mean_abs": 0.011644, "relative": 0.813029 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 71, "start_frame": 355, "end_frame": 374, "center_frame": 364, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.514571, "mean_abs": 0.239683, "relative": 0.886949 }, { "name": "hand_right_joints", "l2": 10.181435, "mean_abs": 0.260539, "relative": 0.802569 }, { "name": "body_joints", "l2": 6.003439, "mean_abs": 0.092572, "relative": 0.81088 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.07891, "mean_abs": 0.01107, "relative": 0.186217 }, { "name": "camera_rotation_matrix", "l2": 3.007717, "mean_abs": 0.218306, "relative": 0.923621 }, { "name": "imu_accel_gyro", "l2": 17.100456, "mean_abs": 1.067124, "relative": 0.922629 }, { "name": "depth_confidence", "l2": 11.224487, "mean_abs": 0.203233, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.381648, "mean_abs": 0.113838, "relative": 0.974501 }, { "name": "video_fisheye_cam1", "l2": 5.481896, "mean_abs": 0.118403, "relative": 0.905837 }, { "name": "video_fisheye_cam2", "l2": 5.70089, "mean_abs": 0.119606, "relative": 0.910062 }, { "name": "video_fisheye_cam3", "l2": 5.595909, "mean_abs": 0.116808, "relative": 0.936967 }, { "name": "video_stereo_left", "l2": 5.580382, "mean_abs": 0.12851, "relative": 0.906823 }, { "name": "video_stereo_right", "l2": 6.440889, "mean_abs": 0.147946, "relative": 0.90803 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.144968, "mean_abs": 0.084755, "relative": 0.767652 }, { "name": "caption_objects_interaction_text", "l2": 1.770626, "mean_abs": 0.011613, "relative": 0.813029 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 72, "start_frame": 360, "end_frame": 379, "center_frame": 369, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.407304, "mean_abs": 0.238672, "relative": 0.8779 }, { "name": "hand_right_joints", "l2": 10.101263, "mean_abs": 0.259516, "relative": 0.79625 }, { "name": "body_joints", "l2": 6.039264, "mean_abs": 0.09491, "relative": 0.815719 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.134197, "mean_abs": 0.018609, "relative": 0.316685 }, { "name": "camera_rotation_matrix", "l2": 3.029429, "mean_abs": 0.225251, "relative": 0.930288 }, { "name": "imu_accel_gyro", "l2": 17.138571, "mean_abs": 1.099921, "relative": 0.924685 }, { "name": "depth_confidence", "l2": 11.355933, "mean_abs": 0.210278, "relative": 0.000377 }, { "name": "video_fisheye_cam0", "l2": 5.376201, "mean_abs": 0.115087, "relative": 0.973515 }, { "name": "video_fisheye_cam1", "l2": 5.53995, "mean_abs": 0.122155, "relative": 0.91543 }, { "name": "video_fisheye_cam2", "l2": 5.67158, "mean_abs": 0.121434, "relative": 0.905383 }, { "name": "video_fisheye_cam3", "l2": 5.61836, "mean_abs": 0.119346, "relative": 0.940727 }, { "name": "video_stereo_left", "l2": 5.558368, "mean_abs": 0.131994, "relative": 0.903246 }, { "name": "video_stereo_right", "l2": 6.4317, "mean_abs": 0.151129, "relative": 0.906735 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.111341, "mean_abs": 0.084438, "relative": 0.755617 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009994, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 73, "start_frame": 365, "end_frame": 384, "center_frame": 374, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.275119, "mean_abs": 0.238486, "relative": 0.86675 }, { "name": "hand_right_joints", "l2": 10.016108, "mean_abs": 0.260199, "relative": 0.789537 }, { "name": "body_joints", "l2": 6.13112, "mean_abs": 0.099663, "relative": 0.828125 }, { "name": "body_contacts", "l2": 3.464655, "mean_abs": 0.081732, "relative": 0.999784 }, { "name": "camera_translation", "l2": 0.215371, "mean_abs": 0.029088, "relative": 0.508245 }, { "name": "camera_rotation_matrix", "l2": 3.080724, "mean_abs": 0.231736, "relative": 0.94604 }, { "name": "imu_accel_gyro", "l2": 17.178738, "mean_abs": 1.137901, "relative": 0.926852 }, { "name": "depth_confidence", "l2": 11.713172, "mean_abs": 0.218932, "relative": 0.000389 }, { "name": "video_fisheye_cam0", "l2": 5.394524, "mean_abs": 0.117677, "relative": 0.976833 }, { "name": "video_fisheye_cam1", "l2": 5.598539, "mean_abs": 0.124788, "relative": 0.925111 }, { "name": "video_fisheye_cam2", "l2": 5.633441, "mean_abs": 0.12363, "relative": 0.899295 }, { "name": "video_fisheye_cam3", "l2": 5.619845, "mean_abs": 0.121239, "relative": 0.940975 }, { "name": "video_stereo_left", "l2": 5.568719, "mean_abs": 0.133942, "relative": 0.904928 }, { "name": "video_stereo_right", "l2": 6.449767, "mean_abs": 0.154418, "relative": 0.909282 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.12896, "mean_abs": 0.085955, "relative": 0.761923 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009994, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 74, "start_frame": 370, "end_frame": 389, "center_frame": 379, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.105679, "mean_abs": 0.23694, "relative": 0.852457 }, { "name": "hand_right_joints", "l2": 9.948689, "mean_abs": 0.25733, "relative": 0.784223 }, { "name": "body_joints", "l2": 6.318282, "mean_abs": 0.106133, "relative": 0.853405 }, { "name": "body_contacts", "l2": 3.464701, "mean_abs": 0.081733, "relative": 0.999798 }, { "name": "camera_translation", "l2": 0.316651, "mean_abs": 0.040885, "relative": 0.747251 }, { "name": "camera_rotation_matrix", "l2": 3.199472, "mean_abs": 0.249845, "relative": 0.982506 }, { "name": "imu_accel_gyro", "l2": 17.275772, "mean_abs": 1.137, "relative": 0.932087 }, { "name": "depth_confidence", "l2": 11.925965, "mean_abs": 0.22267, "relative": 0.000396 }, { "name": "video_fisheye_cam0", "l2": 5.389902, "mean_abs": 0.119124, "relative": 0.975996 }, { "name": "video_fisheye_cam1", "l2": 5.67604, "mean_abs": 0.127012, "relative": 0.937918 }, { "name": "video_fisheye_cam2", "l2": 5.561861, "mean_abs": 0.123768, "relative": 0.887868 }, { "name": "video_fisheye_cam3", "l2": 5.636756, "mean_abs": 0.123746, "relative": 0.943807 }, { "name": "video_stereo_left", "l2": 5.523664, "mean_abs": 0.134341, "relative": 0.897606 }, { "name": "video_stereo_right", "l2": 6.503246, "mean_abs": 0.158114, "relative": 0.916821 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.115332, "mean_abs": 0.083278, "relative": 0.757045 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009994, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 75, "start_frame": 375, "end_frame": 394, "center_frame": 384, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.951443, "mean_abs": 0.237282, "relative": 0.839447 }, { "name": "hand_right_joints", "l2": 9.919042, "mean_abs": 0.256852, "relative": 0.781886 }, { "name": "body_joints", "l2": 6.51832, "mean_abs": 0.111729, "relative": 0.880424 }, { "name": "body_contacts", "l2": 3.464812, "mean_abs": 0.081742, "relative": 0.99983 }, { "name": "camera_translation", "l2": 0.390302, "mean_abs": 0.048497, "relative": 0.921057 }, { "name": "camera_rotation_matrix", "l2": 3.24788, "mean_abs": 0.257128, "relative": 0.997371 }, { "name": "imu_accel_gyro", "l2": 17.403225, "mean_abs": 1.123302, "relative": 0.938964 }, { "name": "depth_confidence", "l2": 12.370242, "mean_abs": 0.226986, "relative": 0.000411 }, { "name": "video_fisheye_cam0", "l2": 5.393731, "mean_abs": 0.119352, "relative": 0.976689 }, { "name": "video_fisheye_cam1", "l2": 5.758222, "mean_abs": 0.129932, "relative": 0.951497 }, { "name": "video_fisheye_cam2", "l2": 5.457019, "mean_abs": 0.123311, "relative": 0.871132 }, { "name": "video_fisheye_cam3", "l2": 5.64953, "mean_abs": 0.12455, "relative": 0.945946 }, { "name": "video_stereo_left", "l2": 5.495221, "mean_abs": 0.134354, "relative": 0.892984 }, { "name": "video_stereo_right", "l2": 6.561372, "mean_abs": 0.16228, "relative": 0.925016 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.130248, "mean_abs": 0.083589, "relative": 0.762384 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009994, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 76, "start_frame": 380, "end_frame": 399, "center_frame": 389, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.800534, "mean_abs": 0.226537, "relative": 0.826717 }, { "name": "hand_right_joints", "l2": 10.091384, "mean_abs": 0.269905, "relative": 0.795471 }, { "name": "body_joints", "l2": 6.695938, "mean_abs": 0.115747, "relative": 0.904415 }, { "name": "body_contacts", "l2": 3.465003, "mean_abs": 0.081743, "relative": 0.999885 }, { "name": "camera_translation", "l2": 0.410053, "mean_abs": 0.049753, "relative": 0.967666 }, { "name": "camera_rotation_matrix", "l2": 3.180576, "mean_abs": 0.257334, "relative": 0.976703 }, { "name": "imu_accel_gyro", "l2": 17.365103, "mean_abs": 1.086692, "relative": 0.936907 }, { "name": "depth_confidence", "l2": 12.464439, "mean_abs": 0.227339, "relative": 0.000414 }, { "name": "video_fisheye_cam0", "l2": 5.26579, "mean_abs": 0.115972, "relative": 0.953522 }, { "name": "video_fisheye_cam1", "l2": 5.758804, "mean_abs": 0.12953, "relative": 0.951594 }, { "name": "video_fisheye_cam2", "l2": 5.259065, "mean_abs": 0.120213, "relative": 0.839532 }, { "name": "video_fisheye_cam3", "l2": 5.636287, "mean_abs": 0.122545, "relative": 0.943728 }, { "name": "video_stereo_left", "l2": 5.386467, "mean_abs": 0.133218, "relative": 0.875311 }, { "name": "video_stereo_right", "l2": 6.514914, "mean_abs": 0.161744, "relative": 0.918466 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.115091, "mean_abs": 0.084369, "relative": 0.756959 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009994, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 77, "start_frame": 385, "end_frame": 404, "center_frame": 394, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "gooseneck kettle", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.628628, "mean_abs": 0.223635, "relative": 0.812216 }, { "name": "hand_right_joints", "l2": 10.287221, "mean_abs": 0.28236, "relative": 0.810908 }, { "name": "body_joints", "l2": 6.831984, "mean_abs": 0.117374, "relative": 0.92279 }, { "name": "body_contacts", "l2": 3.464766, "mean_abs": 0.081745, "relative": 0.999816 }, { "name": "camera_translation", "l2": 0.390231, "mean_abs": 0.046745, "relative": 0.920891 }, { "name": "camera_rotation_matrix", "l2": 3.097351, "mean_abs": 0.245733, "relative": 0.951146 }, { "name": "imu_accel_gyro", "l2": 17.286362, "mean_abs": 1.034728, "relative": 0.932659 }, { "name": "depth_confidence", "l2": 12.489844, "mean_abs": 0.225439, "relative": 0.000415 }, { "name": "video_fisheye_cam0", "l2": 5.181701, "mean_abs": 0.112365, "relative": 0.938295 }, { "name": "video_fisheye_cam1", "l2": 5.749321, "mean_abs": 0.127474, "relative": 0.950027 }, { "name": "video_fisheye_cam2", "l2": 5.116544, "mean_abs": 0.116152, "relative": 0.81678 }, { "name": "video_fisheye_cam3", "l2": 5.622813, "mean_abs": 0.119438, "relative": 0.941472 }, { "name": "video_stereo_left", "l2": 5.326423, "mean_abs": 0.132611, "relative": 0.865554 }, { "name": "video_stereo_right", "l2": 6.39071, "mean_abs": 0.157565, "relative": 0.900956 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.038098, "mean_abs": 0.079015, "relative": 0.729405 }, { "name": "caption_objects_interaction_text", "l2": 1.87404, "mean_abs": 0.015344, "relative": 0.860514 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 78, "start_frame": 390, "end_frame": 409, "center_frame": 399, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "gooseneck kettle", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.514749, "mean_abs": 0.224016, "relative": 0.80261 }, { "name": "hand_right_joints", "l2": 10.614914, "mean_abs": 0.297098, "relative": 0.836739 }, { "name": "body_joints", "l2": 6.891379, "mean_abs": 0.115185, "relative": 0.930813 }, { "name": "body_contacts", "l2": 3.464502, "mean_abs": 0.081737, "relative": 0.99974 }, { "name": "camera_translation", "l2": 0.321475, "mean_abs": 0.038762, "relative": 0.758636 }, { "name": "camera_rotation_matrix", "l2": 3.022541, "mean_abs": 0.238543, "relative": 0.928173 }, { "name": "imu_accel_gyro", "l2": 17.179747, "mean_abs": 0.96399, "relative": 0.926906 }, { "name": "depth_confidence", "l2": 12.335196, "mean_abs": 0.216193, "relative": 0.000409 }, { "name": "video_fisheye_cam0", "l2": 5.06357, "mean_abs": 0.108748, "relative": 0.916904 }, { "name": "video_fisheye_cam1", "l2": 5.703336, "mean_abs": 0.124275, "relative": 0.942428 }, { "name": "video_fisheye_cam2", "l2": 5.012095, "mean_abs": 0.112521, "relative": 0.800106 }, { "name": "video_fisheye_cam3", "l2": 5.63639, "mean_abs": 0.11721, "relative": 0.943746 }, { "name": "video_stereo_left", "l2": 5.328961, "mean_abs": 0.131681, "relative": 0.865967 }, { "name": "video_stereo_right", "l2": 6.385695, "mean_abs": 0.156302, "relative": 0.900249 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.131577, "mean_abs": 0.082962, "relative": 0.762859 }, { "name": "caption_objects_interaction_text", "l2": 1.87404, "mean_abs": 0.015604, "relative": 0.860514 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 79, "start_frame": 395, "end_frame": 414, "center_frame": 404, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "kettle", "coffee dripper", "scale", "gooseneck kettle", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.496292, "mean_abs": 0.230467, "relative": 0.801053 }, { "name": "hand_right_joints", "l2": 11.037094, "mean_abs": 0.312022, "relative": 0.870018 }, { "name": "body_joints", "l2": 6.977882, "mean_abs": 0.113715, "relative": 0.942497 }, { "name": "body_contacts", "l2": 3.464392, "mean_abs": 0.081735, "relative": 0.999708 }, { "name": "camera_translation", "l2": 0.251514, "mean_abs": 0.031456, "relative": 0.593537 }, { "name": "camera_rotation_matrix", "l2": 3.003391, "mean_abs": 0.230612, "relative": 0.922293 }, { "name": "imu_accel_gyro", "l2": 17.102474, "mean_abs": 0.917831, "relative": 0.922737 }, { "name": "depth_confidence", "l2": 12.036451, "mean_abs": 0.208171, "relative": 0.000399 }, { "name": "video_fisheye_cam0", "l2": 5.059725, "mean_abs": 0.107846, "relative": 0.916208 }, { "name": "video_fisheye_cam1", "l2": 5.603852, "mean_abs": 0.120944, "relative": 0.925989 }, { "name": "video_fisheye_cam2", "l2": 5.012216, "mean_abs": 0.111381, "relative": 0.800126 }, { "name": "video_fisheye_cam3", "l2": 5.614576, "mean_abs": 0.116469, "relative": 0.940093 }, { "name": "video_stereo_left", "l2": 5.280628, "mean_abs": 0.127819, "relative": 0.858113 }, { "name": "video_stereo_right", "l2": 6.197649, "mean_abs": 0.14908, "relative": 0.873738 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.077765, "mean_abs": 0.080212, "relative": 0.743601 }, { "name": "caption_objects_interaction_text", "l2": 1.87404, "mean_abs": 0.01547, "relative": 0.860514 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 80, "start_frame": 400, "end_frame": 419, "center_frame": 409, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.554375, "mean_abs": 0.23613, "relative": 0.805952 }, { "name": "hand_right_joints", "l2": 11.31819, "mean_abs": 0.319864, "relative": 0.892176 }, { "name": "body_joints", "l2": 7.0382, "mean_abs": 0.11259, "relative": 0.950644 }, { "name": "body_contacts", "l2": 3.464259, "mean_abs": 0.081729, "relative": 0.99967 }, { "name": "camera_translation", "l2": 0.18636, "mean_abs": 0.024176, "relative": 0.439783 }, { "name": "camera_rotation_matrix", "l2": 2.999428, "mean_abs": 0.227034, "relative": 0.921076 }, { "name": "imu_accel_gyro", "l2": 17.130053, "mean_abs": 0.888701, "relative": 0.924225 }, { "name": "depth_confidence", "l2": 11.566543, "mean_abs": 0.200266, "relative": 0.000384 }, { "name": "video_fisheye_cam0", "l2": 5.091374, "mean_abs": 0.107773, "relative": 0.921939 }, { "name": "video_fisheye_cam1", "l2": 5.50197, "mean_abs": 0.117435, "relative": 0.909154 }, { "name": "video_fisheye_cam2", "l2": 5.081838, "mean_abs": 0.112447, "relative": 0.81124 }, { "name": "video_fisheye_cam3", "l2": 5.619731, "mean_abs": 0.116254, "relative": 0.940956 }, { "name": "video_stereo_left", "l2": 5.292322, "mean_abs": 0.125164, "relative": 0.860013 }, { "name": "video_stereo_right", "l2": 6.017198, "mean_abs": 0.140831, "relative": 0.848299 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.052355, "mean_abs": 0.076334, "relative": 0.734507 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 81, "start_frame": 405, "end_frame": 424, "center_frame": 414, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.609228, "mean_abs": 0.238549, "relative": 0.810579 }, { "name": "hand_right_joints", "l2": 11.547246, "mean_abs": 0.322808, "relative": 0.910232 }, { "name": "body_joints", "l2": 7.070185, "mean_abs": 0.111061, "relative": 0.954964 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081693, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.117479, "mean_abs": 0.015884, "relative": 0.277234 }, { "name": "camera_rotation_matrix", "l2": 3.00189, "mean_abs": 0.227605, "relative": 0.921832 }, { "name": "imu_accel_gyro", "l2": 17.159323, "mean_abs": 0.892653, "relative": 0.925805 }, { "name": "depth_confidence", "l2": 11.497834, "mean_abs": 0.198117, "relative": 0.000382 }, { "name": "video_fisheye_cam0", "l2": 5.13831, "mean_abs": 0.108179, "relative": 0.930438 }, { "name": "video_fisheye_cam1", "l2": 5.455518, "mean_abs": 0.115348, "relative": 0.901478 }, { "name": "video_fisheye_cam2", "l2": 5.175078, "mean_abs": 0.113457, "relative": 0.826124 }, { "name": "video_fisheye_cam3", "l2": 5.620006, "mean_abs": 0.115387, "relative": 0.941002 }, { "name": "video_stereo_left", "l2": 5.364371, "mean_abs": 0.124963, "relative": 0.871721 }, { "name": "video_stereo_right", "l2": 5.946406, "mean_abs": 0.137098, "relative": 0.838319 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.27697, "mean_abs": 0.085927, "relative": 0.814894 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 82, "start_frame": 410, "end_frame": 429, "center_frame": 419, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.689078, "mean_abs": 0.239485, "relative": 0.817315 }, { "name": "hand_right_joints", "l2": 11.82627, "mean_abs": 0.323065, "relative": 0.932226 }, { "name": "body_joints", "l2": 7.07554, "mean_abs": 0.108548, "relative": 0.955687 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081693, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.066028, "mean_abs": 0.009869, "relative": 0.155816 }, { "name": "camera_rotation_matrix", "l2": 3.001708, "mean_abs": 0.227702, "relative": 0.921776 }, { "name": "imu_accel_gyro", "l2": 17.173515, "mean_abs": 0.892, "relative": 0.92657 }, { "name": "depth_confidence", "l2": 11.443389, "mean_abs": 0.197578, "relative": 0.00038 }, { "name": "video_fisheye_cam0", "l2": 5.187912, "mean_abs": 0.108529, "relative": 0.93942 }, { "name": "video_fisheye_cam1", "l2": 5.478342, "mean_abs": 0.116373, "relative": 0.90525 }, { "name": "video_fisheye_cam2", "l2": 5.255195, "mean_abs": 0.11398, "relative": 0.838914 }, { "name": "video_fisheye_cam3", "l2": 5.61191, "mean_abs": 0.114619, "relative": 0.939647 }, { "name": "video_stereo_left", "l2": 5.43385, "mean_abs": 0.125504, "relative": 0.883011 }, { "name": "video_stereo_right", "l2": 5.964892, "mean_abs": 0.137765, "relative": 0.840925 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.005844, "mean_abs": 0.080522, "relative": 0.717861 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 83, "start_frame": 415, "end_frame": 434, "center_frame": 424, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.746851, "mean_abs": 0.24046, "relative": 0.822188 }, { "name": "hand_right_joints", "l2": 12.055023, "mean_abs": 0.324609, "relative": 0.950258 }, { "name": "body_joints", "l2": 7.070397, "mean_abs": 0.107663, "relative": 0.954993 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081693, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.036046, "mean_abs": 0.005732, "relative": 0.085063 }, { "name": "camera_rotation_matrix", "l2": 3.00312, "mean_abs": 0.229352, "relative": 0.922209 }, { "name": "imu_accel_gyro", "l2": 17.18017, "mean_abs": 0.905684, "relative": 0.926929 }, { "name": "depth_confidence", "l2": 11.459223, "mean_abs": 0.199587, "relative": 0.00038 }, { "name": "video_fisheye_cam0", "l2": 5.231576, "mean_abs": 0.10927, "relative": 0.947327 }, { "name": "video_fisheye_cam1", "l2": 5.537949, "mean_abs": 0.118659, "relative": 0.915099 }, { "name": "video_fisheye_cam2", "l2": 5.345337, "mean_abs": 0.115209, "relative": 0.853303 }, { "name": "video_fisheye_cam3", "l2": 5.591638, "mean_abs": 0.114178, "relative": 0.936252 }, { "name": "video_stereo_left", "l2": 5.522382, "mean_abs": 0.129058, "relative": 0.897398 }, { "name": "video_stereo_right", "l2": 6.044815, "mean_abs": 0.141773, "relative": 0.852192 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.01552, "mean_abs": 0.080005, "relative": 0.721325 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 84, "start_frame": 420, "end_frame": 439, "center_frame": 429, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.800752, "mean_abs": 0.243583, "relative": 0.826735 }, { "name": "hand_right_joints", "l2": 11.115651, "mean_abs": 0.348372, "relative": 0.876211 }, { "name": "body_joints", "l2": 6.883744, "mean_abs": 0.109066, "relative": 0.929782 }, { "name": "body_contacts", "l2": 3.463657, "mean_abs": 0.081723, "relative": 0.999496 }, { "name": "camera_translation", "l2": 0.121951, "mean_abs": 0.014874, "relative": 0.287788 }, { "name": "camera_rotation_matrix", "l2": 3.022474, "mean_abs": 0.23582, "relative": 0.928153 }, { "name": "imu_accel_gyro", "l2": 17.174885, "mean_abs": 1.00627, "relative": 0.926644 }, { "name": "depth_confidence", "l2": 12.246173, "mean_abs": 0.212913, "relative": 0.000406 }, { "name": "video_fisheye_cam0", "l2": 5.284696, "mean_abs": 0.113249, "relative": 0.956945 }, { "name": "video_fisheye_cam1", "l2": 5.614779, "mean_abs": 0.122181, "relative": 0.927795 }, { "name": "video_fisheye_cam2", "l2": 5.503014, "mean_abs": 0.121972, "relative": 0.878474 }, { "name": "video_fisheye_cam3", "l2": 5.552759, "mean_abs": 0.114823, "relative": 0.929743 }, { "name": "video_stereo_left", "l2": 5.641941, "mean_abs": 0.137277, "relative": 0.916827 }, { "name": "video_stereo_right", "l2": 6.221944, "mean_abs": 0.152146, "relative": 0.877164 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.02495, "mean_abs": 0.080238, "relative": 0.724699 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 85, "start_frame": 425, "end_frame": 444, "center_frame": 434, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.875466, "mean_abs": 0.246092, "relative": 0.833038 }, { "name": "hand_right_joints", "l2": 10.746642, "mean_abs": 0.344758, "relative": 0.847123 }, { "name": "body_joints", "l2": 6.707621, "mean_abs": 0.110723, "relative": 0.905993 }, { "name": "body_contacts", "l2": 3.463557, "mean_abs": 0.081713, "relative": 0.999468 }, { "name": "camera_translation", "l2": 0.248355, "mean_abs": 0.028447, "relative": 0.586083 }, { "name": "camera_rotation_matrix", "l2": 3.076833, "mean_abs": 0.240537, "relative": 0.944845 }, { "name": "imu_accel_gyro", "l2": 17.230972, "mean_abs": 1.039829, "relative": 0.92967 }, { "name": "depth_confidence", "l2": 12.341931, "mean_abs": 0.219738, "relative": 0.00041 }, { "name": "video_fisheye_cam0", "l2": 5.415807, "mean_abs": 0.118138, "relative": 0.980687 }, { "name": "video_fisheye_cam1", "l2": 5.692019, "mean_abs": 0.126206, "relative": 0.940558 }, { "name": "video_fisheye_cam2", "l2": 5.690833, "mean_abs": 0.12801, "relative": 0.908457 }, { "name": "video_fisheye_cam3", "l2": 5.514188, "mean_abs": 0.116096, "relative": 0.923284 }, { "name": "video_stereo_left", "l2": 5.671679, "mean_abs": 0.140832, "relative": 0.921659 }, { "name": "video_stereo_right", "l2": 6.292434, "mean_abs": 0.155411, "relative": 0.887101 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.063445, "mean_abs": 0.081028, "relative": 0.738476 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 86, "start_frame": 430, "end_frame": 449, "center_frame": 439, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.971376, "mean_abs": 0.247318, "relative": 0.841128 }, { "name": "hand_right_joints", "l2": 10.392783, "mean_abs": 0.335399, "relative": 0.819229 }, { "name": "body_joints", "l2": 6.59886, "mean_abs": 0.109922, "relative": 0.891303 }, { "name": "body_contacts", "l2": 3.463509, "mean_abs": 0.081719, "relative": 0.999454 }, { "name": "camera_translation", "l2": 0.367604, "mean_abs": 0.043004, "relative": 0.867495 }, { "name": "camera_rotation_matrix", "l2": 3.176399, "mean_abs": 0.255713, "relative": 0.97542 }, { "name": "imu_accel_gyro", "l2": 17.313633, "mean_abs": 1.111559, "relative": 0.93413 }, { "name": "depth_confidence", "l2": 12.296823, "mean_abs": 0.223009, "relative": 0.000408 }, { "name": "video_fisheye_cam0", "l2": 5.479136, "mean_abs": 0.122039, "relative": 0.992154 }, { "name": "video_fisheye_cam1", "l2": 5.759768, "mean_abs": 0.130331, "relative": 0.951753 }, { "name": "video_fisheye_cam2", "l2": 5.925627, "mean_abs": 0.134676, "relative": 0.945938 }, { "name": "video_fisheye_cam3", "l2": 5.471596, "mean_abs": 0.11716, "relative": 0.916153 }, { "name": "video_stereo_left", "l2": 5.876812, "mean_abs": 0.146559, "relative": 0.954994 }, { "name": "video_stereo_right", "l2": 6.527585, "mean_abs": 0.162652, "relative": 0.920253 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.070359, "mean_abs": 0.082008, "relative": 0.74095 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 87, "start_frame": 435, "end_frame": 454, "center_frame": 444, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.016482, "mean_abs": 0.245151, "relative": 0.844933 }, { "name": "hand_right_joints", "l2": 9.483637, "mean_abs": 0.286646, "relative": 0.747564 }, { "name": "body_joints", "l2": 6.564006, "mean_abs": 0.110629, "relative": 0.886595 }, { "name": "body_contacts", "l2": 3.463507, "mean_abs": 0.081726, "relative": 0.999453 }, { "name": "camera_translation", "l2": 0.423754, "mean_abs": 0.05071, "relative": 1.0 }, { "name": "camera_rotation_matrix", "l2": 3.216143, "mean_abs": 0.252775, "relative": 0.987625 }, { "name": "imu_accel_gyro", "l2": 17.531416, "mean_abs": 1.164381, "relative": 0.94588 }, { "name": "depth_confidence", "l2": 12.344213, "mean_abs": 0.224002, "relative": 0.00041 }, { "name": "video_fisheye_cam0", "l2": 5.522463, "mean_abs": 0.123015, "relative": 1.0 }, { "name": "video_fisheye_cam1", "l2": 5.774107, "mean_abs": 0.131371, "relative": 0.954122 }, { "name": "video_fisheye_cam2", "l2": 6.087977, "mean_abs": 0.13747, "relative": 0.971855 }, { "name": "video_fisheye_cam3", "l2": 5.457041, "mean_abs": 0.118105, "relative": 0.913716 }, { "name": "video_stereo_left", "l2": 5.947477, "mean_abs": 0.147784, "relative": 0.966477 }, { "name": "video_stereo_right", "l2": 6.597708, "mean_abs": 0.164739, "relative": 0.930138 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.121541, "mean_abs": 0.081947, "relative": 0.759268 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 88, "start_frame": 440, "end_frame": 459, "center_frame": 449, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.005199, "mean_abs": 0.233721, "relative": 0.843981 }, { "name": "hand_right_joints", "l2": 9.273803, "mean_abs": 0.279207, "relative": 0.731024 }, { "name": "body_joints", "l2": 6.087018, "mean_abs": 0.104473, "relative": 0.822169 }, { "name": "body_contacts", "l2": 3.463708, "mean_abs": 0.081722, "relative": 0.999511 }, { "name": "camera_translation", "l2": 0.409807, "mean_abs": 0.050288, "relative": 0.967087 }, { "name": "camera_rotation_matrix", "l2": 3.183126, "mean_abs": 0.244131, "relative": 0.977486 }, { "name": "imu_accel_gyro", "l2": 17.394293, "mean_abs": 1.18547, "relative": 0.938482 }, { "name": "depth_confidence", "l2": 11.799335, "mean_abs": 0.220011, "relative": 0.000392 }, { "name": "video_fisheye_cam0", "l2": 5.505299, "mean_abs": 0.121314, "relative": 0.996892 }, { "name": "video_fisheye_cam1", "l2": 5.772955, "mean_abs": 0.13106, "relative": 0.953932 }, { "name": "video_fisheye_cam2", "l2": 6.216411, "mean_abs": 0.138572, "relative": 0.992358 }, { "name": "video_fisheye_cam3", "l2": 5.468252, "mean_abs": 0.118286, "relative": 0.915593 }, { "name": "video_stereo_left", "l2": 6.007993, "mean_abs": 0.148293, "relative": 0.976311 }, { "name": "video_stereo_right", "l2": 6.666662, "mean_abs": 0.166063, "relative": 0.939859 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.021886, "mean_abs": 0.079115, "relative": 0.723603 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 89, "start_frame": 445, "end_frame": 464, "center_frame": 454, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.978281, "mean_abs": 0.226663, "relative": 0.841711 }, { "name": "hand_right_joints", "l2": 10.235513, "mean_abs": 0.304288, "relative": 0.806832 }, { "name": "body_joints", "l2": 5.905604, "mean_abs": 0.102642, "relative": 0.797665 }, { "name": "body_contacts", "l2": 3.463875, "mean_abs": 0.081717, "relative": 0.999559 }, { "name": "camera_translation", "l2": 0.341151, "mean_abs": 0.042753, "relative": 0.805069 }, { "name": "camera_rotation_matrix", "l2": 3.094209, "mean_abs": 0.226956, "relative": 0.950181 }, { "name": "imu_accel_gyro", "l2": 17.271158, "mean_abs": 1.181019, "relative": 0.931838 }, { "name": "depth_confidence", "l2": 11.25611, "mean_abs": 0.215785, "relative": 0.000374 }, { "name": "video_fisheye_cam0", "l2": 5.469206, "mean_abs": 0.119589, "relative": 0.990356 }, { "name": "video_fisheye_cam1", "l2": 5.763362, "mean_abs": 0.130564, "relative": 0.952347 }, { "name": "video_fisheye_cam2", "l2": 6.264286, "mean_abs": 0.137134, "relative": 1.0 }, { "name": "video_fisheye_cam3", "l2": 5.515806, "mean_abs": 0.118694, "relative": 0.923555 }, { "name": "video_stereo_left", "l2": 6.051217, "mean_abs": 0.148009, "relative": 0.983335 }, { "name": "video_stereo_right", "l2": 6.702928, "mean_abs": 0.164773, "relative": 0.944972 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.063246, "mean_abs": 0.081023, "relative": 0.738405 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 90, "start_frame": 450, "end_frame": 469, "center_frame": 459, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.936555, "mean_abs": 0.22341, "relative": 0.838191 }, { "name": "hand_right_joints", "l2": 11.155736, "mean_abs": 0.325366, "relative": 0.87937 }, { "name": "body_joints", "l2": 6.064044, "mean_abs": 0.105515, "relative": 0.819066 }, { "name": "body_contacts", "l2": 3.463938, "mean_abs": 0.081727, "relative": 0.999578 }, { "name": "camera_translation", "l2": 0.25334, "mean_abs": 0.030439, "relative": 0.597848 }, { "name": "camera_rotation_matrix", "l2": 3.029462, "mean_abs": 0.218811, "relative": 0.930299 }, { "name": "imu_accel_gyro", "l2": 17.17942, "mean_abs": 1.142169, "relative": 0.926889 }, { "name": "depth_confidence", "l2": 10.576133, "mean_abs": 0.204866, "relative": 0.000351 }, { "name": "video_fisheye_cam0", "l2": 5.419169, "mean_abs": 0.116244, "relative": 0.981296 }, { "name": "video_fisheye_cam1", "l2": 5.723475, "mean_abs": 0.129904, "relative": 0.945756 }, { "name": "video_fisheye_cam2", "l2": 6.249023, "mean_abs": 0.134788, "relative": 0.997564 }, { "name": "video_fisheye_cam3", "l2": 5.59143, "mean_abs": 0.118728, "relative": 0.936218 }, { "name": "video_stereo_left", "l2": 5.982606, "mean_abs": 0.14407, "relative": 0.972185 }, { "name": "video_stereo_right", "l2": 6.741522, "mean_abs": 0.164194, "relative": 0.950413 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.045846, "mean_abs": 0.080017, "relative": 0.732177 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 91, "start_frame": 455, "end_frame": 474, "center_frame": 464, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.766309, "mean_abs": 0.219004, "relative": 0.82383 }, { "name": "hand_right_joints", "l2": 11.809031, "mean_abs": 0.321587, "relative": 0.930868 }, { "name": "body_joints", "l2": 6.077364, "mean_abs": 0.102685, "relative": 0.820865 }, { "name": "body_contacts", "l2": 3.463971, "mean_abs": 0.081728, "relative": 0.999587 }, { "name": "camera_translation", "l2": 0.195879, "mean_abs": 0.023501, "relative": 0.462248 }, { "name": "camera_rotation_matrix", "l2": 3.015275, "mean_abs": 0.214108, "relative": 0.925942 }, { "name": "imu_accel_gyro", "l2": 17.169195, "mean_abs": 1.139502, "relative": 0.926337 }, { "name": "depth_confidence", "l2": 10.337468, "mean_abs": 0.198098, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.402657, "mean_abs": 0.114622, "relative": 0.978306 }, { "name": "video_fisheye_cam1", "l2": 5.653211, "mean_abs": 0.125204, "relative": 0.934145 }, { "name": "video_fisheye_cam2", "l2": 6.201102, "mean_abs": 0.131358, "relative": 0.989914 }, { "name": "video_fisheye_cam3", "l2": 5.655606, "mean_abs": 0.11881, "relative": 0.946963 }, { "name": "video_stereo_left", "l2": 6.017744, "mean_abs": 0.141647, "relative": 0.977895 }, { "name": "video_stereo_right", "l2": 6.766898, "mean_abs": 0.161751, "relative": 0.953991 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.082897, "mean_abs": 0.080444, "relative": 0.745438 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 92, "start_frame": 460, "end_frame": 479, "center_frame": 469, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.698153, "mean_abs": 0.212531, "relative": 0.818081 }, { "name": "hand_right_joints", "l2": 12.11363, "mean_abs": 0.310364, "relative": 0.954878 }, { "name": "body_joints", "l2": 5.744175, "mean_abs": 0.091733, "relative": 0.775861 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.131575, "mean_abs": 0.017339, "relative": 0.310498 }, { "name": "camera_rotation_matrix", "l2": 3.005079, "mean_abs": 0.208221, "relative": 0.922811 }, { "name": "imu_accel_gyro", "l2": 17.153454, "mean_abs": 1.144488, "relative": 0.925488 }, { "name": "depth_confidence", "l2": 10.278064, "mean_abs": 0.193934, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.353557, "mean_abs": 0.112501, "relative": 0.969415 }, { "name": "video_fisheye_cam1", "l2": 5.586838, "mean_abs": 0.120888, "relative": 0.923178 }, { "name": "video_fisheye_cam2", "l2": 6.134326, "mean_abs": 0.128043, "relative": 0.979254 }, { "name": "video_fisheye_cam3", "l2": 5.714929, "mean_abs": 0.117919, "relative": 0.956896 }, { "name": "video_stereo_left", "l2": 5.958724, "mean_abs": 0.13589, "relative": 0.968304 }, { "name": "video_stereo_right", "l2": 6.831028, "mean_abs": 0.158987, "relative": 0.963032 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.042894, "mean_abs": 0.078797, "relative": 0.731121 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 93, "start_frame": 465, "end_frame": 484, "center_frame": 474, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.688286, "mean_abs": 0.212288, "relative": 0.817248 }, { "name": "hand_right_joints", "l2": 12.461512, "mean_abs": 0.31194, "relative": 0.9823 }, { "name": "body_joints", "l2": 5.728683, "mean_abs": 0.08813, "relative": 0.773769 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.077736, "mean_abs": 0.010962, "relative": 0.183446 }, { "name": "camera_rotation_matrix", "l2": 3.001484, "mean_abs": 0.202863, "relative": 0.921707 }, { "name": "imu_accel_gyro", "l2": 17.098068, "mean_abs": 1.101848, "relative": 0.9225 }, { "name": "depth_confidence", "l2": 10.230156, "mean_abs": 0.18905, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.348722, "mean_abs": 0.11182, "relative": 0.968539 }, { "name": "video_fisheye_cam1", "l2": 5.580662, "mean_abs": 0.118884, "relative": 0.922157 }, { "name": "video_fisheye_cam2", "l2": 6.062393, "mean_abs": 0.125781, "relative": 0.967771 }, { "name": "video_fisheye_cam3", "l2": 5.761827, "mean_abs": 0.117097, "relative": 0.964748 }, { "name": "video_stereo_left", "l2": 5.886772, "mean_abs": 0.131525, "relative": 0.956612 }, { "name": "video_stereo_right", "l2": 6.870307, "mean_abs": 0.155464, "relative": 0.968569 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.026125, "mean_abs": 0.079005, "relative": 0.72512 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 94, "start_frame": 470, "end_frame": 489, "center_frame": 479, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.682454, "mean_abs": 0.21301, "relative": 0.816756 }, { "name": "hand_right_joints", "l2": 12.51545, "mean_abs": 0.314365, "relative": 0.986552 }, { "name": "body_joints", "l2": 5.791797, "mean_abs": 0.087821, "relative": 0.782293 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030896, "mean_abs": 0.004009, "relative": 0.072911 }, { "name": "camera_rotation_matrix", "l2": 2.998723, "mean_abs": 0.200448, "relative": 0.920859 }, { "name": "imu_accel_gyro", "l2": 17.135416, "mean_abs": 1.104398, "relative": 0.924515 }, { "name": "depth_confidence", "l2": 10.215286, "mean_abs": 0.186832, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.326648, "mean_abs": 0.110953, "relative": 0.964542 }, { "name": "video_fisheye_cam1", "l2": 5.53871, "mean_abs": 0.116851, "relative": 0.915225 }, { "name": "video_fisheye_cam2", "l2": 5.989132, "mean_abs": 0.124005, "relative": 0.956076 }, { "name": "video_fisheye_cam3", "l2": 5.748318, "mean_abs": 0.116466, "relative": 0.962487 }, { "name": "video_stereo_left", "l2": 5.785494, "mean_abs": 0.128358, "relative": 0.940154 }, { "name": "video_stereo_right", "l2": 6.835861, "mean_abs": 0.151617, "relative": 0.963713 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.077752, "mean_abs": 0.083689, "relative": 0.743596 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 95, "start_frame": 475, "end_frame": 494, "center_frame": 484, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.783324, "mean_abs": 0.21492, "relative": 0.825265 }, { "name": "hand_right_joints", "l2": 12.287045, "mean_abs": 0.311941, "relative": 0.968548 }, { "name": "body_joints", "l2": 5.816522, "mean_abs": 0.088224, "relative": 0.785633 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.024281, "mean_abs": 0.003395, "relative": 0.0573 }, { "name": "camera_rotation_matrix", "l2": 3.000937, "mean_abs": 0.203943, "relative": 0.921539 }, { "name": "imu_accel_gyro", "l2": 17.134892, "mean_abs": 1.109464, "relative": 0.924486 }, { "name": "depth_confidence", "l2": 10.244753, "mean_abs": 0.188611, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.307057, "mean_abs": 0.110615, "relative": 0.960995 }, { "name": "video_fisheye_cam1", "l2": 5.519061, "mean_abs": 0.117439, "relative": 0.911978 }, { "name": "video_fisheye_cam2", "l2": 5.916439, "mean_abs": 0.122766, "relative": 0.944471 }, { "name": "video_fisheye_cam3", "l2": 5.715383, "mean_abs": 0.116196, "relative": 0.956972 }, { "name": "video_stereo_left", "l2": 5.703297, "mean_abs": 0.127484, "relative": 0.926797 }, { "name": "video_stereo_right", "l2": 6.813789, "mean_abs": 0.153129, "relative": 0.960601 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.98468, "mean_abs": 0.078428, "relative": 0.710287 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 96, "start_frame": 480, "end_frame": 499, "center_frame": 489, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.807031, "mean_abs": 0.218667, "relative": 0.827265 }, { "name": "hand_right_joints", "l2": 11.956083, "mean_abs": 0.314254, "relative": 0.942459 }, { "name": "body_joints", "l2": 5.872335, "mean_abs": 0.090231, "relative": 0.793172 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.041423, "mean_abs": 0.005999, "relative": 0.097751 }, { "name": "camera_rotation_matrix", "l2": 3.002778, "mean_abs": 0.205905, "relative": 0.922104 }, { "name": "imu_accel_gyro", "l2": 17.127781, "mean_abs": 1.10384, "relative": 0.924103 }, { "name": "depth_confidence", "l2": 10.325886, "mean_abs": 0.191979, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.30407, "mean_abs": 0.110675, "relative": 0.960454 }, { "name": "video_fisheye_cam1", "l2": 5.516648, "mean_abs": 0.117729, "relative": 0.91158 }, { "name": "video_fisheye_cam2", "l2": 5.846974, "mean_abs": 0.120704, "relative": 0.933382 }, { "name": "video_fisheye_cam3", "l2": 5.703043, "mean_abs": 0.116266, "relative": 0.954906 }, { "name": "video_stereo_left", "l2": 5.621793, "mean_abs": 0.126092, "relative": 0.913553 }, { "name": "video_stereo_right", "l2": 6.784604, "mean_abs": 0.154168, "relative": 0.956487 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.013299, "mean_abs": 0.080603, "relative": 0.720529 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 97, "start_frame": 485, "end_frame": 504, "center_frame": 494, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.822907, "mean_abs": 0.22669, "relative": 0.828604 }, { "name": "hand_right_joints", "l2": 11.45847, "mean_abs": 0.313865, "relative": 0.903234 }, { "name": "body_joints", "l2": 6.100079, "mean_abs": 0.09659, "relative": 0.823933 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.071901, "mean_abs": 0.010434, "relative": 0.169677 }, { "name": "camera_rotation_matrix", "l2": 3.002874, "mean_abs": 0.20743, "relative": 0.922134 }, { "name": "imu_accel_gyro", "l2": 17.107714, "mean_abs": 1.119391, "relative": 0.92302 }, { "name": "depth_confidence", "l2": 10.500002, "mean_abs": 0.19627, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.280761, "mean_abs": 0.110024, "relative": 0.956233 }, { "name": "video_fisheye_cam1", "l2": 5.549838, "mean_abs": 0.118807, "relative": 0.917064 }, { "name": "video_fisheye_cam2", "l2": 5.812126, "mean_abs": 0.119704, "relative": 0.927819 }, { "name": "video_fisheye_cam3", "l2": 5.696245, "mean_abs": 0.116149, "relative": 0.953767 }, { "name": "video_stereo_left", "l2": 5.585227, "mean_abs": 0.126553, "relative": 0.907611 }, { "name": "video_stereo_right", "l2": 6.776712, "mean_abs": 0.15535, "relative": 0.955374 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.997201, "mean_abs": 0.077418, "relative": 0.714768 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 98, "start_frame": 490, "end_frame": 509, "center_frame": 499, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.843537, "mean_abs": 0.2359, "relative": 0.830344 }, { "name": "hand_right_joints", "l2": 10.966224, "mean_abs": 0.31059, "relative": 0.864432 }, { "name": "body_joints", "l2": 6.312354, "mean_abs": 0.103346, "relative": 0.852604 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.115727, "mean_abs": 0.015024, "relative": 0.273099 }, { "name": "camera_rotation_matrix", "l2": 3.009083, "mean_abs": 0.214385, "relative": 0.924041 }, { "name": "imu_accel_gyro", "l2": 17.200552, "mean_abs": 1.151997, "relative": 0.928029 }, { "name": "depth_confidence", "l2": 10.695556, "mean_abs": 0.202122, "relative": 0.000355 }, { "name": "video_fisheye_cam0", "l2": 5.277456, "mean_abs": 0.110696, "relative": 0.955634 }, { "name": "video_fisheye_cam1", "l2": 5.647568, "mean_abs": 0.122758, "relative": 0.933213 }, { "name": "video_fisheye_cam2", "l2": 5.774354, "mean_abs": 0.12039, "relative": 0.92179 }, { "name": "video_fisheye_cam3", "l2": 5.738809, "mean_abs": 0.118053, "relative": 0.960894 }, { "name": "video_stereo_left", "l2": 5.556395, "mean_abs": 0.127842, "relative": 0.902925 }, { "name": "video_stereo_right", "l2": 6.898934, "mean_abs": 0.161687, "relative": 0.972605 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.062716, "mean_abs": 0.082043, "relative": 0.738215 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 99, "start_frame": 495, "end_frame": 514, "center_frame": 504, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.824121, "mean_abs": 0.242006, "relative": 0.828707 }, { "name": "hand_right_joints", "l2": 10.459208, "mean_abs": 0.296171, "relative": 0.824465 }, { "name": "body_joints", "l2": 6.25744, "mean_abs": 0.105588, "relative": 0.845187 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.145811, "mean_abs": 0.017988, "relative": 0.344094 }, { "name": "camera_rotation_matrix", "l2": 3.013925, "mean_abs": 0.21955, "relative": 0.925527 }, { "name": "imu_accel_gyro", "l2": 17.275017, "mean_abs": 1.174382, "relative": 0.932047 }, { "name": "depth_confidence", "l2": 10.840288, "mean_abs": 0.204109, "relative": 0.00036 }, { "name": "video_fisheye_cam0", "l2": 5.271238, "mean_abs": 0.111097, "relative": 0.954509 }, { "name": "video_fisheye_cam1", "l2": 5.687907, "mean_abs": 0.123946, "relative": 0.939879 }, { "name": "video_fisheye_cam2", "l2": 5.754229, "mean_abs": 0.120654, "relative": 0.918577 }, { "name": "video_fisheye_cam3", "l2": 5.747679, "mean_abs": 0.118443, "relative": 0.96238 }, { "name": "video_stereo_left", "l2": 5.531205, "mean_abs": 0.128101, "relative": 0.898832 }, { "name": "video_stereo_right", "l2": 6.956098, "mean_abs": 0.164706, "relative": 0.980664 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.05736, "mean_abs": 0.080411, "relative": 0.736298 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 100, "start_frame": 500, "end_frame": 519, "center_frame": 509, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.833373, "mean_abs": 0.24714, "relative": 0.829487 }, { "name": "hand_right_joints", "l2": 9.826668, "mean_abs": 0.27331, "relative": 0.774604 }, { "name": "body_joints", "l2": 6.214353, "mean_abs": 0.104313, "relative": 0.839368 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.154607, "mean_abs": 0.018062, "relative": 0.364851 }, { "name": "camera_rotation_matrix", "l2": 3.013996, "mean_abs": 0.222048, "relative": 0.925549 }, { "name": "imu_accel_gyro", "l2": 17.323833, "mean_abs": 1.15043, "relative": 0.93468 }, { "name": "depth_confidence", "l2": 10.961766, "mean_abs": 0.204931, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.266762, "mean_abs": 0.111315, "relative": 0.953698 }, { "name": "video_fisheye_cam1", "l2": 5.709484, "mean_abs": 0.12419, "relative": 0.943444 }, { "name": "video_fisheye_cam2", "l2": 5.758156, "mean_abs": 0.121175, "relative": 0.919204 }, { "name": "video_fisheye_cam3", "l2": 5.744044, "mean_abs": 0.118705, "relative": 0.961771 }, { "name": "video_stereo_left", "l2": 5.537259, "mean_abs": 0.127966, "relative": 0.899815 }, { "name": "video_stereo_right", "l2": 6.962233, "mean_abs": 0.164337, "relative": 0.981529 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.063122, "mean_abs": 0.079339, "relative": 0.73836 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010753, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 101, "start_frame": 505, "end_frame": 524, "center_frame": 514, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.816606, "mean_abs": 0.249088, "relative": 0.828073 }, { "name": "hand_right_joints", "l2": 9.388421, "mean_abs": 0.258114, "relative": 0.740059 }, { "name": "body_joints", "l2": 6.294756, "mean_abs": 0.104842, "relative": 0.850228 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.126334, "mean_abs": 0.014301, "relative": 0.298131 }, { "name": "camera_rotation_matrix", "l2": 3.008605, "mean_abs": 0.220933, "relative": 0.923894 }, { "name": "imu_accel_gyro", "l2": 17.140612, "mean_abs": 1.092522, "relative": 0.924795 }, { "name": "depth_confidence", "l2": 11.016444, "mean_abs": 0.203019, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.252494, "mean_abs": 0.110629, "relative": 0.951114 }, { "name": "video_fisheye_cam1", "l2": 5.737215, "mean_abs": 0.123934, "relative": 0.948026 }, { "name": "video_fisheye_cam2", "l2": 5.751907, "mean_abs": 0.120622, "relative": 0.918206 }, { "name": "video_fisheye_cam3", "l2": 5.761601, "mean_abs": 0.118247, "relative": 0.964711 }, { "name": "video_stereo_left", "l2": 5.555923, "mean_abs": 0.126709, "relative": 0.902849 }, { "name": "video_stereo_right", "l2": 6.969918, "mean_abs": 0.16186, "relative": 0.982612 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.071202, "mean_abs": 0.08038, "relative": 0.741252 }, { "name": "caption_objects_interaction_text", "l2": 1.885897, "mean_abs": 0.01617, "relative": 0.865958 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 102, "start_frame": 510, "end_frame": 529, "center_frame": 519, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.867949, "mean_abs": 0.252834, "relative": 0.832404 }, { "name": "hand_right_joints", "l2": 9.468637, "mean_abs": 0.259722, "relative": 0.746382 }, { "name": "body_joints", "l2": 6.270712, "mean_abs": 0.104024, "relative": 0.84698 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.09426, "mean_abs": 0.009629, "relative": 0.22244 }, { "name": "camera_rotation_matrix", "l2": 3.005052, "mean_abs": 0.219211, "relative": 0.922803 }, { "name": "imu_accel_gyro", "l2": 17.073086, "mean_abs": 1.013965, "relative": 0.921152 }, { "name": "depth_confidence", "l2": 11.211014, "mean_abs": 0.202764, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.237681, "mean_abs": 0.110156, "relative": 0.948432 }, { "name": "video_fisheye_cam1", "l2": 5.722368, "mean_abs": 0.123402, "relative": 0.945573 }, { "name": "video_fisheye_cam2", "l2": 5.73489, "mean_abs": 0.119942, "relative": 0.91549 }, { "name": "video_fisheye_cam3", "l2": 5.756468, "mean_abs": 0.117782, "relative": 0.963851 }, { "name": "video_stereo_left", "l2": 5.563255, "mean_abs": 0.127156, "relative": 0.90404 }, { "name": "video_stereo_right", "l2": 6.96335, "mean_abs": 0.158113, "relative": 0.981686 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.126179, "mean_abs": 0.082717, "relative": 0.760928 }, { "name": "caption_objects_interaction_text", "l2": 1.885897, "mean_abs": 0.016404, "relative": 0.865958 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 103, "start_frame": 515, "end_frame": 534, "center_frame": 524, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.908281, "mean_abs": 0.256134, "relative": 0.835806 }, { "name": "hand_right_joints", "l2": 9.655393, "mean_abs": 0.264422, "relative": 0.761103 }, { "name": "body_joints", "l2": 6.283326, "mean_abs": 0.104861, "relative": 0.848684 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.07664, "mean_abs": 0.007565, "relative": 0.18086 }, { "name": "camera_rotation_matrix", "l2": 3.005433, "mean_abs": 0.220961, "relative": 0.92292 }, { "name": "imu_accel_gyro", "l2": 17.073597, "mean_abs": 1.015454, "relative": 0.921179 }, { "name": "depth_confidence", "l2": 11.344398, "mean_abs": 0.203763, "relative": 0.000377 }, { "name": "video_fisheye_cam0", "l2": 5.232564, "mean_abs": 0.110093, "relative": 0.947506 }, { "name": "video_fisheye_cam1", "l2": 5.686709, "mean_abs": 0.122661, "relative": 0.939681 }, { "name": "video_fisheye_cam2", "l2": 5.753973, "mean_abs": 0.120154, "relative": 0.918536 }, { "name": "video_fisheye_cam3", "l2": 5.746554, "mean_abs": 0.117544, "relative": 0.962191 }, { "name": "video_stereo_left", "l2": 5.620628, "mean_abs": 0.12753, "relative": 0.913363 }, { "name": "video_stereo_right", "l2": 6.917145, "mean_abs": 0.157128, "relative": 0.975172 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.09073, "mean_abs": 0.082079, "relative": 0.748241 }, { "name": "caption_objects_interaction_text", "l2": 1.885897, "mean_abs": 0.016224, "relative": 0.865958 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 104, "start_frame": 520, "end_frame": 539, "center_frame": 529, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.965566, "mean_abs": 0.259499, "relative": 0.840638 }, { "name": "hand_right_joints", "l2": 9.788448, "mean_abs": 0.267285, "relative": 0.771592 }, { "name": "body_joints", "l2": 6.335717, "mean_abs": 0.106807, "relative": 0.85576 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.051862, "mean_abs": 0.005612, "relative": 0.122388 }, { "name": "camera_rotation_matrix", "l2": 3.003148, "mean_abs": 0.220062, "relative": 0.922218 }, { "name": "imu_accel_gyro", "l2": 17.062025, "mean_abs": 1.02122, "relative": 0.920555 }, { "name": "depth_confidence", "l2": 11.384209, "mean_abs": 0.203839, "relative": 0.000378 }, { "name": "video_fisheye_cam0", "l2": 5.222815, "mean_abs": 0.109509, "relative": 0.94574 }, { "name": "video_fisheye_cam1", "l2": 5.609775, "mean_abs": 0.120281, "relative": 0.926968 }, { "name": "video_fisheye_cam2", "l2": 5.747332, "mean_abs": 0.119557, "relative": 0.917476 }, { "name": "video_fisheye_cam3", "l2": 5.725714, "mean_abs": 0.116599, "relative": 0.958702 }, { "name": "video_stereo_left", "l2": 5.630773, "mean_abs": 0.12605, "relative": 0.915012 }, { "name": "video_stereo_right", "l2": 6.809862, "mean_abs": 0.152956, "relative": 0.960048 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.00433, "mean_abs": 0.077134, "relative": 0.71732 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 105, "start_frame": 525, "end_frame": 544, "center_frame": 534, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.997371, "mean_abs": 0.260627, "relative": 0.843321 }, { "name": "hand_right_joints", "l2": 9.849048, "mean_abs": 0.268829, "relative": 0.776368 }, { "name": "body_joints", "l2": 6.340965, "mean_abs": 0.106955, "relative": 0.856469 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030732, "mean_abs": 0.003905, "relative": 0.072523 }, { "name": "camera_rotation_matrix", "l2": 3.001404, "mean_abs": 0.218323, "relative": 0.921682 }, { "name": "imu_accel_gyro", "l2": 17.002266, "mean_abs": 1.032017, "relative": 0.917331 }, { "name": "depth_confidence", "l2": 11.443344, "mean_abs": 0.202935, "relative": 0.00038 }, { "name": "video_fisheye_cam0", "l2": 5.219286, "mean_abs": 0.108887, "relative": 0.945101 }, { "name": "video_fisheye_cam1", "l2": 5.590075, "mean_abs": 0.118933, "relative": 0.923713 }, { "name": "video_fisheye_cam2", "l2": 5.748449, "mean_abs": 0.118979, "relative": 0.917654 }, { "name": "video_fisheye_cam3", "l2": 5.724946, "mean_abs": 0.116263, "relative": 0.958573 }, { "name": "video_stereo_left", "l2": 5.639104, "mean_abs": 0.124721, "relative": 0.916366 }, { "name": "video_stereo_right", "l2": 6.7801, "mean_abs": 0.150151, "relative": 0.955852 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.117017, "mean_abs": 0.081843, "relative": 0.757649 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 106, "start_frame": 530, "end_frame": 549, "center_frame": 539, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.113571, "mean_abs": 0.263697, "relative": 0.853123 }, { "name": "hand_right_joints", "l2": 9.951229, "mean_abs": 0.270796, "relative": 0.784423 }, { "name": "body_joints", "l2": 6.337948, "mean_abs": 0.106796, "relative": 0.856062 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.037394, "mean_abs": 0.005157, "relative": 0.088244 }, { "name": "camera_rotation_matrix", "l2": 3.001831, "mean_abs": 0.218624, "relative": 0.921814 }, { "name": "imu_accel_gyro", "l2": 16.97575, "mean_abs": 1.06214, "relative": 0.9159 }, { "name": "depth_confidence", "l2": 11.421874, "mean_abs": 0.20329, "relative": 0.000379 }, { "name": "video_fisheye_cam0", "l2": 5.243276, "mean_abs": 0.110021, "relative": 0.949445 }, { "name": "video_fisheye_cam1", "l2": 5.542382, "mean_abs": 0.11828, "relative": 0.915832 }, { "name": "video_fisheye_cam2", "l2": 5.732165, "mean_abs": 0.118986, "relative": 0.915055 }, { "name": "video_fisheye_cam3", "l2": 5.741272, "mean_abs": 0.11635, "relative": 0.961307 }, { "name": "video_stereo_left", "l2": 5.604009, "mean_abs": 0.124654, "relative": 0.910663 }, { "name": "video_stereo_right", "l2": 6.656226, "mean_abs": 0.148362, "relative": 0.938388 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.070384, "mean_abs": 0.080017, "relative": 0.740959 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 107, "start_frame": 535, "end_frame": 554, "center_frame": 544, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.273396, "mean_abs": 0.270267, "relative": 0.866605 }, { "name": "hand_right_joints", "l2": 10.065986, "mean_abs": 0.276354, "relative": 0.793469 }, { "name": "body_joints", "l2": 6.388851, "mean_abs": 0.107907, "relative": 0.862937 }, { "name": "body_contacts", "l2": 3.464071, "mean_abs": 0.081718, "relative": 0.999616 }, { "name": "camera_translation", "l2": 0.142087, "mean_abs": 0.018854, "relative": 0.335305 }, { "name": "camera_rotation_matrix", "l2": 3.010423, "mean_abs": 0.224855, "relative": 0.924452 }, { "name": "imu_accel_gyro", "l2": 16.747583, "mean_abs": 1.056185, "relative": 0.90359 }, { "name": "depth_confidence", "l2": 11.204535, "mean_abs": 0.205617, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.282181, "mean_abs": 0.112048, "relative": 0.95649 }, { "name": "video_fisheye_cam1", "l2": 5.480764, "mean_abs": 0.120173, "relative": 0.90565 }, { "name": "video_fisheye_cam2", "l2": 5.701364, "mean_abs": 0.120916, "relative": 0.910138 }, { "name": "video_fisheye_cam3", "l2": 5.78133, "mean_abs": 0.1186, "relative": 0.968014 }, { "name": "video_stereo_left", "l2": 5.562187, "mean_abs": 0.129013, "relative": 0.903866 }, { "name": "video_stereo_right", "l2": 6.561394, "mean_abs": 0.153147, "relative": 0.925019 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.093086, "mean_abs": 0.080155, "relative": 0.749084 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 108, "start_frame": 540, "end_frame": 559, "center_frame": 549, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.449704, "mean_abs": 0.27727, "relative": 0.881477 }, { "name": "hand_right_joints", "l2": 10.225061, "mean_abs": 0.283187, "relative": 0.806008 }, { "name": "body_joints", "l2": 6.479847, "mean_abs": 0.111656, "relative": 0.875228 }, { "name": "body_contacts", "l2": 3.464035, "mean_abs": 0.081722, "relative": 0.999605 }, { "name": "camera_translation", "l2": 0.267152, "mean_abs": 0.038411, "relative": 0.630441 }, { "name": "camera_rotation_matrix", "l2": 3.018507, "mean_abs": 0.226204, "relative": 0.926934 }, { "name": "imu_accel_gyro", "l2": 16.909693, "mean_abs": 1.134033, "relative": 0.912336 }, { "name": "depth_confidence", "l2": 10.966587, "mean_abs": 0.207626, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.317987, "mean_abs": 0.11359, "relative": 0.962974 }, { "name": "video_fisheye_cam1", "l2": 5.388313, "mean_abs": 0.119974, "relative": 0.890373 }, { "name": "video_fisheye_cam2", "l2": 5.639268, "mean_abs": 0.121465, "relative": 0.900225 }, { "name": "video_fisheye_cam3", "l2": 5.870654, "mean_abs": 0.122541, "relative": 0.98297 }, { "name": "video_stereo_left", "l2": 5.542181, "mean_abs": 0.131813, "relative": 0.900616 }, { "name": "video_stereo_right", "l2": 6.48312, "mean_abs": 0.155356, "relative": 0.913984 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.143344, "mean_abs": 0.083442, "relative": 0.767071 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 109, "start_frame": 545, "end_frame": 564, "center_frame": 554, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.651754, "mean_abs": 0.283479, "relative": 0.898521 }, { "name": "hand_right_joints", "l2": 10.378881, "mean_abs": 0.285752, "relative": 0.818133 }, { "name": "body_joints", "l2": 6.568633, "mean_abs": 0.115496, "relative": 0.88722 }, { "name": "body_contacts", "l2": 3.464016, "mean_abs": 0.08172, "relative": 0.9996 }, { "name": "camera_translation", "l2": 0.370627, "mean_abs": 0.055704, "relative": 0.874629 }, { "name": "camera_rotation_matrix", "l2": 3.028418, "mean_abs": 0.224491, "relative": 0.929978 }, { "name": "imu_accel_gyro", "l2": 17.33251, "mean_abs": 1.208383, "relative": 0.935149 }, { "name": "depth_confidence", "l2": 10.657198, "mean_abs": 0.20636, "relative": 0.000354 }, { "name": "video_fisheye_cam0", "l2": 5.340589, "mean_abs": 0.114249, "relative": 0.967066 }, { "name": "video_fisheye_cam1", "l2": 5.300688, "mean_abs": 0.119413, "relative": 0.875894 }, { "name": "video_fisheye_cam2", "l2": 5.580839, "mean_abs": 0.121663, "relative": 0.890898 }, { "name": "video_fisheye_cam3", "l2": 5.949798, "mean_abs": 0.126798, "relative": 0.996222 }, { "name": "video_stereo_left", "l2": 5.594302, "mean_abs": 0.134145, "relative": 0.909085 }, { "name": "video_stereo_right", "l2": 6.44305, "mean_abs": 0.156572, "relative": 0.908335 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.14805, "mean_abs": 0.082911, "relative": 0.768755 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 110, "start_frame": 550, "end_frame": 569, "center_frame": 559, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.803373, "mean_abs": 0.283422, "relative": 0.911311 }, { "name": "hand_right_joints", "l2": 10.563405, "mean_abs": 0.287255, "relative": 0.832679 }, { "name": "body_joints", "l2": 6.567856, "mean_abs": 0.114576, "relative": 0.887115 }, { "name": "body_contacts", "l2": 3.464011, "mean_abs": 0.0817, "relative": 0.999599 }, { "name": "camera_translation", "l2": 0.403252, "mean_abs": 0.061611, "relative": 0.951617 }, { "name": "camera_rotation_matrix", "l2": 3.032664, "mean_abs": 0.21923, "relative": 0.931282 }, { "name": "imu_accel_gyro", "l2": 17.443836, "mean_abs": 1.253408, "relative": 0.941155 }, { "name": "depth_confidence", "l2": 10.34037, "mean_abs": 0.202479, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.367254, "mean_abs": 0.114254, "relative": 0.971895 }, { "name": "video_fisheye_cam1", "l2": 5.177242, "mean_abs": 0.116348, "relative": 0.855496 }, { "name": "video_fisheye_cam2", "l2": 5.530041, "mean_abs": 0.120334, "relative": 0.882789 }, { "name": "video_fisheye_cam3", "l2": 5.972362, "mean_abs": 0.128501, "relative": 1.0 }, { "name": "video_stereo_left", "l2": 5.488927, "mean_abs": 0.131162, "relative": 0.891962 }, { "name": "video_stereo_right", "l2": 6.247217, "mean_abs": 0.15143, "relative": 0.880727 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.498771, "mean_abs": 0.107565, "relative": 0.894273 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 111, "start_frame": 555, "end_frame": 574, "center_frame": 564, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.936306, "mean_abs": 0.278112, "relative": 0.922524 }, { "name": "hand_right_joints", "l2": 10.654538, "mean_abs": 0.283048, "relative": 0.839863 }, { "name": "body_joints", "l2": 6.481626, "mean_abs": 0.109592, "relative": 0.875468 }, { "name": "body_contacts", "l2": 3.464022, "mean_abs": 0.081722, "relative": 0.999602 }, { "name": "camera_translation", "l2": 0.341859, "mean_abs": 0.05194, "relative": 0.80674 }, { "name": "camera_rotation_matrix", "l2": 3.031816, "mean_abs": 0.216151, "relative": 0.931022 }, { "name": "imu_accel_gyro", "l2": 17.759787, "mean_abs": 1.273807, "relative": 0.958202 }, { "name": "depth_confidence", "l2": 9.65774, "mean_abs": 0.191047, "relative": 0.000321 }, { "name": "video_fisheye_cam0", "l2": 5.381154, "mean_abs": 0.113676, "relative": 0.974412 }, { "name": "video_fisheye_cam1", "l2": 5.023357, "mean_abs": 0.112903, "relative": 0.830067 }, { "name": "video_fisheye_cam2", "l2": 5.455436, "mean_abs": 0.116732, "relative": 0.870879 }, { "name": "video_fisheye_cam3", "l2": 5.918245, "mean_abs": 0.128091, "relative": 0.990939 }, { "name": "video_stereo_left", "l2": 5.323811, "mean_abs": 0.124921, "relative": 0.86513 }, { "name": "video_stereo_right", "l2": 5.955779, "mean_abs": 0.142691, "relative": 0.83964 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.384065, "mean_abs": 0.096479, "relative": 0.853221 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 112, "start_frame": 560, "end_frame": 579, "center_frame": 569, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.089306, "mean_abs": 0.277534, "relative": 0.93543 }, { "name": "hand_right_joints", "l2": 10.777754, "mean_abs": 0.276768, "relative": 0.849575 }, { "name": "body_joints", "l2": 6.458738, "mean_abs": 0.106062, "relative": 0.872377 }, { "name": "body_contacts", "l2": 3.464041, "mean_abs": 0.081723, "relative": 0.999607 }, { "name": "camera_translation", "l2": 0.223527, "mean_abs": 0.033317, "relative": 0.527492 }, { "name": "camera_rotation_matrix", "l2": 3.016302, "mean_abs": 0.210228, "relative": 0.926257 }, { "name": "imu_accel_gyro", "l2": 17.461729, "mean_abs": 1.196164, "relative": 0.94212 }, { "name": "depth_confidence", "l2": 9.222869, "mean_abs": 0.179276, "relative": 0.000306 }, { "name": "video_fisheye_cam0", "l2": 5.382441, "mean_abs": 0.11306, "relative": 0.974645 }, { "name": "video_fisheye_cam1", "l2": 4.948391, "mean_abs": 0.110439, "relative": 0.81768 }, { "name": "video_fisheye_cam2", "l2": 5.368268, "mean_abs": 0.112298, "relative": 0.856964 }, { "name": "video_fisheye_cam3", "l2": 5.856348, "mean_abs": 0.126132, "relative": 0.980575 }, { "name": "video_stereo_left", "l2": 5.13807, "mean_abs": 0.117241, "relative": 0.834947 }, { "name": "video_stereo_right", "l2": 5.756546, "mean_abs": 0.134692, "relative": 0.811552 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.305899, "mean_abs": 0.093557, "relative": 0.825247 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 113, "start_frame": 565, "end_frame": 584, "center_frame": 574, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.166062, "mean_abs": 0.276556, "relative": 0.941905 }, { "name": "hand_right_joints", "l2": 10.831981, "mean_abs": 0.276241, "relative": 0.85385 }, { "name": "body_joints", "l2": 6.468808, "mean_abs": 0.103342, "relative": 0.873737 }, { "name": "body_contacts", "l2": 3.464002, "mean_abs": 0.081715, "relative": 0.999596 }, { "name": "camera_translation", "l2": 0.128465, "mean_abs": 0.018616, "relative": 0.30316 }, { "name": "camera_rotation_matrix", "l2": 3.009571, "mean_abs": 0.207676, "relative": 0.92419 }, { "name": "imu_accel_gyro", "l2": 17.408613, "mean_abs": 1.161837, "relative": 0.939255 }, { "name": "depth_confidence", "l2": 8.927107, "mean_abs": 0.169544, "relative": 0.000296 }, { "name": "video_fisheye_cam0", "l2": 5.364869, "mean_abs": 0.112378, "relative": 0.971463 }, { "name": "video_fisheye_cam1", "l2": 4.931551, "mean_abs": 0.109633, "relative": 0.814897 }, { "name": "video_fisheye_cam2", "l2": 5.363618, "mean_abs": 0.111494, "relative": 0.856222 }, { "name": "video_fisheye_cam3", "l2": 5.759899, "mean_abs": 0.122731, "relative": 0.964426 }, { "name": "video_stereo_left", "l2": 5.041032, "mean_abs": 0.113159, "relative": 0.819178 }, { "name": "video_stereo_right", "l2": 5.682139, "mean_abs": 0.130644, "relative": 0.801062 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.215972, "mean_abs": 0.089722, "relative": 0.793063 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 114, "start_frame": 570, "end_frame": 589, "center_frame": 579, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.156576, "mean_abs": 0.275294, "relative": 0.941105 }, { "name": "hand_right_joints", "l2": 10.840862, "mean_abs": 0.278153, "relative": 0.85455 }, { "name": "body_joints", "l2": 6.457403, "mean_abs": 0.102349, "relative": 0.872196 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.082924, "mean_abs": 0.012568, "relative": 0.195688 }, { "name": "camera_rotation_matrix", "l2": 3.00437, "mean_abs": 0.205709, "relative": 0.922593 }, { "name": "imu_accel_gyro", "l2": 17.433035, "mean_abs": 1.164289, "relative": 0.940572 }, { "name": "depth_confidence", "l2": 8.712956, "mean_abs": 0.163825, "relative": 0.000289 }, { "name": "video_fisheye_cam0", "l2": 5.357603, "mean_abs": 0.11213, "relative": 0.970147 }, { "name": "video_fisheye_cam1", "l2": 4.964347, "mean_abs": 0.109988, "relative": 0.820316 }, { "name": "video_fisheye_cam2", "l2": 5.376548, "mean_abs": 0.111273, "relative": 0.858286 }, { "name": "video_fisheye_cam3", "l2": 5.670056, "mean_abs": 0.118516, "relative": 0.949383 }, { "name": "video_stereo_left", "l2": 4.995702, "mean_abs": 0.111854, "relative": 0.811811 }, { "name": "video_stereo_right", "l2": 5.697692, "mean_abs": 0.131901, "relative": 0.803255 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.428346, "mean_abs": 0.097511, "relative": 0.869069 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 115, "start_frame": 575, "end_frame": 594, "center_frame": 584, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.18481, "mean_abs": 0.275339, "relative": 0.943486 }, { "name": "hand_right_joints", "l2": 10.837984, "mean_abs": 0.277579, "relative": 0.854323 }, { "name": "body_joints", "l2": 6.442496, "mean_abs": 0.101864, "relative": 0.870183 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.058432, "mean_abs": 0.008153, "relative": 0.137892 }, { "name": "camera_rotation_matrix", "l2": 3.001732, "mean_abs": 0.203364, "relative": 0.921783 }, { "name": "imu_accel_gyro", "l2": 17.203424, "mean_abs": 1.132919, "relative": 0.928184 }, { "name": "depth_confidence", "l2": 8.565535, "mean_abs": 0.15804, "relative": 0.000284 }, { "name": "video_fisheye_cam0", "l2": 5.351473, "mean_abs": 0.111428, "relative": 0.969037 }, { "name": "video_fisheye_cam1", "l2": 5.015015, "mean_abs": 0.110051, "relative": 0.828689 }, { "name": "video_fisheye_cam2", "l2": 5.395318, "mean_abs": 0.110841, "relative": 0.861282 }, { "name": "video_fisheye_cam3", "l2": 5.635507, "mean_abs": 0.116994, "relative": 0.943598 }, { "name": "video_stereo_left", "l2": 4.990867, "mean_abs": 0.11052, "relative": 0.811026 }, { "name": "video_stereo_right", "l2": 5.75145, "mean_abs": 0.131202, "relative": 0.810834 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.123313, "mean_abs": 0.084163, "relative": 0.759902 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 116, "start_frame": 580, "end_frame": 599, "center_frame": 589, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.187287, "mean_abs": 0.274793, "relative": 0.943695 }, { "name": "hand_right_joints", "l2": 10.888114, "mean_abs": 0.279464, "relative": 0.858275 }, { "name": "body_joints", "l2": 6.4058, "mean_abs": 0.1018, "relative": 0.865226 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.035356, "mean_abs": 0.005192, "relative": 0.083434 }, { "name": "camera_rotation_matrix", "l2": 3.001817, "mean_abs": 0.202971, "relative": 0.921809 }, { "name": "imu_accel_gyro", "l2": 16.948021, "mean_abs": 1.12938, "relative": 0.914404 }, { "name": "depth_confidence", "l2": 8.564582, "mean_abs": 0.158547, "relative": 0.000284 }, { "name": "video_fisheye_cam0", "l2": 5.345937, "mean_abs": 0.110841, "relative": 0.968035 }, { "name": "video_fisheye_cam1", "l2": 5.091703, "mean_abs": 0.109597, "relative": 0.841361 }, { "name": "video_fisheye_cam2", "l2": 5.413656, "mean_abs": 0.110893, "relative": 0.86421 }, { "name": "video_fisheye_cam3", "l2": 5.62483, "mean_abs": 0.116546, "relative": 0.94181 }, { "name": "video_stereo_left", "l2": 5.015337, "mean_abs": 0.111023, "relative": 0.815002 }, { "name": "video_stereo_right", "l2": 5.815053, "mean_abs": 0.131368, "relative": 0.8198 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.398628, "mean_abs": 0.101929, "relative": 0.858433 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 117, "start_frame": 585, "end_frame": 604, "center_frame": 594, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.178213, "mean_abs": 0.275557, "relative": 0.94293 }, { "name": "hand_right_joints", "l2": 10.942972, "mean_abs": 0.280947, "relative": 0.862599 }, { "name": "body_joints", "l2": 6.340517, "mean_abs": 0.100705, "relative": 0.856408 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01638, "mean_abs": 0.002583, "relative": 0.038653 }, { "name": "camera_rotation_matrix", "l2": 2.999718, "mean_abs": 0.202097, "relative": 0.921165 }, { "name": "imu_accel_gyro", "l2": 16.942812, "mean_abs": 1.113173, "relative": 0.914123 }, { "name": "depth_confidence", "l2": 8.513356, "mean_abs": 0.156842, "relative": 0.000283 }, { "name": "video_fisheye_cam0", "l2": 5.341616, "mean_abs": 0.110407, "relative": 0.967252 }, { "name": "video_fisheye_cam1", "l2": 5.128593, "mean_abs": 0.109084, "relative": 0.847457 }, { "name": "video_fisheye_cam2", "l2": 5.437157, "mean_abs": 0.110547, "relative": 0.867961 }, { "name": "video_fisheye_cam3", "l2": 5.611302, "mean_abs": 0.116157, "relative": 0.939545 }, { "name": "video_stereo_left", "l2": 5.028972, "mean_abs": 0.110175, "relative": 0.817218 }, { "name": "video_stereo_right", "l2": 5.824979, "mean_abs": 0.13013, "relative": 0.8212 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.252822, "mean_abs": 0.093041, "relative": 0.806251 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 118, "start_frame": 590, "end_frame": 609, "center_frame": 599, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.192707, "mean_abs": 0.274459, "relative": 0.944153 }, { "name": "hand_right_joints", "l2": 10.995321, "mean_abs": 0.28244, "relative": 0.866725 }, { "name": "body_joints", "l2": 6.286346, "mean_abs": 0.100103, "relative": 0.849092 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011088, "mean_abs": 0.001784, "relative": 0.026166 }, { "name": "camera_rotation_matrix", "l2": 3.000509, "mean_abs": 0.201635, "relative": 0.921408 }, { "name": "imu_accel_gyro", "l2": 16.974266, "mean_abs": 1.115549, "relative": 0.91582 }, { "name": "depth_confidence", "l2": 8.552757, "mean_abs": 0.156358, "relative": 0.000284 }, { "name": "video_fisheye_cam0", "l2": 5.337258, "mean_abs": 0.110148, "relative": 0.966463 }, { "name": "video_fisheye_cam1", "l2": 5.144979, "mean_abs": 0.109155, "relative": 0.850164 }, { "name": "video_fisheye_cam2", "l2": 5.456796, "mean_abs": 0.11066, "relative": 0.871096 }, { "name": "video_fisheye_cam3", "l2": 5.604933, "mean_abs": 0.116377, "relative": 0.938478 }, { "name": "video_stereo_left", "l2": 5.060133, "mean_abs": 0.110678, "relative": 0.822282 }, { "name": "video_stereo_right", "l2": 5.824832, "mean_abs": 0.129697, "relative": 0.821179 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.484583, "mean_abs": 0.120266, "relative": 0.889195 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 119, "start_frame": 595, "end_frame": 614, "center_frame": 604, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.267663, "mean_abs": 0.274206, "relative": 0.950475 }, { "name": "hand_right_joints", "l2": 11.055633, "mean_abs": 0.281954, "relative": 0.87148 }, { "name": "body_joints", "l2": 6.230209, "mean_abs": 0.098694, "relative": 0.841509 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01222, "mean_abs": 0.001802, "relative": 0.028837 }, { "name": "camera_rotation_matrix", "l2": 3.000304, "mean_abs": 0.201528, "relative": 0.921345 }, { "name": "imu_accel_gyro", "l2": 17.010504, "mean_abs": 1.139275, "relative": 0.917775 }, { "name": "depth_confidence", "l2": 8.583853, "mean_abs": 0.156986, "relative": 0.000285 }, { "name": "video_fisheye_cam0", "l2": 5.333827, "mean_abs": 0.10998, "relative": 0.965842 }, { "name": "video_fisheye_cam1", "l2": 5.150518, "mean_abs": 0.109245, "relative": 0.85108 }, { "name": "video_fisheye_cam2", "l2": 5.46921, "mean_abs": 0.110994, "relative": 0.873078 }, { "name": "video_fisheye_cam3", "l2": 5.571918, "mean_abs": 0.114788, "relative": 0.93295 }, { "name": "video_stereo_left", "l2": 5.076952, "mean_abs": 0.111163, "relative": 0.825015 }, { "name": "video_stereo_right", "l2": 5.824545, "mean_abs": 0.129628, "relative": 0.821139 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.307297, "mean_abs": 0.107005, "relative": 0.825747 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 120, "start_frame": 600, "end_frame": 619, "center_frame": 609, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.279126, "mean_abs": 0.274131, "relative": 0.951442 }, { "name": "hand_right_joints", "l2": 11.134337, "mean_abs": 0.281719, "relative": 0.877684 }, { "name": "body_joints", "l2": 6.176323, "mean_abs": 0.096815, "relative": 0.834231 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.051255, "mean_abs": 0.0066, "relative": 0.120954 }, { "name": "camera_rotation_matrix", "l2": 3.003065, "mean_abs": 0.205546, "relative": 0.922193 }, { "name": "imu_accel_gyro", "l2": 17.124054, "mean_abs": 1.1779, "relative": 0.923902 }, { "name": "depth_confidence", "l2": 8.807841, "mean_abs": 0.163937, "relative": 0.000292 }, { "name": "video_fisheye_cam0", "l2": 5.336911, "mean_abs": 0.110945, "relative": 0.966401 }, { "name": "video_fisheye_cam1", "l2": 5.169174, "mean_abs": 0.111684, "relative": 0.854162 }, { "name": "video_fisheye_cam2", "l2": 5.490876, "mean_abs": 0.112283, "relative": 0.876537 }, { "name": "video_fisheye_cam3", "l2": 5.569293, "mean_abs": 0.116109, "relative": 0.932511 }, { "name": "video_stereo_left", "l2": 5.13199, "mean_abs": 0.114026, "relative": 0.833959 }, { "name": "video_stereo_right", "l2": 5.879123, "mean_abs": 0.132406, "relative": 0.828833 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.512742, "mean_abs": 0.118028, "relative": 0.899273 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 121, "start_frame": 605, "end_frame": 624, "center_frame": 614, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.381538, "mean_abs": 0.277282, "relative": 0.960081 }, { "name": "hand_right_joints", "l2": 11.131261, "mean_abs": 0.279597, "relative": 0.877441 }, { "name": "body_joints", "l2": 6.090454, "mean_abs": 0.096653, "relative": 0.822633 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.115527, "mean_abs": 0.016014, "relative": 0.272628 }, { "name": "camera_rotation_matrix", "l2": 3.014278, "mean_abs": 0.213077, "relative": 0.925636 }, { "name": "imu_accel_gyro", "l2": 17.164305, "mean_abs": 1.208732, "relative": 0.926073 }, { "name": "depth_confidence", "l2": 8.940869, "mean_abs": 0.170461, "relative": 0.000297 }, { "name": "video_fisheye_cam0", "l2": 5.361754, "mean_abs": 0.113065, "relative": 0.970899 }, { "name": "video_fisheye_cam1", "l2": 5.239713, "mean_abs": 0.116351, "relative": 0.865818 }, { "name": "video_fisheye_cam2", "l2": 5.514604, "mean_abs": 0.115231, "relative": 0.880324 }, { "name": "video_fisheye_cam3", "l2": 5.590899, "mean_abs": 0.119054, "relative": 0.936129 }, { "name": "video_stereo_left", "l2": 5.253328, "mean_abs": 0.11866, "relative": 0.853676 }, { "name": "video_stereo_right", "l2": 6.003406, "mean_abs": 0.137517, "relative": 0.846354 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.353846, "mean_abs": 0.112278, "relative": 0.842406 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 122, "start_frame": 610, "end_frame": 629, "center_frame": 619, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.504617, "mean_abs": 0.278249, "relative": 0.970464 }, { "name": "hand_right_joints", "l2": 11.116979, "mean_abs": 0.279953, "relative": 0.876315 }, { "name": "body_joints", "l2": 6.007525, "mean_abs": 0.097034, "relative": 0.811432 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.165613, "mean_abs": 0.023361, "relative": 0.390823 }, { "name": "camera_rotation_matrix", "l2": 3.024674, "mean_abs": 0.216784, "relative": 0.928828 }, { "name": "imu_accel_gyro", "l2": 17.136183, "mean_abs": 1.226815, "relative": 0.924556 }, { "name": "depth_confidence", "l2": 9.098805, "mean_abs": 0.176874, "relative": 0.000302 }, { "name": "video_fisheye_cam0", "l2": 5.380961, "mean_abs": 0.114138, "relative": 0.974377 }, { "name": "video_fisheye_cam1", "l2": 5.286442, "mean_abs": 0.118741, "relative": 0.87354 }, { "name": "video_fisheye_cam2", "l2": 5.552818, "mean_abs": 0.117632, "relative": 0.886425 }, { "name": "video_fisheye_cam3", "l2": 5.580092, "mean_abs": 0.120071, "relative": 0.934319 }, { "name": "video_stereo_left", "l2": 5.306407, "mean_abs": 0.121163, "relative": 0.862302 }, { "name": "video_stereo_right", "l2": 6.064005, "mean_abs": 0.141394, "relative": 0.854897 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.345711, "mean_abs": 0.100683, "relative": 0.839495 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 123, "start_frame": 615, "end_frame": 634, "center_frame": 624, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.516214, "mean_abs": 0.276155, "relative": 0.971442 }, { "name": "hand_right_joints", "l2": 11.229328, "mean_abs": 0.282956, "relative": 0.885171 }, { "name": "body_joints", "l2": 5.927292, "mean_abs": 0.095367, "relative": 0.800595 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.190787, "mean_abs": 0.027857, "relative": 0.450229 }, { "name": "camera_rotation_matrix", "l2": 3.034657, "mean_abs": 0.216746, "relative": 0.931894 }, { "name": "imu_accel_gyro", "l2": 17.312706, "mean_abs": 1.246421, "relative": 0.93408 }, { "name": "depth_confidence", "l2": 9.202257, "mean_abs": 0.17972, "relative": 0.000305 }, { "name": "video_fisheye_cam0", "l2": 5.38724, "mean_abs": 0.114934, "relative": 0.975514 }, { "name": "video_fisheye_cam1", "l2": 5.309989, "mean_abs": 0.119322, "relative": 0.877431 }, { "name": "video_fisheye_cam2", "l2": 5.629524, "mean_abs": 0.120348, "relative": 0.89867 }, { "name": "video_fisheye_cam3", "l2": 5.536556, "mean_abs": 0.118964, "relative": 0.927029 }, { "name": "video_stereo_left", "l2": 5.368051, "mean_abs": 0.123294, "relative": 0.872319 }, { "name": "video_stereo_right", "l2": 6.135629, "mean_abs": 0.143797, "relative": 0.864995 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.275351, "mean_abs": 0.098417, "relative": 0.814314 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 124, "start_frame": 620, "end_frame": 639, "center_frame": 629, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.588333, "mean_abs": 0.268327, "relative": 0.977525 }, { "name": "hand_right_joints", "l2": 11.206472, "mean_abs": 0.274409, "relative": 0.88337 }, { "name": "body_joints", "l2": 5.83086, "mean_abs": 0.09566, "relative": 0.78757 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.163795, "mean_abs": 0.024792, "relative": 0.386534 }, { "name": "camera_rotation_matrix", "l2": 3.034757, "mean_abs": 0.211187, "relative": 0.931925 }, { "name": "imu_accel_gyro", "l2": 17.030239, "mean_abs": 1.182315, "relative": 0.91884 }, { "name": "depth_confidence", "l2": 9.251893, "mean_abs": 0.18086, "relative": 0.000307 }, { "name": "video_fisheye_cam0", "l2": 5.386185, "mean_abs": 0.115343, "relative": 0.975323 }, { "name": "video_fisheye_cam1", "l2": 5.335394, "mean_abs": 0.119314, "relative": 0.881629 }, { "name": "video_fisheye_cam2", "l2": 5.656529, "mean_abs": 0.121523, "relative": 0.902981 }, { "name": "video_fisheye_cam3", "l2": 5.521863, "mean_abs": 0.118311, "relative": 0.924569 }, { "name": "video_stereo_left", "l2": 5.441608, "mean_abs": 0.124383, "relative": 0.884272 }, { "name": "video_stereo_right", "l2": 6.23705, "mean_abs": 0.145731, "relative": 0.879293 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.08105, "mean_abs": 0.085304, "relative": 0.744777 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 125, "start_frame": 625, "end_frame": 644, "center_frame": 634, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.558399, "mean_abs": 0.265626, "relative": 0.975 }, { "name": "hand_right_joints", "l2": 11.230816, "mean_abs": 0.267759, "relative": 0.885289 }, { "name": "body_joints", "l2": 5.612336, "mean_abs": 0.088566, "relative": 0.758054 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.117501, "mean_abs": 0.017504, "relative": 0.277287 }, { "name": "camera_rotation_matrix", "l2": 3.02143, "mean_abs": 0.20247, "relative": 0.927832 }, { "name": "imu_accel_gyro", "l2": 17.029928, "mean_abs": 1.146775, "relative": 0.918823 }, { "name": "depth_confidence", "l2": 9.308679, "mean_abs": 0.181286, "relative": 0.000309 }, { "name": "video_fisheye_cam0", "l2": 5.330482, "mean_abs": 0.11366, "relative": 0.965236 }, { "name": "video_fisheye_cam1", "l2": 5.329955, "mean_abs": 0.117936, "relative": 0.88073 }, { "name": "video_fisheye_cam2", "l2": 5.618837, "mean_abs": 0.119298, "relative": 0.896964 }, { "name": "video_fisheye_cam3", "l2": 5.427153, "mean_abs": 0.114446, "relative": 0.908711 }, { "name": "video_stereo_left", "l2": 5.409598, "mean_abs": 0.123238, "relative": 0.87907 }, { "name": "video_stereo_right", "l2": 6.219789, "mean_abs": 0.1447, "relative": 0.87686 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.23649, "mean_abs": 0.095169, "relative": 0.800406 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 126, "start_frame": 630, "end_frame": 649, "center_frame": 639, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.521871, "mean_abs": 0.269012, "relative": 0.971919 }, { "name": "hand_right_joints", "l2": 11.379316, "mean_abs": 0.266578, "relative": 0.896995 }, { "name": "body_joints", "l2": 5.566233, "mean_abs": 0.086877, "relative": 0.751827 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.084544, "mean_abs": 0.011693, "relative": 0.199512 }, { "name": "camera_rotation_matrix", "l2": 3.014722, "mean_abs": 0.197074, "relative": 0.925772 }, { "name": "imu_accel_gyro", "l2": 17.060806, "mean_abs": 1.121667, "relative": 0.920489 }, { "name": "depth_confidence", "l2": 9.373582, "mean_abs": 0.181679, "relative": 0.000311 }, { "name": "video_fisheye_cam0", "l2": 5.306468, "mean_abs": 0.112647, "relative": 0.960888 }, { "name": "video_fisheye_cam1", "l2": 5.333991, "mean_abs": 0.117344, "relative": 0.881397 }, { "name": "video_fisheye_cam2", "l2": 5.630783, "mean_abs": 0.11791, "relative": 0.898871 }, { "name": "video_fisheye_cam3", "l2": 5.377669, "mean_abs": 0.112528, "relative": 0.900426 }, { "name": "video_stereo_left", "l2": 5.368959, "mean_abs": 0.121574, "relative": 0.872467 }, { "name": "video_stereo_right", "l2": 6.211336, "mean_abs": 0.144325, "relative": 0.875668 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.193919, "mean_abs": 0.088093, "relative": 0.785171 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 127, "start_frame": 635, "end_frame": 654, "center_frame": 644, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.445161, "mean_abs": 0.265253, "relative": 0.965448 }, { "name": "hand_right_joints", "l2": 11.572441, "mean_abs": 0.2747, "relative": 0.912218 }, { "name": "body_joints", "l2": 5.53189, "mean_abs": 0.085405, "relative": 0.747188 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.046449, "mean_abs": 0.006803, "relative": 0.109612 }, { "name": "camera_rotation_matrix", "l2": 3.006129, "mean_abs": 0.190136, "relative": 0.923133 }, { "name": "imu_accel_gyro", "l2": 17.136959, "mean_abs": 1.137887, "relative": 0.924598 }, { "name": "depth_confidence", "l2": 9.524881, "mean_abs": 0.183138, "relative": 0.000316 }, { "name": "video_fisheye_cam0", "l2": 5.275879, "mean_abs": 0.111076, "relative": 0.955349 }, { "name": "video_fisheye_cam1", "l2": 5.326195, "mean_abs": 0.115595, "relative": 0.880109 }, { "name": "video_fisheye_cam2", "l2": 5.648514, "mean_abs": 0.116826, "relative": 0.901701 }, { "name": "video_fisheye_cam3", "l2": 5.365237, "mean_abs": 0.110643, "relative": 0.898344 }, { "name": "video_stereo_left", "l2": 5.386719, "mean_abs": 0.121345, "relative": 0.875353 }, { "name": "video_stereo_right", "l2": 6.243686, "mean_abs": 0.144081, "relative": 0.880229 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.122416, "mean_abs": 0.084568, "relative": 0.759581 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 128, "start_frame": 640, "end_frame": 659, "center_frame": 649, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.285414, "mean_abs": 0.25782, "relative": 0.951973 }, { "name": "hand_right_joints", "l2": 11.618901, "mean_abs": 0.278873, "relative": 0.91588 }, { "name": "body_joints", "l2": 5.562661, "mean_abs": 0.086111, "relative": 0.751344 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.039975, "mean_abs": 0.005825, "relative": 0.094335 }, { "name": "camera_rotation_matrix", "l2": 3.002423, "mean_abs": 0.190411, "relative": 0.921995 }, { "name": "imu_accel_gyro", "l2": 17.406279, "mean_abs": 1.156869, "relative": 0.939129 }, { "name": "depth_confidence", "l2": 9.742139, "mean_abs": 0.188945, "relative": 0.000323 }, { "name": "video_fisheye_cam0", "l2": 5.22621, "mean_abs": 0.108844, "relative": 0.946355 }, { "name": "video_fisheye_cam1", "l2": 5.281473, "mean_abs": 0.113156, "relative": 0.872719 }, { "name": "video_fisheye_cam2", "l2": 5.661168, "mean_abs": 0.116359, "relative": 0.903721 }, { "name": "video_fisheye_cam3", "l2": 5.327837, "mean_abs": 0.109713, "relative": 0.892082 }, { "name": "video_stereo_left", "l2": 5.323127, "mean_abs": 0.119879, "relative": 0.865019 }, { "name": "video_stereo_right", "l2": 6.194865, "mean_abs": 0.143048, "relative": 0.873346 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.11912, "mean_abs": 0.081462, "relative": 0.758401 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 129, "start_frame": 645, "end_frame": 664, "center_frame": 654, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.113816, "mean_abs": 0.25532, "relative": 0.937498 }, { "name": "hand_right_joints", "l2": 11.48135, "mean_abs": 0.283709, "relative": 0.905038 }, { "name": "body_joints", "l2": 5.665426, "mean_abs": 0.085964, "relative": 0.765225 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.085629, "mean_abs": 0.011823, "relative": 0.202073 }, { "name": "camera_rotation_matrix", "l2": 3.008148, "mean_abs": 0.199016, "relative": 0.923753 }, { "name": "imu_accel_gyro", "l2": 17.433838, "mean_abs": 1.161635, "relative": 0.940616 }, { "name": "depth_confidence", "l2": 9.967113, "mean_abs": 0.195133, "relative": 0.000331 }, { "name": "video_fisheye_cam0", "l2": 5.210448, "mean_abs": 0.108871, "relative": 0.943501 }, { "name": "video_fisheye_cam1", "l2": 5.312555, "mean_abs": 0.115613, "relative": 0.877855 }, { "name": "video_fisheye_cam2", "l2": 5.693149, "mean_abs": 0.11749, "relative": 0.908827 }, { "name": "video_fisheye_cam3", "l2": 5.380932, "mean_abs": 0.111232, "relative": 0.900972 }, { "name": "video_stereo_left", "l2": 5.321911, "mean_abs": 0.119661, "relative": 0.864821 }, { "name": "video_stereo_right", "l2": 6.205337, "mean_abs": 0.143821, "relative": 0.874822 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.253993, "mean_abs": 0.084624, "relative": 0.80667 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 130, "start_frame": 650, "end_frame": 669, "center_frame": 659, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.941243, "mean_abs": 0.249121, "relative": 0.922941 }, { "name": "hand_right_joints", "l2": 11.386814, "mean_abs": 0.289379, "relative": 0.897586 }, { "name": "body_joints", "l2": 5.778973, "mean_abs": 0.089973, "relative": 0.780561 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.150503, "mean_abs": 0.019469, "relative": 0.355165 }, { "name": "camera_rotation_matrix", "l2": 3.028574, "mean_abs": 0.215237, "relative": 0.930026 }, { "name": "imu_accel_gyro", "l2": 17.270628, "mean_abs": 1.179138, "relative": 0.93181 }, { "name": "depth_confidence", "l2": 10.353147, "mean_abs": 0.203346, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.203283, "mean_abs": 0.110306, "relative": 0.942203 }, { "name": "video_fisheye_cam1", "l2": 5.384521, "mean_abs": 0.119708, "relative": 0.889747 }, { "name": "video_fisheye_cam2", "l2": 5.768978, "mean_abs": 0.121637, "relative": 0.920931 }, { "name": "video_fisheye_cam3", "l2": 5.436824, "mean_abs": 0.114038, "relative": 0.910331 }, { "name": "video_stereo_left", "l2": 5.385909, "mean_abs": 0.122946, "relative": 0.875221 }, { "name": "video_stereo_right", "l2": 6.255665, "mean_abs": 0.145934, "relative": 0.881918 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.158699, "mean_abs": 0.085011, "relative": 0.772566 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 131, "start_frame": 655, "end_frame": 674, "center_frame": 664, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.821424, "mean_abs": 0.237679, "relative": 0.912833 }, { "name": "hand_right_joints", "l2": 11.261426, "mean_abs": 0.285201, "relative": 0.887702 }, { "name": "body_joints", "l2": 5.870107, "mean_abs": 0.092446, "relative": 0.792871 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.172416, "mean_abs": 0.022698, "relative": 0.406876 }, { "name": "camera_rotation_matrix", "l2": 3.030389, "mean_abs": 0.222976, "relative": 0.930583 }, { "name": "imu_accel_gyro", "l2": 17.294167, "mean_abs": 1.170818, "relative": 0.93308 }, { "name": "depth_confidence", "l2": 10.733257, "mean_abs": 0.208286, "relative": 0.000356 }, { "name": "video_fisheye_cam0", "l2": 5.194839, "mean_abs": 0.110414, "relative": 0.940674 }, { "name": "video_fisheye_cam1", "l2": 5.421377, "mean_abs": 0.121142, "relative": 0.895837 }, { "name": "video_fisheye_cam2", "l2": 5.833063, "mean_abs": 0.123658, "relative": 0.931162 }, { "name": "video_fisheye_cam3", "l2": 5.475046, "mean_abs": 0.115112, "relative": 0.91673 }, { "name": "video_stereo_left", "l2": 5.456695, "mean_abs": 0.125911, "relative": 0.886724 }, { "name": "video_stereo_right", "l2": 6.3299, "mean_abs": 0.148567, "relative": 0.892383 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.387716, "mean_abs": 0.101305, "relative": 0.854528 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 132, "start_frame": 660, "end_frame": 679, "center_frame": 669, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.698122, "mean_abs": 0.231492, "relative": 0.902432 }, { "name": "hand_right_joints", "l2": 11.129161, "mean_abs": 0.28128, "relative": 0.877276 }, { "name": "body_joints", "l2": 5.914989, "mean_abs": 0.090822, "relative": 0.798933 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.157163, "mean_abs": 0.021833, "relative": 0.370883 }, { "name": "camera_rotation_matrix", "l2": 3.017155, "mean_abs": 0.220942, "relative": 0.926519 }, { "name": "imu_accel_gyro", "l2": 17.350483, "mean_abs": 1.147621, "relative": 0.936118 }, { "name": "depth_confidence", "l2": 10.900615, "mean_abs": 0.209204, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.182642, "mean_abs": 0.109509, "relative": 0.938466 }, { "name": "video_fisheye_cam1", "l2": 5.458359, "mean_abs": 0.12115, "relative": 0.901948 }, { "name": "video_fisheye_cam2", "l2": 5.856218, "mean_abs": 0.123419, "relative": 0.934858 }, { "name": "video_fisheye_cam3", "l2": 5.524886, "mean_abs": 0.115101, "relative": 0.925076 }, { "name": "video_stereo_left", "l2": 5.428124, "mean_abs": 0.124958, "relative": 0.882081 }, { "name": "video_stereo_right", "l2": 6.385106, "mean_abs": 0.148901, "relative": 0.900166 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.34256, "mean_abs": 0.097678, "relative": 0.838367 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 133, "start_frame": 665, "end_frame": 684, "center_frame": 674, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.629717, "mean_abs": 0.228551, "relative": 0.896662 }, { "name": "hand_right_joints", "l2": 10.880366, "mean_abs": 0.271173, "relative": 0.857664 }, { "name": "body_joints", "l2": 5.959214, "mean_abs": 0.087273, "relative": 0.804906 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.114384, "mean_abs": 0.017885, "relative": 0.26993 }, { "name": "camera_rotation_matrix", "l2": 3.006478, "mean_abs": 0.215377, "relative": 0.923241 }, { "name": "imu_accel_gyro", "l2": 17.168016, "mean_abs": 1.114318, "relative": 0.926274 }, { "name": "depth_confidence", "l2": 11.118579, "mean_abs": 0.209078, "relative": 0.000369 }, { "name": "video_fisheye_cam0", "l2": 5.156868, "mean_abs": 0.108066, "relative": 0.933798 }, { "name": "video_fisheye_cam1", "l2": 5.493331, "mean_abs": 0.119272, "relative": 0.907727 }, { "name": "video_fisheye_cam2", "l2": 5.862163, "mean_abs": 0.121721, "relative": 0.935807 }, { "name": "video_fisheye_cam3", "l2": 5.560525, "mean_abs": 0.114572, "relative": 0.931043 }, { "name": "video_stereo_left", "l2": 5.408317, "mean_abs": 0.12286, "relative": 0.878862 }, { "name": "video_stereo_right", "l2": 6.432402, "mean_abs": 0.148607, "relative": 0.906834 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.242327, "mean_abs": 0.09346, "relative": 0.802495 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 134, "start_frame": 670, "end_frame": 689, "center_frame": 679, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.649339, "mean_abs": 0.22568, "relative": 0.898317 }, { "name": "hand_right_joints", "l2": 10.902403, "mean_abs": 0.275759, "relative": 0.859401 }, { "name": "body_joints", "l2": 5.968454, "mean_abs": 0.084265, "relative": 0.806154 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.1022, "mean_abs": 0.013103, "relative": 0.241178 }, { "name": "camera_rotation_matrix", "l2": 2.999467, "mean_abs": 0.209779, "relative": 0.921088 }, { "name": "imu_accel_gyro", "l2": 17.264219, "mean_abs": 1.076008, "relative": 0.931464 }, { "name": "depth_confidence", "l2": 11.296099, "mean_abs": 0.207124, "relative": 0.000375 }, { "name": "video_fisheye_cam0", "l2": 5.137882, "mean_abs": 0.106995, "relative": 0.93036 }, { "name": "video_fisheye_cam1", "l2": 5.53289, "mean_abs": 0.117705, "relative": 0.914263 }, { "name": "video_fisheye_cam2", "l2": 5.858183, "mean_abs": 0.119919, "relative": 0.935172 }, { "name": "video_fisheye_cam3", "l2": 5.608824, "mean_abs": 0.114802, "relative": 0.93913 }, { "name": "video_stereo_left", "l2": 5.399127, "mean_abs": 0.120331, "relative": 0.877369 }, { "name": "video_stereo_right", "l2": 6.448909, "mean_abs": 0.146161, "relative": 0.909161 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.258737, "mean_abs": 0.090273, "relative": 0.808368 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 135, "start_frame": 675, "end_frame": 694, "center_frame": 684, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.644957, "mean_abs": 0.224193, "relative": 0.897947 }, { "name": "hand_right_joints", "l2": 10.966084, "mean_abs": 0.277839, "relative": 0.864421 }, { "name": "body_joints", "l2": 5.999756, "mean_abs": 0.082923, "relative": 0.810382 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.097072, "mean_abs": 0.01025, "relative": 0.229075 }, { "name": "camera_rotation_matrix", "l2": 2.999487, "mean_abs": 0.209781, "relative": 0.921094 }, { "name": "imu_accel_gyro", "l2": 17.36244, "mean_abs": 1.04958, "relative": 0.936763 }, { "name": "depth_confidence", "l2": 11.322454, "mean_abs": 0.205982, "relative": 0.000376 }, { "name": "video_fisheye_cam0", "l2": 5.143591, "mean_abs": 0.106862, "relative": 0.931394 }, { "name": "video_fisheye_cam1", "l2": 5.54697, "mean_abs": 0.117351, "relative": 0.91659 }, { "name": "video_fisheye_cam2", "l2": 5.840637, "mean_abs": 0.119502, "relative": 0.932371 }, { "name": "video_fisheye_cam3", "l2": 5.650821, "mean_abs": 0.115128, "relative": 0.946162 }, { "name": "video_stereo_left", "l2": 5.372892, "mean_abs": 0.117698, "relative": 0.873106 }, { "name": "video_stereo_right", "l2": 6.482775, "mean_abs": 0.144688, "relative": 0.913935 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.383027, "mean_abs": 0.096934, "relative": 0.852849 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 136, "start_frame": 680, "end_frame": 699, "center_frame": 689, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.635866, "mean_abs": 0.21986, "relative": 0.897181 }, { "name": "hand_right_joints", "l2": 11.010211, "mean_abs": 0.276097, "relative": 0.867899 }, { "name": "body_joints", "l2": 6.02081, "mean_abs": 0.081028, "relative": 0.813226 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.067357, "mean_abs": 0.008316, "relative": 0.158953 }, { "name": "camera_rotation_matrix", "l2": 3.001094, "mean_abs": 0.210606, "relative": 0.921587 }, { "name": "imu_accel_gyro", "l2": 17.078451, "mean_abs": 0.9874, "relative": 0.921441 }, { "name": "depth_confidence", "l2": 11.3298, "mean_abs": 0.206052, "relative": 0.000376 }, { "name": "video_fisheye_cam0", "l2": 5.146548, "mean_abs": 0.10691, "relative": 0.93193 }, { "name": "video_fisheye_cam1", "l2": 5.549731, "mean_abs": 0.117838, "relative": 0.917046 }, { "name": "video_fisheye_cam2", "l2": 5.802223, "mean_abs": 0.118633, "relative": 0.926239 }, { "name": "video_fisheye_cam3", "l2": 5.666252, "mean_abs": 0.11531, "relative": 0.948746 }, { "name": "video_stereo_left", "l2": 5.352726, "mean_abs": 0.117527, "relative": 0.869829 }, { "name": "video_stereo_right", "l2": 6.502821, "mean_abs": 0.145661, "relative": 0.916761 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.262208, "mean_abs": 0.086544, "relative": 0.80961 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 137, "start_frame": 685, "end_frame": 704, "center_frame": 694, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.643456, "mean_abs": 0.218814, "relative": 0.897821 }, { "name": "hand_right_joints", "l2": 11.013905, "mean_abs": 0.274147, "relative": 0.86819 }, { "name": "body_joints", "l2": 6.00798, "mean_abs": 0.080011, "relative": 0.811493 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.048346, "mean_abs": 0.006608, "relative": 0.114089 }, { "name": "camera_rotation_matrix", "l2": 3.000923, "mean_abs": 0.210121, "relative": 0.921535 }, { "name": "imu_accel_gyro", "l2": 17.131306, "mean_abs": 0.981953, "relative": 0.924293 }, { "name": "depth_confidence", "l2": 11.306185, "mean_abs": 0.20575, "relative": 0.000375 }, { "name": "video_fisheye_cam0", "l2": 5.149095, "mean_abs": 0.106992, "relative": 0.932391 }, { "name": "video_fisheye_cam1", "l2": 5.514032, "mean_abs": 0.117256, "relative": 0.911147 }, { "name": "video_fisheye_cam2", "l2": 5.782572, "mean_abs": 0.11816, "relative": 0.923102 }, { "name": "video_fisheye_cam3", "l2": 5.671873, "mean_abs": 0.115339, "relative": 0.949687 }, { "name": "video_stereo_left", "l2": 5.352668, "mean_abs": 0.117254, "relative": 0.869819 }, { "name": "video_stereo_right", "l2": 6.489679, "mean_abs": 0.144892, "relative": 0.914909 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.053999, "mean_abs": 0.077658, "relative": 0.735095 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 138, "start_frame": 690, "end_frame": 709, "center_frame": 699, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.643641, "mean_abs": 0.218507, "relative": 0.897837 }, { "name": "hand_right_joints", "l2": 10.993439, "mean_abs": 0.272863, "relative": 0.866577 }, { "name": "body_joints", "l2": 5.973223, "mean_abs": 0.079349, "relative": 0.806798 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.025812, "mean_abs": 0.003848, "relative": 0.060913 }, { "name": "camera_rotation_matrix", "l2": 2.999848, "mean_abs": 0.207606, "relative": 0.921205 }, { "name": "imu_accel_gyro", "l2": 17.133327, "mean_abs": 0.979162, "relative": 0.924402 }, { "name": "depth_confidence", "l2": 11.195663, "mean_abs": 0.203456, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.149997, "mean_abs": 0.106718, "relative": 0.932554 }, { "name": "video_fisheye_cam1", "l2": 5.498039, "mean_abs": 0.116233, "relative": 0.908505 }, { "name": "video_fisheye_cam2", "l2": 5.767402, "mean_abs": 0.116903, "relative": 0.92068 }, { "name": "video_fisheye_cam3", "l2": 5.645997, "mean_abs": 0.115427, "relative": 0.945354 }, { "name": "video_stereo_left", "l2": 5.345018, "mean_abs": 0.115696, "relative": 0.868576 }, { "name": "video_stereo_right", "l2": 6.483066, "mean_abs": 0.142809, "relative": 0.913976 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.06434, "mean_abs": 0.080023, "relative": 0.738796 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 139, "start_frame": 695, "end_frame": 714, "center_frame": 704, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.66881, "mean_abs": 0.217606, "relative": 0.89996 }, { "name": "hand_right_joints", "l2": 10.965776, "mean_abs": 0.270603, "relative": 0.864397 }, { "name": "body_joints", "l2": 5.947197, "mean_abs": 0.080086, "relative": 0.803283 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016834, "mean_abs": 0.002579, "relative": 0.039726 }, { "name": "camera_rotation_matrix", "l2": 3.001025, "mean_abs": 0.207793, "relative": 0.921566 }, { "name": "imu_accel_gyro", "l2": 17.074226, "mean_abs": 0.980162, "relative": 0.921213 }, { "name": "depth_confidence", "l2": 11.174499, "mean_abs": 0.203965, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.150463, "mean_abs": 0.10666, "relative": 0.932639 }, { "name": "video_fisheye_cam1", "l2": 5.484077, "mean_abs": 0.116257, "relative": 0.906197 }, { "name": "video_fisheye_cam2", "l2": 5.757642, "mean_abs": 0.116993, "relative": 0.919122 }, { "name": "video_fisheye_cam3", "l2": 5.613557, "mean_abs": 0.114484, "relative": 0.939922 }, { "name": "video_stereo_left", "l2": 5.341221, "mean_abs": 0.115974, "relative": 0.867959 }, { "name": "video_stereo_right", "l2": 6.477895, "mean_abs": 0.143585, "relative": 0.913247 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.065759, "mean_abs": 0.07929, "relative": 0.739304 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 140, "start_frame": 700, "end_frame": 719, "center_frame": 709, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.667978, "mean_abs": 0.216726, "relative": 0.899889 }, { "name": "hand_right_joints", "l2": 10.946211, "mean_abs": 0.26826, "relative": 0.862854 }, { "name": "body_joints", "l2": 5.932661, "mean_abs": 0.080536, "relative": 0.80132 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.02082, "mean_abs": 0.002714, "relative": 0.049133 }, { "name": "camera_rotation_matrix", "l2": 3.000932, "mean_abs": 0.208189, "relative": 0.921538 }, { "name": "imu_accel_gyro", "l2": 17.093426, "mean_abs": 0.971047, "relative": 0.922249 }, { "name": "depth_confidence", "l2": 11.168551, "mean_abs": 0.204078, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.154596, "mean_abs": 0.106993, "relative": 0.933387 }, { "name": "video_fisheye_cam1", "l2": 5.477311, "mean_abs": 0.116547, "relative": 0.905079 }, { "name": "video_fisheye_cam2", "l2": 5.745785, "mean_abs": 0.117237, "relative": 0.917229 }, { "name": "video_fisheye_cam3", "l2": 5.594107, "mean_abs": 0.114012, "relative": 0.936666 }, { "name": "video_stereo_left", "l2": 5.328177, "mean_abs": 0.116305, "relative": 0.865839 }, { "name": "video_stereo_right", "l2": 6.458767, "mean_abs": 0.144234, "relative": 0.910551 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.042688, "mean_abs": 0.076619, "relative": 0.731047 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 141, "start_frame": 705, "end_frame": 724, "center_frame": 714, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.646409, "mean_abs": 0.214178, "relative": 0.89807 }, { "name": "hand_right_joints", "l2": 10.91879, "mean_abs": 0.268037, "relative": 0.860693 }, { "name": "body_joints", "l2": 5.918882, "mean_abs": 0.080092, "relative": 0.799459 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030478, "mean_abs": 0.003944, "relative": 0.071925 }, { "name": "camera_rotation_matrix", "l2": 3.00115, "mean_abs": 0.207914, "relative": 0.921604 }, { "name": "imu_accel_gyro", "l2": 17.055431, "mean_abs": 0.966556, "relative": 0.920199 }, { "name": "depth_confidence", "l2": 11.151011, "mean_abs": 0.203993, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.156998, "mean_abs": 0.10721, "relative": 0.933822 }, { "name": "video_fisheye_cam1", "l2": 5.46554, "mean_abs": 0.116342, "relative": 0.903134 }, { "name": "video_fisheye_cam2", "l2": 5.73705, "mean_abs": 0.117248, "relative": 0.915835 }, { "name": "video_fisheye_cam3", "l2": 5.575774, "mean_abs": 0.11391, "relative": 0.933596 }, { "name": "video_stereo_left", "l2": 5.312679, "mean_abs": 0.116477, "relative": 0.863321 }, { "name": "video_stereo_right", "l2": 6.421659, "mean_abs": 0.143671, "relative": 0.905319 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.060582, "mean_abs": 0.079395, "relative": 0.737451 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 142, "start_frame": 710, "end_frame": 729, "center_frame": 719, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.639905, "mean_abs": 0.214296, "relative": 0.897521 }, { "name": "hand_right_joints", "l2": 10.90271, "mean_abs": 0.267899, "relative": 0.859425 }, { "name": "body_joints", "l2": 5.905941, "mean_abs": 0.079113, "relative": 0.797711 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028122, "mean_abs": 0.003889, "relative": 0.066363 }, { "name": "camera_rotation_matrix", "l2": 3.000186, "mean_abs": 0.206049, "relative": 0.921309 }, { "name": "imu_accel_gyro", "l2": 17.058331, "mean_abs": 0.971297, "relative": 0.920356 }, { "name": "depth_confidence", "l2": 11.127781, "mean_abs": 0.202822, "relative": 0.000369 }, { "name": "video_fisheye_cam0", "l2": 5.160119, "mean_abs": 0.106997, "relative": 0.934387 }, { "name": "video_fisheye_cam1", "l2": 5.466374, "mean_abs": 0.115521, "relative": 0.903272 }, { "name": "video_fisheye_cam2", "l2": 5.727163, "mean_abs": 0.116506, "relative": 0.914256 }, { "name": "video_fisheye_cam3", "l2": 5.572985, "mean_abs": 0.113355, "relative": 0.933129 }, { "name": "video_stereo_left", "l2": 5.289546, "mean_abs": 0.1152, "relative": 0.859562 }, { "name": "video_stereo_right", "l2": 6.40227, "mean_abs": 0.141827, "relative": 0.902586 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.047208, "mean_abs": 0.078982, "relative": 0.732665 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 143, "start_frame": 715, "end_frame": 734, "center_frame": 724, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.629395, "mean_abs": 0.215995, "relative": 0.896635 }, { "name": "hand_right_joints", "l2": 10.895866, "mean_abs": 0.268499, "relative": 0.858886 }, { "name": "body_joints", "l2": 5.908081, "mean_abs": 0.078692, "relative": 0.798 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.033172, "mean_abs": 0.004228, "relative": 0.078282 }, { "name": "camera_rotation_matrix", "l2": 2.998885, "mean_abs": 0.204134, "relative": 0.920909 }, { "name": "imu_accel_gyro", "l2": 17.087334, "mean_abs": 0.982504, "relative": 0.92192 }, { "name": "depth_confidence", "l2": 11.077485, "mean_abs": 0.202355, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.166555, "mean_abs": 0.106761, "relative": 0.935553 }, { "name": "video_fisheye_cam1", "l2": 5.468493, "mean_abs": 0.114833, "relative": 0.903622 }, { "name": "video_fisheye_cam2", "l2": 5.722139, "mean_abs": 0.115936, "relative": 0.913454 }, { "name": "video_fisheye_cam3", "l2": 5.585266, "mean_abs": 0.113745, "relative": 0.935185 }, { "name": "video_stereo_left", "l2": 5.27489, "mean_abs": 0.113898, "relative": 0.85718 }, { "name": "video_stereo_right", "l2": 6.39581, "mean_abs": 0.140006, "relative": 0.901675 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.08473, "mean_abs": 0.079081, "relative": 0.746093 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 144, "start_frame": 720, "end_frame": 739, "center_frame": 729, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.604985, "mean_abs": 0.215491, "relative": 0.894576 }, { "name": "hand_right_joints", "l2": 10.8746, "mean_abs": 0.266881, "relative": 0.857209 }, { "name": "body_joints", "l2": 5.913402, "mean_abs": 0.078958, "relative": 0.798718 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.033692, "mean_abs": 0.004371, "relative": 0.079507 }, { "name": "camera_rotation_matrix", "l2": 2.999697, "mean_abs": 0.204355, "relative": 0.921158 }, { "name": "imu_accel_gyro", "l2": 17.102016, "mean_abs": 0.990391, "relative": 0.922713 }, { "name": "depth_confidence", "l2": 11.047233, "mean_abs": 0.203041, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.174921, "mean_abs": 0.107025, "relative": 0.937067 }, { "name": "video_fisheye_cam1", "l2": 5.46522, "mean_abs": 0.114873, "relative": 0.903082 }, { "name": "video_fisheye_cam2", "l2": 5.724977, "mean_abs": 0.115964, "relative": 0.913907 }, { "name": "video_fisheye_cam3", "l2": 5.597007, "mean_abs": 0.114381, "relative": 0.937151 }, { "name": "video_stereo_left", "l2": 5.284263, "mean_abs": 0.114687, "relative": 0.858703 }, { "name": "video_stereo_right", "l2": 6.405516, "mean_abs": 0.140949, "relative": 0.903043 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.952567, "mean_abs": 0.075694, "relative": 0.698794 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 145, "start_frame": 725, "end_frame": 744, "center_frame": 734, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.632874, "mean_abs": 0.218652, "relative": 0.896928 }, { "name": "hand_right_joints", "l2": 10.901416, "mean_abs": 0.268176, "relative": 0.859323 }, { "name": "body_joints", "l2": 5.901145, "mean_abs": 0.078907, "relative": 0.797063 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.033229, "mean_abs": 0.00491, "relative": 0.078416 }, { "name": "camera_rotation_matrix", "l2": 3.000727, "mean_abs": 0.205013, "relative": 0.921475 }, { "name": "imu_accel_gyro", "l2": 17.128353, "mean_abs": 0.999315, "relative": 0.924134 }, { "name": "depth_confidence", "l2": 11.013315, "mean_abs": 0.20345, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.182118, "mean_abs": 0.107172, "relative": 0.938371 }, { "name": "video_fisheye_cam1", "l2": 5.463861, "mean_abs": 0.115113, "relative": 0.902857 }, { "name": "video_fisheye_cam2", "l2": 5.734687, "mean_abs": 0.116257, "relative": 0.915458 }, { "name": "video_fisheye_cam3", "l2": 5.609979, "mean_abs": 0.114773, "relative": 0.939323 }, { "name": "video_stereo_left", "l2": 5.302671, "mean_abs": 0.11553, "relative": 0.861695 }, { "name": "video_stereo_right", "l2": 6.429978, "mean_abs": 0.141952, "relative": 0.906492 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.000384, "mean_abs": 0.075312, "relative": 0.715908 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 146, "start_frame": 730, "end_frame": 749, "center_frame": 739, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.663634, "mean_abs": 0.221323, "relative": 0.899523 }, { "name": "hand_right_joints", "l2": 10.885627, "mean_abs": 0.269066, "relative": 0.858079 }, { "name": "body_joints", "l2": 5.854235, "mean_abs": 0.07846, "relative": 0.790727 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.026902, "mean_abs": 0.003994, "relative": 0.063485 }, { "name": "camera_rotation_matrix", "l2": 3.000602, "mean_abs": 0.203959, "relative": 0.921436 }, { "name": "imu_accel_gyro", "l2": 17.062071, "mean_abs": 0.998058, "relative": 0.920557 }, { "name": "depth_confidence", "l2": 10.961183, "mean_abs": 0.202475, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.191179, "mean_abs": 0.107355, "relative": 0.940012 }, { "name": "video_fisheye_cam1", "l2": 5.458287, "mean_abs": 0.114749, "relative": 0.901936 }, { "name": "video_fisheye_cam2", "l2": 5.740708, "mean_abs": 0.116031, "relative": 0.916419 }, { "name": "video_fisheye_cam3", "l2": 5.573545, "mean_abs": 0.113944, "relative": 0.933223 }, { "name": "video_stereo_left", "l2": 5.312648, "mean_abs": 0.11506, "relative": 0.863316 }, { "name": "video_stereo_right", "l2": 6.442541, "mean_abs": 0.141431, "relative": 0.908263 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.95585, "mean_abs": 0.073686, "relative": 0.699969 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 147, "start_frame": 735, "end_frame": 754, "center_frame": 744, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.694229, "mean_abs": 0.223692, "relative": 0.902104 }, { "name": "hand_right_joints", "l2": 10.878381, "mean_abs": 0.270218, "relative": 0.857507 }, { "name": "body_joints", "l2": 5.78188, "mean_abs": 0.078013, "relative": 0.780954 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014519, "mean_abs": 0.002241, "relative": 0.034264 }, { "name": "camera_rotation_matrix", "l2": 3.000427, "mean_abs": 0.202546, "relative": 0.921383 }, { "name": "imu_accel_gyro", "l2": 17.08353, "mean_abs": 0.988127, "relative": 0.921715 }, { "name": "depth_confidence", "l2": 10.928511, "mean_abs": 0.201108, "relative": 0.000363 }, { "name": "video_fisheye_cam0", "l2": 5.201015, "mean_abs": 0.107242, "relative": 0.941793 }, { "name": "video_fisheye_cam1", "l2": 5.454541, "mean_abs": 0.114429, "relative": 0.901317 }, { "name": "video_fisheye_cam2", "l2": 5.74504, "mean_abs": 0.115884, "relative": 0.91711 }, { "name": "video_fisheye_cam3", "l2": 5.57224, "mean_abs": 0.114208, "relative": 0.933004 }, { "name": "video_stereo_left", "l2": 5.324869, "mean_abs": 0.114611, "relative": 0.865302 }, { "name": "video_stereo_right", "l2": 6.436261, "mean_abs": 0.141022, "relative": 0.907378 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.979793, "mean_abs": 0.074993, "relative": 0.708538 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 148, "start_frame": 740, "end_frame": 759, "center_frame": 749, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.738692, "mean_abs": 0.224092, "relative": 0.905855 }, { "name": "hand_right_joints", "l2": 10.923048, "mean_abs": 0.270938, "relative": 0.861028 }, { "name": "body_joints", "l2": 5.710869, "mean_abs": 0.077181, "relative": 0.771362 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010038, "mean_abs": 0.001408, "relative": 0.023689 }, { "name": "camera_rotation_matrix", "l2": 3.000234, "mean_abs": 0.203334, "relative": 0.921323 }, { "name": "imu_accel_gyro", "l2": 17.100718, "mean_abs": 0.995716, "relative": 0.922643 }, { "name": "depth_confidence", "l2": 10.879979, "mean_abs": 0.201131, "relative": 0.000361 }, { "name": "video_fisheye_cam0", "l2": 5.208713, "mean_abs": 0.107845, "relative": 0.943187 }, { "name": "video_fisheye_cam1", "l2": 5.461009, "mean_abs": 0.115382, "relative": 0.902386 }, { "name": "video_fisheye_cam2", "l2": 5.743189, "mean_abs": 0.116549, "relative": 0.916815 }, { "name": "video_fisheye_cam3", "l2": 5.538317, "mean_abs": 0.113025, "relative": 0.927324 }, { "name": "video_stereo_left", "l2": 5.325035, "mean_abs": 0.116444, "relative": 0.865329 }, { "name": "video_stereo_right", "l2": 6.433625, "mean_abs": 0.14289, "relative": 0.907006 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.005191, "mean_abs": 0.076059, "relative": 0.717628 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011074, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 149, "start_frame": 745, "end_frame": 764, "center_frame": 754, "action": "Position kettle to pour", "subtask": "Pick up and position kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.774803, "mean_abs": 0.223608, "relative": 0.908901 }, { "name": "hand_right_joints", "l2": 10.905432, "mean_abs": 0.267525, "relative": 0.85964 }, { "name": "body_joints", "l2": 5.622139, "mean_abs": 0.07594, "relative": 0.759378 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028829, "mean_abs": 0.003198, "relative": 0.068032 }, { "name": "camera_rotation_matrix", "l2": 3.00114, "mean_abs": 0.20491, "relative": 0.921601 }, { "name": "imu_accel_gyro", "l2": 17.077868, "mean_abs": 0.990691, "relative": 0.92141 }, { "name": "depth_confidence", "l2": 10.874494, "mean_abs": 0.201958, "relative": 0.000361 }, { "name": "video_fisheye_cam0", "l2": 5.215157, "mean_abs": 0.108391, "relative": 0.944353 }, { "name": "video_fisheye_cam1", "l2": 5.465979, "mean_abs": 0.116317, "relative": 0.903207 }, { "name": "video_fisheye_cam2", "l2": 5.741132, "mean_abs": 0.117444, "relative": 0.916486 }, { "name": "video_fisheye_cam3", "l2": 5.522438, "mean_abs": 0.113626, "relative": 0.924666 }, { "name": "video_stereo_left", "l2": 5.327112, "mean_abs": 0.118354, "relative": 0.865666 }, { "name": "video_stereo_right", "l2": 6.42568, "mean_abs": 0.144706, "relative": 0.905886 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.959426, "mean_abs": 0.074651, "relative": 0.701249 }, { "name": "caption_objects_interaction_text", "l2": 1.928611, "mean_abs": 0.018873, "relative": 0.885571 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 150, "start_frame": 750, "end_frame": 769, "center_frame": 759, "action": "Position kettle to pour", "subtask": "", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.779783, "mean_abs": 0.222771, "relative": 0.909321 }, { "name": "hand_right_joints", "l2": 10.870311, "mean_abs": 0.261657, "relative": 0.856871 }, { "name": "body_joints", "l2": 5.511926, "mean_abs": 0.074414, "relative": 0.744491 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.049359, "mean_abs": 0.005494, "relative": 0.11648 }, { "name": "camera_rotation_matrix", "l2": 3.002598, "mean_abs": 0.205705, "relative": 0.922049 }, { "name": "imu_accel_gyro", "l2": 17.019613, "mean_abs": 0.998944, "relative": 0.918267 }, { "name": "depth_confidence", "l2": 10.851224, "mean_abs": 0.202021, "relative": 0.00036 }, { "name": "video_fisheye_cam0", "l2": 5.222565, "mean_abs": 0.108531, "relative": 0.945695 }, { "name": "video_fisheye_cam1", "l2": 5.473572, "mean_abs": 0.116695, "relative": 0.904462 }, { "name": "video_fisheye_cam2", "l2": 5.73955, "mean_abs": 0.117895, "relative": 0.916234 }, { "name": "video_fisheye_cam3", "l2": 5.517438, "mean_abs": 0.113375, "relative": 0.923829 }, { "name": "video_stereo_left", "l2": 5.326301, "mean_abs": 0.11878, "relative": 0.865534 }, { "name": "video_stereo_right", "l2": 6.41115, "mean_abs": 0.144647, "relative": 0.903838 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.983081, "mean_abs": 0.077563, "relative": 0.709715 }, { "name": "caption_objects_interaction_text", "l2": 1.928611, "mean_abs": 0.019285, "relative": 0.885571 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 151, "start_frame": 755, "end_frame": 774, "center_frame": 764, "action": "Position kettle to pour", "subtask": "Position kettle to pour", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.787375, "mean_abs": 0.226783, "relative": 0.909961 }, { "name": "hand_right_joints", "l2": 10.904818, "mean_abs": 0.257779, "relative": 0.859591 }, { "name": "body_joints", "l2": 5.420144, "mean_abs": 0.0752, "relative": 0.732094 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.062301, "mean_abs": 0.007273, "relative": 0.147021 }, { "name": "camera_rotation_matrix", "l2": 3.002494, "mean_abs": 0.206377, "relative": 0.922017 }, { "name": "imu_accel_gyro", "l2": 17.024473, "mean_abs": 1.011942, "relative": 0.918529 }, { "name": "depth_confidence", "l2": 10.811746, "mean_abs": 0.201331, "relative": 0.000359 }, { "name": "video_fisheye_cam0", "l2": 5.227494, "mean_abs": 0.108516, "relative": 0.946587 }, { "name": "video_fisheye_cam1", "l2": 5.477247, "mean_abs": 0.116342, "relative": 0.905069 }, { "name": "video_fisheye_cam2", "l2": 5.739886, "mean_abs": 0.11772, "relative": 0.916287 }, { "name": "video_fisheye_cam3", "l2": 5.503224, "mean_abs": 0.112728, "relative": 0.921449 }, { "name": "video_stereo_left", "l2": 5.317713, "mean_abs": 0.117825, "relative": 0.864139 }, { "name": "video_stereo_right", "l2": 6.37637, "mean_abs": 0.142824, "relative": 0.898934 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.981237, "mean_abs": 0.077711, "relative": 0.709055 }, { "name": "caption_objects_interaction_text", "l2": 1.928611, "mean_abs": 0.019138, "relative": 0.885571 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 152, "start_frame": 760, "end_frame": 779, "center_frame": 769, "action": "Position kettle to pour", "subtask": "Position kettle to pour", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.791936, "mean_abs": 0.228878, "relative": 0.910346 }, { "name": "hand_right_joints", "l2": 11.028616, "mean_abs": 0.260323, "relative": 0.86935 }, { "name": "body_joints", "l2": 5.355643, "mean_abs": 0.076928, "relative": 0.723382 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.065892, "mean_abs": 0.007888, "relative": 0.155496 }, { "name": "camera_rotation_matrix", "l2": 3.00277, "mean_abs": 0.206073, "relative": 0.922102 }, { "name": "imu_accel_gyro", "l2": 17.046354, "mean_abs": 1.033099, "relative": 0.91971 }, { "name": "depth_confidence", "l2": 10.76878, "mean_abs": 0.20147, "relative": 0.000357 }, { "name": "video_fisheye_cam0", "l2": 5.225702, "mean_abs": 0.108336, "relative": 0.946263 }, { "name": "video_fisheye_cam1", "l2": 5.470255, "mean_abs": 0.115588, "relative": 0.903913 }, { "name": "video_fisheye_cam2", "l2": 5.744608, "mean_abs": 0.117337, "relative": 0.917041 }, { "name": "video_fisheye_cam3", "l2": 5.471932, "mean_abs": 0.111539, "relative": 0.916209 }, { "name": "video_stereo_left", "l2": 5.316916, "mean_abs": 0.117558, "relative": 0.864009 }, { "name": "video_stereo_right", "l2": 6.352244, "mean_abs": 0.141404, "relative": 0.895533 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.123666, "mean_abs": 0.08471, "relative": 0.760028 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01266, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 153, "start_frame": 765, "end_frame": 784, "center_frame": 774, "action": "Position kettle to pour", "subtask": "Position kettle to pour", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.734303, "mean_abs": 0.227646, "relative": 0.905484 }, { "name": "hand_right_joints", "l2": 11.139984, "mean_abs": 0.264338, "relative": 0.878129 }, { "name": "body_joints", "l2": 5.317402, "mean_abs": 0.077724, "relative": 0.718217 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.056514, "mean_abs": 0.006689, "relative": 0.133365 }, { "name": "camera_rotation_matrix", "l2": 3.002227, "mean_abs": 0.205359, "relative": 0.921935 }, { "name": "imu_accel_gyro", "l2": 17.054737, "mean_abs": 1.046503, "relative": 0.920162 }, { "name": "depth_confidence", "l2": 10.729259, "mean_abs": 0.201347, "relative": 0.000356 }, { "name": "video_fisheye_cam0", "l2": 5.22488, "mean_abs": 0.108547, "relative": 0.946114 }, { "name": "video_fisheye_cam1", "l2": 5.465061, "mean_abs": 0.115363, "relative": 0.903055 }, { "name": "video_fisheye_cam2", "l2": 5.759026, "mean_abs": 0.117369, "relative": 0.919343 }, { "name": "video_fisheye_cam3", "l2": 5.426907, "mean_abs": 0.110513, "relative": 0.90867 }, { "name": "video_stereo_left", "l2": 5.329321, "mean_abs": 0.118025, "relative": 0.866025 }, { "name": "video_stereo_right", "l2": 6.36169, "mean_abs": 0.142302, "relative": 0.896865 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.072459, "mean_abs": 0.081889, "relative": 0.741702 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01266, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 154, "start_frame": 770, "end_frame": 789, "center_frame": 779, "action": "Position kettle to pour", "subtask": "Position kettle to pour", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.735101, "mean_abs": 0.226593, "relative": 0.905552 }, { "name": "hand_right_joints", "l2": 11.291751, "mean_abs": 0.26735, "relative": 0.890092 }, { "name": "body_joints", "l2": 5.270425, "mean_abs": 0.076658, "relative": 0.711872 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.039292, "mean_abs": 0.004575, "relative": 0.092724 }, { "name": "camera_rotation_matrix", "l2": 3.001932, "mean_abs": 0.205742, "relative": 0.921845 }, { "name": "imu_accel_gyro", "l2": 17.077482, "mean_abs": 1.046513, "relative": 0.921389 }, { "name": "depth_confidence", "l2": 10.71183, "mean_abs": 0.200076, "relative": 0.000356 }, { "name": "video_fisheye_cam0", "l2": 5.227537, "mean_abs": 0.108757, "relative": 0.946595 }, { "name": "video_fisheye_cam1", "l2": 5.454994, "mean_abs": 0.115357, "relative": 0.901392 }, { "name": "video_fisheye_cam2", "l2": 5.773888, "mean_abs": 0.117782, "relative": 0.921715 }, { "name": "video_fisheye_cam3", "l2": 5.418494, "mean_abs": 0.110028, "relative": 0.907261 }, { "name": "video_stereo_left", "l2": 5.336142, "mean_abs": 0.118206, "relative": 0.867134 }, { "name": "video_stereo_right", "l2": 6.366103, "mean_abs": 0.142915, "relative": 0.897487 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.087585, "mean_abs": 0.080376, "relative": 0.747115 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01266, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 155, "start_frame": 775, "end_frame": 794, "center_frame": 784, "action": "Position kettle to pour", "subtask": "Position kettle to pour", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.746322, "mean_abs": 0.22685, "relative": 0.906498 }, { "name": "hand_right_joints", "l2": 11.467384, "mean_abs": 0.267924, "relative": 0.903937 }, { "name": "body_joints", "l2": 5.242383, "mean_abs": 0.075482, "relative": 0.708084 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.026364, "mean_abs": 0.003815, "relative": 0.062215 }, { "name": "camera_rotation_matrix", "l2": 3.001608, "mean_abs": 0.206716, "relative": 0.921745 }, { "name": "imu_accel_gyro", "l2": 17.097498, "mean_abs": 1.055848, "relative": 0.922469 }, { "name": "depth_confidence", "l2": 10.702054, "mean_abs": 0.199141, "relative": 0.000355 }, { "name": "video_fisheye_cam0", "l2": 5.230599, "mean_abs": 0.108802, "relative": 0.94715 }, { "name": "video_fisheye_cam1", "l2": 5.443462, "mean_abs": 0.11515, "relative": 0.899486 }, { "name": "video_fisheye_cam2", "l2": 5.791946, "mean_abs": 0.117828, "relative": 0.924598 }, { "name": "video_fisheye_cam3", "l2": 5.433457, "mean_abs": 0.110603, "relative": 0.909767 }, { "name": "video_stereo_left", "l2": 5.346098, "mean_abs": 0.117922, "relative": 0.868752 }, { "name": "video_stereo_right", "l2": 6.371626, "mean_abs": 0.142743, "relative": 0.898266 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.102199, "mean_abs": 0.082432, "relative": 0.752345 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01266, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 156, "start_frame": 780, "end_frame": 799, "center_frame": 789, "action": "Position kettle to pour", "subtask": "Position kettle to pour", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.755882, "mean_abs": 0.230078, "relative": 0.907305 }, { "name": "hand_right_joints", "l2": 11.645325, "mean_abs": 0.268498, "relative": 0.917963 }, { "name": "body_joints", "l2": 5.260634, "mean_abs": 0.075726, "relative": 0.71055 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.033824, "mean_abs": 0.004286, "relative": 0.079819 }, { "name": "camera_rotation_matrix", "l2": 3.000947, "mean_abs": 0.207287, "relative": 0.921542 }, { "name": "imu_accel_gyro", "l2": 17.105303, "mean_abs": 1.07392, "relative": 0.92289 }, { "name": "depth_confidence", "l2": 10.741578, "mean_abs": 0.198832, "relative": 0.000356 }, { "name": "video_fisheye_cam0", "l2": 5.2355, "mean_abs": 0.108761, "relative": 0.948037 }, { "name": "video_fisheye_cam1", "l2": 5.432359, "mean_abs": 0.114914, "relative": 0.897651 }, { "name": "video_fisheye_cam2", "l2": 5.80329, "mean_abs": 0.117898, "relative": 0.926409 }, { "name": "video_fisheye_cam3", "l2": 5.463886, "mean_abs": 0.111448, "relative": 0.914862 }, { "name": "video_stereo_left", "l2": 5.345069, "mean_abs": 0.117318, "relative": 0.868584 }, { "name": "video_stereo_right", "l2": 6.372193, "mean_abs": 0.142295, "relative": 0.898345 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.03012, "mean_abs": 0.080863, "relative": 0.726549 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01266, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 157, "start_frame": 785, "end_frame": 804, "center_frame": 794, "action": "Position kettle to pour", "subtask": "Position kettle to pour", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.789947, "mean_abs": 0.230391, "relative": 0.910178 }, { "name": "hand_right_joints", "l2": 11.696603, "mean_abs": 0.271408, "relative": 0.922005 }, { "name": "body_joints", "l2": 5.284648, "mean_abs": 0.075471, "relative": 0.713793 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.044321, "mean_abs": 0.005621, "relative": 0.104592 }, { "name": "camera_rotation_matrix", "l2": 3.000495, "mean_abs": 0.207055, "relative": 0.921403 }, { "name": "imu_accel_gyro", "l2": 17.099831, "mean_abs": 1.076683, "relative": 0.922595 }, { "name": "depth_confidence", "l2": 10.738502, "mean_abs": 0.197814, "relative": 0.000356 }, { "name": "video_fisheye_cam0", "l2": 5.241539, "mean_abs": 0.108581, "relative": 0.949131 }, { "name": "video_fisheye_cam1", "l2": 5.418136, "mean_abs": 0.114296, "relative": 0.895301 }, { "name": "video_fisheye_cam2", "l2": 5.81152, "mean_abs": 0.117584, "relative": 0.927723 }, { "name": "video_fisheye_cam3", "l2": 5.486198, "mean_abs": 0.111637, "relative": 0.918598 }, { "name": "video_stereo_left", "l2": 5.343143, "mean_abs": 0.116175, "relative": 0.868271 }, { "name": "video_stereo_right", "l2": 6.356672, "mean_abs": 0.140552, "relative": 0.896157 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.208422, "mean_abs": 0.088845, "relative": 0.790361 }, { "name": "caption_objects_interaction_text", "l2": 1.986051, "mean_abs": 0.020094, "relative": 0.911946 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 158, "start_frame": 790, "end_frame": 809, "center_frame": 799, "action": "", "subtask": "", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.797733, "mean_abs": 0.231108, "relative": 0.910835 }, { "name": "hand_right_joints", "l2": 11.788424, "mean_abs": 0.271748, "relative": 0.929243 }, { "name": "body_joints", "l2": 5.307271, "mean_abs": 0.076471, "relative": 0.716849 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.045993, "mean_abs": 0.006067, "relative": 0.108537 }, { "name": "camera_rotation_matrix", "l2": 3.001233, "mean_abs": 0.20573, "relative": 0.92163 }, { "name": "imu_accel_gyro", "l2": 17.078167, "mean_abs": 1.070585, "relative": 0.921426 }, { "name": "depth_confidence", "l2": 10.732872, "mean_abs": 0.197228, "relative": 0.000356 }, { "name": "video_fisheye_cam0", "l2": 5.252236, "mean_abs": 0.108725, "relative": 0.951068 }, { "name": "video_fisheye_cam1", "l2": 5.404417, "mean_abs": 0.113536, "relative": 0.893034 }, { "name": "video_fisheye_cam2", "l2": 5.80271, "mean_abs": 0.117322, "relative": 0.926316 }, { "name": "video_fisheye_cam3", "l2": 5.533093, "mean_abs": 0.11237, "relative": 0.92645 }, { "name": "video_stereo_left", "l2": 5.328437, "mean_abs": 0.11506, "relative": 0.865882 }, { "name": "video_stereo_right", "l2": 6.334939, "mean_abs": 0.138931, "relative": 0.893094 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.461694, "mean_abs": 0.102978, "relative": 0.881003 }, { "name": "caption_objects_interaction_text", "l2": 1.986051, "mean_abs": 0.020239, "relative": 0.911946 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 159, "start_frame": 795, "end_frame": 814, "center_frame": 804, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.832123, "mean_abs": 0.233945, "relative": 0.913736 }, { "name": "hand_right_joints", "l2": 11.915081, "mean_abs": 0.279516, "relative": 0.939227 }, { "name": "body_joints", "l2": 5.369993, "mean_abs": 0.080066, "relative": 0.725321 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.057903, "mean_abs": 0.00824, "relative": 0.136642 }, { "name": "camera_rotation_matrix", "l2": 3.001289, "mean_abs": 0.206305, "relative": 0.921647 }, { "name": "imu_accel_gyro", "l2": 17.010798, "mean_abs": 1.078423, "relative": 0.917791 }, { "name": "depth_confidence", "l2": 10.695361, "mean_abs": 0.199121, "relative": 0.000355 }, { "name": "video_fisheye_cam0", "l2": 5.26384, "mean_abs": 0.109668, "relative": 0.953169 }, { "name": "video_fisheye_cam1", "l2": 5.385037, "mean_abs": 0.114215, "relative": 0.889832 }, { "name": "video_fisheye_cam2", "l2": 5.771148, "mean_abs": 0.117657, "relative": 0.921278 }, { "name": "video_fisheye_cam3", "l2": 5.600146, "mean_abs": 0.11355, "relative": 0.937677 }, { "name": "video_stereo_left", "l2": 5.315373, "mean_abs": 0.117444, "relative": 0.863759 }, { "name": "video_stereo_right", "l2": 6.305162, "mean_abs": 0.142468, "relative": 0.888896 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.237571, "mean_abs": 0.090948, "relative": 0.800793 }, { "name": "caption_objects_interaction_text", "l2": 1.986051, "mean_abs": 0.01973, "relative": 0.911946 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 160, "start_frame": 800, "end_frame": 819, "center_frame": 809, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.919827, "mean_abs": 0.245226, "relative": 0.921134 }, { "name": "hand_right_joints", "l2": 11.831227, "mean_abs": 0.280258, "relative": 0.932617 }, { "name": "body_joints", "l2": 5.58271, "mean_abs": 0.087371, "relative": 0.754052 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.096262, "mean_abs": 0.014386, "relative": 0.227166 }, { "name": "camera_rotation_matrix", "l2": 3.005735, "mean_abs": 0.208164, "relative": 0.923012 }, { "name": "imu_accel_gyro", "l2": 16.950384, "mean_abs": 1.091613, "relative": 0.914532 }, { "name": "depth_confidence", "l2": 10.610142, "mean_abs": 0.20096, "relative": 0.000352 }, { "name": "video_fisheye_cam0", "l2": 5.282048, "mean_abs": 0.110924, "relative": 0.956466 }, { "name": "video_fisheye_cam1", "l2": 5.359334, "mean_abs": 0.115167, "relative": 0.885585 }, { "name": "video_fisheye_cam2", "l2": 5.751052, "mean_abs": 0.118344, "relative": 0.91807 }, { "name": "video_fisheye_cam3", "l2": 5.682478, "mean_abs": 0.117169, "relative": 0.951462 }, { "name": "video_stereo_left", "l2": 5.330116, "mean_abs": 0.120525, "relative": 0.866154 }, { "name": "video_stereo_right", "l2": 6.310002, "mean_abs": 0.145666, "relative": 0.889578 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.242239, "mean_abs": 0.09298, "relative": 0.802464 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 161, "start_frame": 805, "end_frame": 824, "center_frame": 814, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.059628, "mean_abs": 0.252677, "relative": 0.932927 }, { "name": "hand_right_joints", "l2": 11.761386, "mean_abs": 0.287169, "relative": 0.927112 }, { "name": "body_joints", "l2": 5.923608, "mean_abs": 0.100214, "relative": 0.800097 }, { "name": "body_contacts", "l2": 3.464043, "mean_abs": 0.081719, "relative": 0.999608 }, { "name": "camera_translation", "l2": 0.156764, "mean_abs": 0.023335, "relative": 0.369941 }, { "name": "camera_rotation_matrix", "l2": 3.009383, "mean_abs": 0.209418, "relative": 0.924133 }, { "name": "imu_accel_gyro", "l2": 17.051094, "mean_abs": 1.112321, "relative": 0.919965 }, { "name": "depth_confidence", "l2": 10.49437, "mean_abs": 0.200971, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.300658, "mean_abs": 0.11203, "relative": 0.959836 }, { "name": "video_fisheye_cam1", "l2": 5.329314, "mean_abs": 0.11562, "relative": 0.880624 }, { "name": "video_fisheye_cam2", "l2": 5.717816, "mean_abs": 0.11845, "relative": 0.912764 }, { "name": "video_fisheye_cam3", "l2": 5.719262, "mean_abs": 0.119788, "relative": 0.957621 }, { "name": "video_stereo_left", "l2": 5.34933, "mean_abs": 0.122597, "relative": 0.869277 }, { "name": "video_stereo_right", "l2": 6.348277, "mean_abs": 0.14854, "relative": 0.894974 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.20339, "mean_abs": 0.089255, "relative": 0.78856 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 162, "start_frame": 810, "end_frame": 829, "center_frame": 819, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.190795, "mean_abs": 0.256861, "relative": 0.943991 }, { "name": "hand_right_joints", "l2": 11.711699, "mean_abs": 0.287707, "relative": 0.923195 }, { "name": "body_joints", "l2": 6.167294, "mean_abs": 0.107354, "relative": 0.833011 }, { "name": "body_contacts", "l2": 3.464043, "mean_abs": 0.081697, "relative": 0.999608 }, { "name": "camera_translation", "l2": 0.214169, "mean_abs": 0.03142, "relative": 0.50541 }, { "name": "camera_rotation_matrix", "l2": 3.008209, "mean_abs": 0.208594, "relative": 0.923772 }, { "name": "imu_accel_gyro", "l2": 17.133936, "mean_abs": 1.130398, "relative": 0.924435 }, { "name": "depth_confidence", "l2": 10.325766, "mean_abs": 0.199725, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.306875, "mean_abs": 0.11213, "relative": 0.960962 }, { "name": "video_fisheye_cam1", "l2": 5.279931, "mean_abs": 0.115171, "relative": 0.872464 }, { "name": "video_fisheye_cam2", "l2": 5.67892, "mean_abs": 0.117855, "relative": 0.906555 }, { "name": "video_fisheye_cam3", "l2": 5.726582, "mean_abs": 0.121066, "relative": 0.958847 }, { "name": "video_stereo_left", "l2": 5.326922, "mean_abs": 0.122181, "relative": 0.865635 }, { "name": "video_stereo_right", "l2": 6.308097, "mean_abs": 0.147112, "relative": 0.889309 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.215359, "mean_abs": 0.090217, "relative": 0.792844 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 163, "start_frame": 815, "end_frame": 834, "center_frame": 824, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.276291, "mean_abs": 0.256338, "relative": 0.951203 }, { "name": "hand_right_joints", "l2": 11.650691, "mean_abs": 0.288469, "relative": 0.918386 }, { "name": "body_joints", "l2": 6.362873, "mean_abs": 0.109986, "relative": 0.859428 }, { "name": "body_contacts", "l2": 3.464043, "mean_abs": 0.081694, "relative": 0.999608 }, { "name": "camera_translation", "l2": 0.222525, "mean_abs": 0.031601, "relative": 0.525127 }, { "name": "camera_rotation_matrix", "l2": 3.002904, "mean_abs": 0.203648, "relative": 0.922143 }, { "name": "imu_accel_gyro", "l2": 17.183241, "mean_abs": 1.107862, "relative": 0.927095 }, { "name": "depth_confidence", "l2": 10.080261, "mean_abs": 0.194894, "relative": 0.000335 }, { "name": "video_fisheye_cam0", "l2": 5.319976, "mean_abs": 0.111599, "relative": 0.963334 }, { "name": "video_fisheye_cam1", "l2": 5.228591, "mean_abs": 0.113752, "relative": 0.863981 }, { "name": "video_fisheye_cam2", "l2": 5.638898, "mean_abs": 0.116399, "relative": 0.900166 }, { "name": "video_fisheye_cam3", "l2": 5.739402, "mean_abs": 0.121486, "relative": 0.960994 }, { "name": "video_stereo_left", "l2": 5.271599, "mean_abs": 0.120042, "relative": 0.856645 }, { "name": "video_stereo_right", "l2": 6.230876, "mean_abs": 0.144141, "relative": 0.878423 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.048005, "mean_abs": 0.079608, "relative": 0.73295 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 164, "start_frame": 820, "end_frame": 839, "center_frame": 829, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.373411, "mean_abs": 0.257178, "relative": 0.959396 }, { "name": "hand_right_joints", "l2": 11.381686, "mean_abs": 0.27203, "relative": 0.897181 }, { "name": "body_joints", "l2": 6.556136, "mean_abs": 0.107925, "relative": 0.885532 }, { "name": "body_contacts", "l2": 3.464043, "mean_abs": 0.081693, "relative": 0.999608 }, { "name": "camera_translation", "l2": 0.184553, "mean_abs": 0.024623, "relative": 0.435519 }, { "name": "camera_rotation_matrix", "l2": 3.000588, "mean_abs": 0.197296, "relative": 0.921432 }, { "name": "imu_accel_gyro", "l2": 17.102659, "mean_abs": 1.07013, "relative": 0.922747 }, { "name": "depth_confidence", "l2": 9.856137, "mean_abs": 0.188455, "relative": 0.000327 }, { "name": "video_fisheye_cam0", "l2": 5.325456, "mean_abs": 0.110681, "relative": 0.964326 }, { "name": "video_fisheye_cam1", "l2": 5.16709, "mean_abs": 0.11142, "relative": 0.853818 }, { "name": "video_fisheye_cam2", "l2": 5.59667, "mean_abs": 0.11434, "relative": 0.893425 }, { "name": "video_fisheye_cam3", "l2": 5.683048, "mean_abs": 0.120259, "relative": 0.951558 }, { "name": "video_stereo_left", "l2": 5.210042, "mean_abs": 0.116232, "relative": 0.846642 }, { "name": "video_stereo_right", "l2": 6.155252, "mean_abs": 0.139244, "relative": 0.867761 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.086201, "mean_abs": 0.082705, "relative": 0.74662 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 165, "start_frame": 825, "end_frame": 844, "center_frame": 834, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.413672, "mean_abs": 0.256253, "relative": 0.962792 }, { "name": "hand_right_joints", "l2": 11.370933, "mean_abs": 0.272374, "relative": 0.896334 }, { "name": "body_joints", "l2": 6.673811, "mean_abs": 0.107813, "relative": 0.901426 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.111882, "mean_abs": 0.014819, "relative": 0.264026 }, { "name": "camera_rotation_matrix", "l2": 3.001287, "mean_abs": 0.194732, "relative": 0.921646 }, { "name": "imu_accel_gyro", "l2": 17.237556, "mean_abs": 1.053192, "relative": 0.930026 }, { "name": "depth_confidence", "l2": 9.686816, "mean_abs": 0.183039, "relative": 0.000321 }, { "name": "video_fisheye_cam0", "l2": 5.321447, "mean_abs": 0.110002, "relative": 0.9636 }, { "name": "video_fisheye_cam1", "l2": 5.115541, "mean_abs": 0.109366, "relative": 0.8453 }, { "name": "video_fisheye_cam2", "l2": 5.562623, "mean_abs": 0.11282, "relative": 0.88799 }, { "name": "video_fisheye_cam3", "l2": 5.604311, "mean_abs": 0.116607, "relative": 0.938374 }, { "name": "video_stereo_left", "l2": 5.165147, "mean_abs": 0.112733, "relative": 0.839347 }, { "name": "video_stereo_right", "l2": 6.119401, "mean_abs": 0.135863, "relative": 0.862707 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.00965, "mean_abs": 0.078085, "relative": 0.719224 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 166, "start_frame": 830, "end_frame": 849, "center_frame": 839, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.403367, "mean_abs": 0.258375, "relative": 0.961923 }, { "name": "hand_right_joints", "l2": 11.362397, "mean_abs": 0.27481, "relative": 0.895661 }, { "name": "body_joints", "l2": 6.777264, "mean_abs": 0.108923, "relative": 0.915399 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.041397, "mean_abs": 0.005732, "relative": 0.097691 }, { "name": "camera_rotation_matrix", "l2": 2.999942, "mean_abs": 0.192154, "relative": 0.921234 }, { "name": "imu_accel_gyro", "l2": 17.278587, "mean_abs": 1.043376, "relative": 0.932239 }, { "name": "depth_confidence", "l2": 9.546521, "mean_abs": 0.1777, "relative": 0.000317 }, { "name": "video_fisheye_cam0", "l2": 5.317676, "mean_abs": 0.1097, "relative": 0.962917 }, { "name": "video_fisheye_cam1", "l2": 5.073881, "mean_abs": 0.107504, "relative": 0.838416 }, { "name": "video_fisheye_cam2", "l2": 5.540687, "mean_abs": 0.111806, "relative": 0.884488 }, { "name": "video_fisheye_cam3", "l2": 5.548725, "mean_abs": 0.114235, "relative": 0.929067 }, { "name": "video_stereo_left", "l2": 5.131965, "mean_abs": 0.110536, "relative": 0.833954 }, { "name": "video_stereo_right", "l2": 6.08246, "mean_abs": 0.133622, "relative": 0.857499 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.0116, "mean_abs": 0.077293, "relative": 0.719921 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 167, "start_frame": 835, "end_frame": 854, "center_frame": 844, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.359066, "mean_abs": 0.256662, "relative": 0.958186 }, { "name": "hand_right_joints", "l2": 11.331625, "mean_abs": 0.276367, "relative": 0.893235 }, { "name": "body_joints", "l2": 6.835524, "mean_abs": 0.11069, "relative": 0.923269 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013626, "mean_abs": 0.00206, "relative": 0.032156 }, { "name": "camera_rotation_matrix", "l2": 2.999978, "mean_abs": 0.190279, "relative": 0.921245 }, { "name": "imu_accel_gyro", "l2": 17.106991, "mean_abs": 1.024935, "relative": 0.922981 }, { "name": "depth_confidence", "l2": 9.501935, "mean_abs": 0.175661, "relative": 0.000315 }, { "name": "video_fisheye_cam0", "l2": 5.298328, "mean_abs": 0.109351, "relative": 0.959414 }, { "name": "video_fisheye_cam1", "l2": 5.072228, "mean_abs": 0.107067, "relative": 0.838143 }, { "name": "video_fisheye_cam2", "l2": 5.52747, "mean_abs": 0.111023, "relative": 0.882378 }, { "name": "video_fisheye_cam3", "l2": 5.520122, "mean_abs": 0.112932, "relative": 0.924278 }, { "name": "video_stereo_left", "l2": 5.116335, "mean_abs": 0.10954, "relative": 0.831415 }, { "name": "video_stereo_right", "l2": 6.082347, "mean_abs": 0.132882, "relative": 0.857483 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.067832, "mean_abs": 0.078618, "relative": 0.740046 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 168, "start_frame": 840, "end_frame": 859, "center_frame": 849, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.339023, "mean_abs": 0.256254, "relative": 0.956495 }, { "name": "hand_right_joints", "l2": 11.266011, "mean_abs": 0.278499, "relative": 0.888063 }, { "name": "body_joints", "l2": 6.880578, "mean_abs": 0.11185, "relative": 0.929354 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016664, "mean_abs": 0.002299, "relative": 0.039325 }, { "name": "camera_rotation_matrix", "l2": 3.000112, "mean_abs": 0.190269, "relative": 0.921286 }, { "name": "imu_accel_gyro", "l2": 17.085815, "mean_abs": 1.018897, "relative": 0.921839 }, { "name": "depth_confidence", "l2": 9.482618, "mean_abs": 0.175308, "relative": 0.000315 }, { "name": "video_fisheye_cam0", "l2": 5.290892, "mean_abs": 0.10919, "relative": 0.958067 }, { "name": "video_fisheye_cam1", "l2": 5.105458, "mean_abs": 0.108767, "relative": 0.843634 }, { "name": "video_fisheye_cam2", "l2": 5.521963, "mean_abs": 0.110832, "relative": 0.881499 }, { "name": "video_fisheye_cam3", "l2": 5.521174, "mean_abs": 0.113605, "relative": 0.924454 }, { "name": "video_stereo_left", "l2": 5.113044, "mean_abs": 0.109373, "relative": 0.83088 }, { "name": "video_stereo_right", "l2": 6.121073, "mean_abs": 0.134909, "relative": 0.862943 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.077293, "mean_abs": 0.07872, "relative": 0.743432 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 169, "start_frame": 845, "end_frame": 864, "center_frame": 854, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.314435, "mean_abs": 0.256451, "relative": 0.954421 }, { "name": "hand_right_joints", "l2": 11.229321, "mean_abs": 0.278511, "relative": 0.885171 }, { "name": "body_joints", "l2": 6.898432, "mean_abs": 0.111379, "relative": 0.931766 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019567, "mean_abs": 0.00269, "relative": 0.046175 }, { "name": "camera_rotation_matrix", "l2": 2.999937, "mean_abs": 0.189867, "relative": 0.921232 }, { "name": "imu_accel_gyro", "l2": 17.078878, "mean_abs": 1.011649, "relative": 0.921464 }, { "name": "depth_confidence", "l2": 9.476306, "mean_abs": 0.175345, "relative": 0.000315 }, { "name": "video_fisheye_cam0", "l2": 5.279345, "mean_abs": 0.10913, "relative": 0.955976 }, { "name": "video_fisheye_cam1", "l2": 5.120396, "mean_abs": 0.108876, "relative": 0.846102 }, { "name": "video_fisheye_cam2", "l2": 5.514048, "mean_abs": 0.110811, "relative": 0.880236 }, { "name": "video_fisheye_cam3", "l2": 5.513243, "mean_abs": 0.112934, "relative": 0.923126 }, { "name": "video_stereo_left", "l2": 5.102938, "mean_abs": 0.109106, "relative": 0.829238 }, { "name": "video_stereo_right", "l2": 6.142127, "mean_abs": 0.134763, "relative": 0.865911 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.031778, "mean_abs": 0.077465, "relative": 0.727143 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 170, "start_frame": 850, "end_frame": 869, "center_frame": 859, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.309831, "mean_abs": 0.25603, "relative": 0.954032 }, { "name": "hand_right_joints", "l2": 11.188503, "mean_abs": 0.27617, "relative": 0.881953 }, { "name": "body_joints", "l2": 6.897979, "mean_abs": 0.10988, "relative": 0.931704 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020183, "mean_abs": 0.002704, "relative": 0.04763 }, { "name": "camera_rotation_matrix", "l2": 2.999928, "mean_abs": 0.190195, "relative": 0.921229 }, { "name": "imu_accel_gyro", "l2": 17.077227, "mean_abs": 1.015065, "relative": 0.921375 }, { "name": "depth_confidence", "l2": 9.496689, "mean_abs": 0.175796, "relative": 0.000315 }, { "name": "video_fisheye_cam0", "l2": 5.270541, "mean_abs": 0.108837, "relative": 0.954382 }, { "name": "video_fisheye_cam1", "l2": 5.1451, "mean_abs": 0.109844, "relative": 0.850184 }, { "name": "video_fisheye_cam2", "l2": 5.517209, "mean_abs": 0.111078, "relative": 0.88074 }, { "name": "video_fisheye_cam3", "l2": 5.516545, "mean_abs": 0.113424, "relative": 0.923679 }, { "name": "video_stereo_left", "l2": 5.107958, "mean_abs": 0.109533, "relative": 0.830053 }, { "name": "video_stereo_right", "l2": 6.179324, "mean_abs": 0.136416, "relative": 0.871155 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.078356, "mean_abs": 0.081499, "relative": 0.743812 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 171, "start_frame": 855, "end_frame": 874, "center_frame": 864, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.306972, "mean_abs": 0.256551, "relative": 0.953791 }, { "name": "hand_right_joints", "l2": 11.195774, "mean_abs": 0.276899, "relative": 0.882527 }, { "name": "body_joints", "l2": 6.918779, "mean_abs": 0.109324, "relative": 0.934514 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013948, "mean_abs": 0.00194, "relative": 0.032916 }, { "name": "camera_rotation_matrix", "l2": 2.999815, "mean_abs": 0.189922, "relative": 0.921195 }, { "name": "imu_accel_gyro", "l2": 17.080465, "mean_abs": 1.008201, "relative": 0.92155 }, { "name": "depth_confidence", "l2": 9.503214, "mean_abs": 0.176418, "relative": 0.000315 }, { "name": "video_fisheye_cam0", "l2": 5.262077, "mean_abs": 0.10853, "relative": 0.95285 }, { "name": "video_fisheye_cam1", "l2": 5.166764, "mean_abs": 0.109368, "relative": 0.853764 }, { "name": "video_fisheye_cam2", "l2": 5.519354, "mean_abs": 0.110927, "relative": 0.881083 }, { "name": "video_fisheye_cam3", "l2": 5.518507, "mean_abs": 0.113411, "relative": 0.924007 }, { "name": "video_stereo_left", "l2": 5.110667, "mean_abs": 0.109394, "relative": 0.830494 }, { "name": "video_stereo_right", "l2": 6.200114, "mean_abs": 0.135892, "relative": 0.874086 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.059291, "mean_abs": 0.08245, "relative": 0.736989 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 172, "start_frame": 860, "end_frame": 879, "center_frame": 869, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.317634, "mean_abs": 0.256516, "relative": 0.954691 }, { "name": "hand_right_joints", "l2": 11.224071, "mean_abs": 0.278824, "relative": 0.884757 }, { "name": "body_joints", "l2": 6.929681, "mean_abs": 0.109711, "relative": 0.935986 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007439, "mean_abs": 0.001105, "relative": 0.017556 }, { "name": "camera_rotation_matrix", "l2": 2.999912, "mean_abs": 0.190234, "relative": 0.921224 }, { "name": "imu_accel_gyro", "l2": 17.070024, "mean_abs": 1.005824, "relative": 0.920987 }, { "name": "depth_confidence", "l2": 9.503912, "mean_abs": 0.17564, "relative": 0.000315 }, { "name": "video_fisheye_cam0", "l2": 5.263914, "mean_abs": 0.108564, "relative": 0.953182 }, { "name": "video_fisheye_cam1", "l2": 5.17682, "mean_abs": 0.109518, "relative": 0.855426 }, { "name": "video_fisheye_cam2", "l2": 5.523164, "mean_abs": 0.111035, "relative": 0.881691 }, { "name": "video_fisheye_cam3", "l2": 5.495472, "mean_abs": 0.113197, "relative": 0.920151 }, { "name": "video_stereo_left", "l2": 5.114067, "mean_abs": 0.10968, "relative": 0.831046 }, { "name": "video_stereo_right", "l2": 6.20545, "mean_abs": 0.136127, "relative": 0.874838 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.047264, "mean_abs": 0.081232, "relative": 0.732685 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 173, "start_frame": 865, "end_frame": 884, "center_frame": 874, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.314008, "mean_abs": 0.256742, "relative": 0.954385 }, { "name": "hand_right_joints", "l2": 11.276393, "mean_abs": 0.278039, "relative": 0.888881 }, { "name": "body_joints", "l2": 6.925621, "mean_abs": 0.10984, "relative": 0.935438 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010799, "mean_abs": 0.001424, "relative": 0.025483 }, { "name": "camera_rotation_matrix", "l2": 2.999986, "mean_abs": 0.190456, "relative": 0.921247 }, { "name": "imu_accel_gyro", "l2": 17.081743, "mean_abs": 1.012912, "relative": 0.921619 }, { "name": "depth_confidence", "l2": 9.525834, "mean_abs": 0.176394, "relative": 0.000316 }, { "name": "video_fisheye_cam0", "l2": 5.264641, "mean_abs": 0.109041, "relative": 0.953314 }, { "name": "video_fisheye_cam1", "l2": 5.174884, "mean_abs": 0.109801, "relative": 0.855106 }, { "name": "video_fisheye_cam2", "l2": 5.527403, "mean_abs": 0.111271, "relative": 0.882368 }, { "name": "video_fisheye_cam3", "l2": 5.488253, "mean_abs": 0.114371, "relative": 0.918942 }, { "name": "video_stereo_left", "l2": 5.119153, "mean_abs": 0.109888, "relative": 0.831873 }, { "name": "video_stereo_right", "l2": 6.204407, "mean_abs": 0.136345, "relative": 0.874691 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.15462, "mean_abs": 0.085443, "relative": 0.771106 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 174, "start_frame": 870, "end_frame": 889, "center_frame": 879, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.254303, "mean_abs": 0.258925, "relative": 0.949348 }, { "name": "hand_right_joints", "l2": 11.261493, "mean_abs": 0.279528, "relative": 0.887707 }, { "name": "body_joints", "l2": 6.956599, "mean_abs": 0.110582, "relative": 0.939622 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013931, "mean_abs": 0.001791, "relative": 0.032875 }, { "name": "camera_rotation_matrix", "l2": 3.000262, "mean_abs": 0.190652, "relative": 0.921332 }, { "name": "imu_accel_gyro", "l2": 17.15497, "mean_abs": 1.026906, "relative": 0.92557 }, { "name": "depth_confidence", "l2": 9.51242, "mean_abs": 0.176203, "relative": 0.000316 }, { "name": "video_fisheye_cam0", "l2": 5.262895, "mean_abs": 0.108569, "relative": 0.952998 }, { "name": "video_fisheye_cam1", "l2": 5.168589, "mean_abs": 0.109393, "relative": 0.854066 }, { "name": "video_fisheye_cam2", "l2": 5.523917, "mean_abs": 0.111251, "relative": 0.881811 }, { "name": "video_fisheye_cam3", "l2": 5.470693, "mean_abs": 0.112256, "relative": 0.916002 }, { "name": "video_stereo_left", "l2": 5.11228, "mean_abs": 0.109418, "relative": 0.830756 }, { "name": "video_stereo_right", "l2": 6.188134, "mean_abs": 0.135585, "relative": 0.872397 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.194823, "mean_abs": 0.088433, "relative": 0.785494 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 175, "start_frame": 875, "end_frame": 894, "center_frame": 884, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.208393, "mean_abs": 0.260791, "relative": 0.945476 }, { "name": "hand_right_joints", "l2": 11.166683, "mean_abs": 0.28249, "relative": 0.880233 }, { "name": "body_joints", "l2": 6.988589, "mean_abs": 0.111674, "relative": 0.943943 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009738, "mean_abs": 0.00166, "relative": 0.022979 }, { "name": "camera_rotation_matrix", "l2": 3.000073, "mean_abs": 0.190028, "relative": 0.921274 }, { "name": "imu_accel_gyro", "l2": 17.161663, "mean_abs": 1.018104, "relative": 0.925931 }, { "name": "depth_confidence", "l2": 9.510971, "mean_abs": 0.176173, "relative": 0.000316 }, { "name": "video_fisheye_cam0", "l2": 5.262958, "mean_abs": 0.108151, "relative": 0.953009 }, { "name": "video_fisheye_cam1", "l2": 5.165284, "mean_abs": 0.109269, "relative": 0.853519 }, { "name": "video_fisheye_cam2", "l2": 5.524758, "mean_abs": 0.111131, "relative": 0.881945 }, { "name": "video_fisheye_cam3", "l2": 5.431895, "mean_abs": 0.110954, "relative": 0.909505 }, { "name": "video_stereo_left", "l2": 5.11097, "mean_abs": 0.109563, "relative": 0.830543 }, { "name": "video_stereo_right", "l2": 6.189734, "mean_abs": 0.135701, "relative": 0.872623 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.125054, "mean_abs": 0.08499, "relative": 0.760525 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 176, "start_frame": 880, "end_frame": 899, "center_frame": 889, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.112418, "mean_abs": 0.26087, "relative": 0.93738 }, { "name": "hand_right_joints", "l2": 11.101859, "mean_abs": 0.283906, "relative": 0.875124 }, { "name": "body_joints", "l2": 6.989059, "mean_abs": 0.11274, "relative": 0.944007 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.037759, "mean_abs": 0.005783, "relative": 0.089107 }, { "name": "camera_rotation_matrix", "l2": 2.999998, "mean_abs": 0.19042, "relative": 0.921251 }, { "name": "imu_accel_gyro", "l2": 17.126472, "mean_abs": 1.025347, "relative": 0.924032 }, { "name": "depth_confidence", "l2": 9.49369, "mean_abs": 0.17694, "relative": 0.000315 }, { "name": "video_fisheye_cam0", "l2": 5.256516, "mean_abs": 0.108102, "relative": 0.951843 }, { "name": "video_fisheye_cam1", "l2": 5.182189, "mean_abs": 0.110319, "relative": 0.856313 }, { "name": "video_fisheye_cam2", "l2": 5.533306, "mean_abs": 0.111848, "relative": 0.88331 }, { "name": "video_fisheye_cam3", "l2": 5.406675, "mean_abs": 0.110873, "relative": 0.905283 }, { "name": "video_stereo_left", "l2": 5.11566, "mean_abs": 0.11065, "relative": 0.831305 }, { "name": "video_stereo_right", "l2": 6.197402, "mean_abs": 0.137017, "relative": 0.873704 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.053616, "mean_abs": 0.080634, "relative": 0.734958 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 177, "start_frame": 885, "end_frame": 904, "center_frame": 894, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.071718, "mean_abs": 0.259771, "relative": 0.933947 }, { "name": "hand_right_joints", "l2": 10.976954, "mean_abs": 0.281179, "relative": 0.865278 }, { "name": "body_joints", "l2": 6.960402, "mean_abs": 0.11332, "relative": 0.940136 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.078908, "mean_abs": 0.011951, "relative": 0.186211 }, { "name": "camera_rotation_matrix", "l2": 3.000937, "mean_abs": 0.193006, "relative": 0.921539 }, { "name": "imu_accel_gyro", "l2": 17.12607, "mean_abs": 1.03474, "relative": 0.92401 }, { "name": "depth_confidence", "l2": 9.5409, "mean_abs": 0.17969, "relative": 0.000317 }, { "name": "video_fisheye_cam0", "l2": 5.256768, "mean_abs": 0.108664, "relative": 0.951888 }, { "name": "video_fisheye_cam1", "l2": 5.21553, "mean_abs": 0.111979, "relative": 0.861822 }, { "name": "video_fisheye_cam2", "l2": 5.539023, "mean_abs": 0.11285, "relative": 0.884223 }, { "name": "video_fisheye_cam3", "l2": 5.406099, "mean_abs": 0.112178, "relative": 0.905186 }, { "name": "video_stereo_left", "l2": 5.12358, "mean_abs": 0.112479, "relative": 0.832592 }, { "name": "video_stereo_right", "l2": 6.222305, "mean_abs": 0.139133, "relative": 0.877214 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.013965, "mean_abs": 0.083141, "relative": 0.720768 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 178, "start_frame": 890, "end_frame": 909, "center_frame": 899, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.943751, "mean_abs": 0.25598, "relative": 0.923152 }, { "name": "hand_right_joints", "l2": 10.844923, "mean_abs": 0.275905, "relative": 0.85487 }, { "name": "body_joints", "l2": 6.921647, "mean_abs": 0.112788, "relative": 0.934901 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.124636, "mean_abs": 0.018358, "relative": 0.294124 }, { "name": "camera_rotation_matrix", "l2": 3.004709, "mean_abs": 0.195173, "relative": 0.922697 }, { "name": "imu_accel_gyro", "l2": 17.129183, "mean_abs": 1.050165, "relative": 0.924178 }, { "name": "depth_confidence", "l2": 9.558854, "mean_abs": 0.181694, "relative": 0.000317 }, { "name": "video_fisheye_cam0", "l2": 5.244966, "mean_abs": 0.108976, "relative": 0.949751 }, { "name": "video_fisheye_cam1", "l2": 5.245965, "mean_abs": 0.114228, "relative": 0.866851 }, { "name": "video_fisheye_cam2", "l2": 5.528902, "mean_abs": 0.114022, "relative": 0.882607 }, { "name": "video_fisheye_cam3", "l2": 5.404788, "mean_abs": 0.113564, "relative": 0.904966 }, { "name": "video_stereo_left", "l2": 5.120894, "mean_abs": 0.114466, "relative": 0.832155 }, { "name": "video_stereo_right", "l2": 6.233256, "mean_abs": 0.141145, "relative": 0.878758 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.962854, "mean_abs": 0.075549, "relative": 0.702476 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 179, "start_frame": 895, "end_frame": 914, "center_frame": 904, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.908257, "mean_abs": 0.252345, "relative": 0.920158 }, { "name": "hand_right_joints", "l2": 10.782304, "mean_abs": 0.273503, "relative": 0.849934 }, { "name": "body_joints", "l2": 6.826726, "mean_abs": 0.11065, "relative": 0.92208 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.162717, "mean_abs": 0.022313, "relative": 0.38399 }, { "name": "camera_rotation_matrix", "l2": 3.009913, "mean_abs": 0.197024, "relative": 0.924295 }, { "name": "imu_accel_gyro", "l2": 17.094788, "mean_abs": 1.060391, "relative": 0.922323 }, { "name": "depth_confidence", "l2": 9.578843, "mean_abs": 0.183205, "relative": 0.000318 }, { "name": "video_fisheye_cam0", "l2": 5.236964, "mean_abs": 0.109604, "relative": 0.948302 }, { "name": "video_fisheye_cam1", "l2": 5.292014, "mean_abs": 0.116411, "relative": 0.874461 }, { "name": "video_fisheye_cam2", "l2": 5.506941, "mean_abs": 0.115057, "relative": 0.879101 }, { "name": "video_fisheye_cam3", "l2": 5.39788, "mean_abs": 0.114408, "relative": 0.90381 }, { "name": "video_stereo_left", "l2": 5.103748, "mean_abs": 0.115666, "relative": 0.829369 }, { "name": "video_stereo_right", "l2": 6.243875, "mean_abs": 0.143911, "relative": 0.880255 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.982184, "mean_abs": 0.076051, "relative": 0.709394 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 180, "start_frame": 900, "end_frame": 919, "center_frame": 909, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.904391, "mean_abs": 0.254962, "relative": 0.919832 }, { "name": "hand_right_joints", "l2": 10.764128, "mean_abs": 0.275215, "relative": 0.848501 }, { "name": "body_joints", "l2": 6.701461, "mean_abs": 0.106267, "relative": 0.905161 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.17851, "mean_abs": 0.022016, "relative": 0.421258 }, { "name": "camera_rotation_matrix", "l2": 3.012766, "mean_abs": 0.19632, "relative": 0.925172 }, { "name": "imu_accel_gyro", "l2": 17.047062, "mean_abs": 1.045435, "relative": 0.919748 }, { "name": "depth_confidence", "l2": 9.632001, "mean_abs": 0.184484, "relative": 0.00032 }, { "name": "video_fisheye_cam0", "l2": 5.227034, "mean_abs": 0.109846, "relative": 0.946504 }, { "name": "video_fisheye_cam1", "l2": 5.331688, "mean_abs": 0.117912, "relative": 0.881016 }, { "name": "video_fisheye_cam2", "l2": 5.479428, "mean_abs": 0.115099, "relative": 0.874709 }, { "name": "video_fisheye_cam3", "l2": 5.396265, "mean_abs": 0.114233, "relative": 0.903539 }, { "name": "video_stereo_left", "l2": 5.060667, "mean_abs": 0.115638, "relative": 0.822368 }, { "name": "video_stereo_right", "l2": 6.242372, "mean_abs": 0.145971, "relative": 0.880044 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.985278, "mean_abs": 0.075497, "relative": 0.710501 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 181, "start_frame": 905, "end_frame": 924, "center_frame": 914, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.909078, "mean_abs": 0.255341, "relative": 0.920227 }, { "name": "hand_right_joints", "l2": 10.829954, "mean_abs": 0.279404, "relative": 0.85369 }, { "name": "body_joints", "l2": 6.556697, "mean_abs": 0.099619, "relative": 0.885608 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.174191, "mean_abs": 0.01881, "relative": 0.411067 }, { "name": "camera_rotation_matrix", "l2": 3.01112, "mean_abs": 0.193037, "relative": 0.924666 }, { "name": "imu_accel_gyro", "l2": 16.991074, "mean_abs": 1.047167, "relative": 0.916727 }, { "name": "depth_confidence", "l2": 9.619452, "mean_abs": 0.183611, "relative": 0.000319 }, { "name": "video_fisheye_cam0", "l2": 5.213222, "mean_abs": 0.109392, "relative": 0.944003 }, { "name": "video_fisheye_cam1", "l2": 5.34391, "mean_abs": 0.118084, "relative": 0.883036 }, { "name": "video_fisheye_cam2", "l2": 5.446044, "mean_abs": 0.114315, "relative": 0.86938 }, { "name": "video_fisheye_cam3", "l2": 5.389395, "mean_abs": 0.113294, "relative": 0.902389 }, { "name": "video_stereo_left", "l2": 5.020892, "mean_abs": 0.114621, "relative": 0.815905 }, { "name": "video_stereo_right", "l2": 6.216307, "mean_abs": 0.145862, "relative": 0.876369 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.134676, "mean_abs": 0.081575, "relative": 0.763969 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 182, "start_frame": 910, "end_frame": 929, "center_frame": 919, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.948659, "mean_abs": 0.258013, "relative": 0.923566 }, { "name": "hand_right_joints", "l2": 10.900488, "mean_abs": 0.283226, "relative": 0.85925 }, { "name": "body_joints", "l2": 6.451688, "mean_abs": 0.095146, "relative": 0.871424 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.159311, "mean_abs": 0.016958, "relative": 0.375951 }, { "name": "camera_rotation_matrix", "l2": 3.006992, "mean_abs": 0.188128, "relative": 0.923398 }, { "name": "imu_accel_gyro", "l2": 17.0427, "mean_abs": 1.048189, "relative": 0.919512 }, { "name": "depth_confidence", "l2": 9.584286, "mean_abs": 0.181735, "relative": 0.000318 }, { "name": "video_fisheye_cam0", "l2": 5.183214, "mean_abs": 0.108587, "relative": 0.938569 }, { "name": "video_fisheye_cam1", "l2": 5.329128, "mean_abs": 0.116777, "relative": 0.880593 }, { "name": "video_fisheye_cam2", "l2": 5.395798, "mean_abs": 0.112773, "relative": 0.861359 }, { "name": "video_fisheye_cam3", "l2": 5.372471, "mean_abs": 0.111641, "relative": 0.899555 }, { "name": "video_stereo_left", "l2": 4.968061, "mean_abs": 0.113327, "relative": 0.80732 }, { "name": "video_stereo_right", "l2": 6.093868, "mean_abs": 0.141915, "relative": 0.859108 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.055353, "mean_abs": 0.07945, "relative": 0.73558 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 183, "start_frame": 915, "end_frame": 934, "center_frame": 924, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.992531, "mean_abs": 0.258152, "relative": 0.927267 }, { "name": "hand_right_joints", "l2": 11.041932, "mean_abs": 0.288039, "relative": 0.8704 }, { "name": "body_joints", "l2": 6.401901, "mean_abs": 0.093545, "relative": 0.8647 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.133505, "mean_abs": 0.015393, "relative": 0.315053 }, { "name": "camera_rotation_matrix", "l2": 3.004107, "mean_abs": 0.185073, "relative": 0.922512 }, { "name": "imu_accel_gyro", "l2": 17.0627, "mean_abs": 1.051153, "relative": 0.920591 }, { "name": "depth_confidence", "l2": 9.555308, "mean_abs": 0.179795, "relative": 0.000317 }, { "name": "video_fisheye_cam0", "l2": 5.173654, "mean_abs": 0.10842, "relative": 0.936838 }, { "name": "video_fisheye_cam1", "l2": 5.308058, "mean_abs": 0.115004, "relative": 0.877112 }, { "name": "video_fisheye_cam2", "l2": 5.370031, "mean_abs": 0.111908, "relative": 0.857246 }, { "name": "video_fisheye_cam3", "l2": 5.382757, "mean_abs": 0.1114, "relative": 0.901278 }, { "name": "video_stereo_left", "l2": 4.920107, "mean_abs": 0.111628, "relative": 0.799527 }, { "name": "video_stereo_right", "l2": 6.012901, "mean_abs": 0.137813, "relative": 0.847693 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.173407, "mean_abs": 0.091707, "relative": 0.77783 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 184, "start_frame": 920, "end_frame": 939, "center_frame": 929, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.017417, "mean_abs": 0.257778, "relative": 0.929366 }, { "name": "hand_right_joints", "l2": 11.157887, "mean_abs": 0.291083, "relative": 0.87954 }, { "name": "body_joints", "l2": 6.373491, "mean_abs": 0.092742, "relative": 0.860862 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.114834, "mean_abs": 0.015261, "relative": 0.270991 }, { "name": "camera_rotation_matrix", "l2": 3.002807, "mean_abs": 0.181987, "relative": 0.922113 }, { "name": "imu_accel_gyro", "l2": 17.216881, "mean_abs": 1.075785, "relative": 0.92891 }, { "name": "depth_confidence", "l2": 9.470864, "mean_abs": 0.177688, "relative": 0.000314 }, { "name": "video_fisheye_cam0", "l2": 5.169426, "mean_abs": 0.10873, "relative": 0.936073 }, { "name": "video_fisheye_cam1", "l2": 5.278816, "mean_abs": 0.113152, "relative": 0.87228 }, { "name": "video_fisheye_cam2", "l2": 5.357353, "mean_abs": 0.111577, "relative": 0.855222 }, { "name": "video_fisheye_cam3", "l2": 5.440129, "mean_abs": 0.112566, "relative": 0.910884 }, { "name": "video_stereo_left", "l2": 4.884628, "mean_abs": 0.109743, "relative": 0.793762 }, { "name": "video_stereo_right", "l2": 5.916072, "mean_abs": 0.135389, "relative": 0.834042 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.160637, "mean_abs": 0.089826, "relative": 0.773259 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 185, "start_frame": 925, "end_frame": 944, "center_frame": 934, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.059586, "mean_abs": 0.258552, "relative": 0.932923 }, { "name": "hand_right_joints", "l2": 11.245589, "mean_abs": 0.293143, "relative": 0.886453 }, { "name": "body_joints", "l2": 6.343912, "mean_abs": 0.092152, "relative": 0.856867 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.094992, "mean_abs": 0.01302, "relative": 0.224168 }, { "name": "camera_rotation_matrix", "l2": 3.002228, "mean_abs": 0.179962, "relative": 0.921935 }, { "name": "imu_accel_gyro", "l2": 17.114367, "mean_abs": 1.079266, "relative": 0.923379 }, { "name": "depth_confidence", "l2": 9.432532, "mean_abs": 0.176435, "relative": 0.000313 }, { "name": "video_fisheye_cam0", "l2": 5.171483, "mean_abs": 0.108728, "relative": 0.936445 }, { "name": "video_fisheye_cam1", "l2": 5.263929, "mean_abs": 0.112462, "relative": 0.86982 }, { "name": "video_fisheye_cam2", "l2": 5.369187, "mean_abs": 0.111613, "relative": 0.857111 }, { "name": "video_fisheye_cam3", "l2": 5.448298, "mean_abs": 0.111802, "relative": 0.912252 }, { "name": "video_stereo_left", "l2": 4.903485, "mean_abs": 0.109783, "relative": 0.796826 }, { "name": "video_stereo_right", "l2": 5.874901, "mean_abs": 0.134552, "relative": 0.828238 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.184093, "mean_abs": 0.090472, "relative": 0.781654 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 186, "start_frame": 930, "end_frame": 949, "center_frame": 939, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.112875, "mean_abs": 0.260207, "relative": 0.937418 }, { "name": "hand_right_joints", "l2": 11.343208, "mean_abs": 0.293436, "relative": 0.894148 }, { "name": "body_joints", "l2": 6.291753, "mean_abs": 0.091168, "relative": 0.849822 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.075479, "mean_abs": 0.010592, "relative": 0.178121 }, { "name": "camera_rotation_matrix", "l2": 3.001864, "mean_abs": 0.178284, "relative": 0.921824 }, { "name": "imu_accel_gyro", "l2": 17.079781, "mean_abs": 1.067621, "relative": 0.921513 }, { "name": "depth_confidence", "l2": 9.409907, "mean_abs": 0.176085, "relative": 0.000312 }, { "name": "video_fisheye_cam0", "l2": 5.166334, "mean_abs": 0.108687, "relative": 0.935513 }, { "name": "video_fisheye_cam1", "l2": 5.243285, "mean_abs": 0.111531, "relative": 0.866409 }, { "name": "video_fisheye_cam2", "l2": 5.394982, "mean_abs": 0.111959, "relative": 0.861229 }, { "name": "video_fisheye_cam3", "l2": 5.456901, "mean_abs": 0.111572, "relative": 0.913692 }, { "name": "video_stereo_left", "l2": 4.930137, "mean_abs": 0.110101, "relative": 0.801157 }, { "name": "video_stereo_right", "l2": 5.836785, "mean_abs": 0.133086, "relative": 0.822864 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.292155, "mean_abs": 0.097464, "relative": 0.820328 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 187, "start_frame": 935, "end_frame": 954, "center_frame": 944, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.139729, "mean_abs": 0.261353, "relative": 0.939684 }, { "name": "hand_right_joints", "l2": 11.404177, "mean_abs": 0.290595, "relative": 0.898954 }, { "name": "body_joints", "l2": 6.238452, "mean_abs": 0.089824, "relative": 0.842623 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.055862, "mean_abs": 0.00772, "relative": 0.131826 }, { "name": "camera_rotation_matrix", "l2": 3.001459, "mean_abs": 0.176532, "relative": 0.921699 }, { "name": "imu_accel_gyro", "l2": 17.135288, "mean_abs": 1.068424, "relative": 0.924508 }, { "name": "depth_confidence", "l2": 9.393354, "mean_abs": 0.175811, "relative": 0.000312 }, { "name": "video_fisheye_cam0", "l2": 5.168304, "mean_abs": 0.108584, "relative": 0.935869 }, { "name": "video_fisheye_cam1", "l2": 5.23081, "mean_abs": 0.110729, "relative": 0.864347 }, { "name": "video_fisheye_cam2", "l2": 5.431012, "mean_abs": 0.112624, "relative": 0.86698 }, { "name": "video_fisheye_cam3", "l2": 5.478172, "mean_abs": 0.111441, "relative": 0.917254 }, { "name": "video_stereo_left", "l2": 4.964778, "mean_abs": 0.110322, "relative": 0.806786 }, { "name": "video_stereo_right", "l2": 5.814407, "mean_abs": 0.131911, "relative": 0.819709 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.109498, "mean_abs": 0.080634, "relative": 0.754958 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 188, "start_frame": 940, "end_frame": 959, "center_frame": 949, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.162965, "mean_abs": 0.261811, "relative": 0.941644 }, { "name": "hand_right_joints", "l2": 11.465128, "mean_abs": 0.289316, "relative": 0.903759 }, { "name": "body_joints", "l2": 6.183807, "mean_abs": 0.088265, "relative": 0.835242 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.042884, "mean_abs": 0.005946, "relative": 0.1012 }, { "name": "camera_rotation_matrix", "l2": 3.000791, "mean_abs": 0.174675, "relative": 0.921494 }, { "name": "imu_accel_gyro", "l2": 17.045748, "mean_abs": 1.064718, "relative": 0.919677 }, { "name": "depth_confidence", "l2": 9.36214, "mean_abs": 0.174934, "relative": 0.000311 }, { "name": "video_fisheye_cam0", "l2": 5.173596, "mean_abs": 0.108445, "relative": 0.936828 }, { "name": "video_fisheye_cam1", "l2": 5.232838, "mean_abs": 0.110562, "relative": 0.864682 }, { "name": "video_fisheye_cam2", "l2": 5.452669, "mean_abs": 0.112682, "relative": 0.870437 }, { "name": "video_fisheye_cam3", "l2": 5.467614, "mean_abs": 0.110947, "relative": 0.915486 }, { "name": "video_stereo_left", "l2": 4.987473, "mean_abs": 0.11027, "relative": 0.810474 }, { "name": "video_stereo_right", "l2": 5.81801, "mean_abs": 0.131508, "relative": 0.820217 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.094758, "mean_abs": 0.078942, "relative": 0.749682 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 189, "start_frame": 945, "end_frame": 964, "center_frame": 954, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.205987, "mean_abs": 0.260854, "relative": 0.945273 }, { "name": "hand_right_joints", "l2": 11.483993, "mean_abs": 0.286608, "relative": 0.905246 }, { "name": "body_joints", "l2": 6.118465, "mean_abs": 0.086281, "relative": 0.826416 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.032453, "mean_abs": 0.004837, "relative": 0.076585 }, { "name": "camera_rotation_matrix", "l2": 3.000693, "mean_abs": 0.17335, "relative": 0.921464 }, { "name": "imu_accel_gyro", "l2": 17.016338, "mean_abs": 1.06781, "relative": 0.91809 }, { "name": "depth_confidence", "l2": 9.345168, "mean_abs": 0.175011, "relative": 0.00031 }, { "name": "video_fisheye_cam0", "l2": 5.181742, "mean_abs": 0.108392, "relative": 0.938303 }, { "name": "video_fisheye_cam1", "l2": 5.240581, "mean_abs": 0.110216, "relative": 0.865962 }, { "name": "video_fisheye_cam2", "l2": 5.479819, "mean_abs": 0.112415, "relative": 0.874772 }, { "name": "video_fisheye_cam3", "l2": 5.462653, "mean_abs": 0.110582, "relative": 0.914655 }, { "name": "video_stereo_left", "l2": 5.024756, "mean_abs": 0.110576, "relative": 0.816533 }, { "name": "video_stereo_right", "l2": 5.845592, "mean_abs": 0.131751, "relative": 0.824106 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.091206, "mean_abs": 0.082087, "relative": 0.748411 }, { "name": "caption_objects_interaction_text", "l2": 1.820931, "mean_abs": 0.01304, "relative": 0.836127 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 190, "start_frame": 950, "end_frame": 969, "center_frame": 959, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.235075, "mean_abs": 0.262806, "relative": 0.947726 }, { "name": "hand_right_joints", "l2": 11.492339, "mean_abs": 0.287024, "relative": 0.905904 }, { "name": "body_joints", "l2": 6.080519, "mean_abs": 0.085667, "relative": 0.821291 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.026168, "mean_abs": 0.004042, "relative": 0.061754 }, { "name": "camera_rotation_matrix", "l2": 3.000768, "mean_abs": 0.172152, "relative": 0.921487 }, { "name": "imu_accel_gyro", "l2": 17.040091, "mean_abs": 1.084278, "relative": 0.919372 }, { "name": "depth_confidence", "l2": 9.31276, "mean_abs": 0.174461, "relative": 0.000309 }, { "name": "video_fisheye_cam0", "l2": 5.162695, "mean_abs": 0.10809, "relative": 0.934854 }, { "name": "video_fisheye_cam1", "l2": 5.24447, "mean_abs": 0.110115, "relative": 0.866604 }, { "name": "video_fisheye_cam2", "l2": 5.502503, "mean_abs": 0.112272, "relative": 0.878393 }, { "name": "video_fisheye_cam3", "l2": 5.479841, "mean_abs": 0.110985, "relative": 0.917533 }, { "name": "video_stereo_left", "l2": 5.065883, "mean_abs": 0.111543, "relative": 0.823216 }, { "name": "video_stereo_right", "l2": 5.872999, "mean_abs": 0.132673, "relative": 0.82797 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.007277, "mean_abs": 0.076847, "relative": 0.718374 }, { "name": "caption_objects_interaction_text", "l2": 1.820931, "mean_abs": 0.013141, "relative": 0.836127 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 191, "start_frame": 955, "end_frame": 974, "center_frame": 964, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.244565, "mean_abs": 0.265176, "relative": 0.948527 }, { "name": "hand_right_joints", "l2": 11.484858, "mean_abs": 0.289194, "relative": 0.905314 }, { "name": "body_joints", "l2": 6.096095, "mean_abs": 0.086989, "relative": 0.823395 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020002, "mean_abs": 0.002846, "relative": 0.047202 }, { "name": "camera_rotation_matrix", "l2": 3.000576, "mean_abs": 0.171271, "relative": 0.921428 }, { "name": "imu_accel_gyro", "l2": 17.068344, "mean_abs": 1.074841, "relative": 0.920896 }, { "name": "depth_confidence", "l2": 9.273561, "mean_abs": 0.173748, "relative": 0.000308 }, { "name": "video_fisheye_cam0", "l2": 5.174937, "mean_abs": 0.108515, "relative": 0.93707 }, { "name": "video_fisheye_cam1", "l2": 5.252698, "mean_abs": 0.110139, "relative": 0.867964 }, { "name": "video_fisheye_cam2", "l2": 5.526255, "mean_abs": 0.11236, "relative": 0.882184 }, { "name": "video_fisheye_cam3", "l2": 5.491213, "mean_abs": 0.11097, "relative": 0.919437 }, { "name": "video_stereo_left", "l2": 5.092169, "mean_abs": 0.111959, "relative": 0.827488 }, { "name": "video_stereo_right", "l2": 5.906141, "mean_abs": 0.133309, "relative": 0.832642 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.04562, "mean_abs": 0.078676, "relative": 0.732097 }, { "name": "caption_objects_interaction_text", "l2": 1.820931, "mean_abs": 0.01304, "relative": 0.836127 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 192, "start_frame": 960, "end_frame": 979, "center_frame": 969, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.246371, "mean_abs": 0.267566, "relative": 0.948679 }, { "name": "hand_right_joints", "l2": 11.440726, "mean_abs": 0.291801, "relative": 0.901835 }, { "name": "body_joints", "l2": 6.124323, "mean_abs": 0.088359, "relative": 0.827207 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017964, "mean_abs": 0.002484, "relative": 0.042392 }, { "name": "camera_rotation_matrix", "l2": 3.000926, "mean_abs": 0.170993, "relative": 0.921536 }, { "name": "imu_accel_gyro", "l2": 17.102083, "mean_abs": 1.094866, "relative": 0.922716 }, { "name": "depth_confidence", "l2": 9.213196, "mean_abs": 0.172052, "relative": 0.000306 }, { "name": "video_fisheye_cam0", "l2": 5.180343, "mean_abs": 0.108989, "relative": 0.938049 }, { "name": "video_fisheye_cam1", "l2": 5.258971, "mean_abs": 0.110134, "relative": 0.869001 }, { "name": "video_fisheye_cam2", "l2": 5.537174, "mean_abs": 0.112638, "relative": 0.883927 }, { "name": "video_fisheye_cam3", "l2": 5.504742, "mean_abs": 0.111023, "relative": 0.921703 }, { "name": "video_stereo_left", "l2": 5.119426, "mean_abs": 0.112493, "relative": 0.831917 }, { "name": "video_stereo_right", "l2": 5.944635, "mean_abs": 0.134001, "relative": 0.838069 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.092734, "mean_abs": 0.08051, "relative": 0.748958 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 193, "start_frame": 965, "end_frame": 984, "center_frame": 974, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.211255, "mean_abs": 0.269159, "relative": 0.945717 }, { "name": "hand_right_joints", "l2": 11.36311, "mean_abs": 0.291882, "relative": 0.895717 }, { "name": "body_joints", "l2": 6.173728, "mean_abs": 0.090217, "relative": 0.83388 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013165, "mean_abs": 0.001992, "relative": 0.031068 }, { "name": "camera_rotation_matrix", "l2": 3.000344, "mean_abs": 0.169957, "relative": 0.921357 }, { "name": "imu_accel_gyro", "l2": 17.095604, "mean_abs": 1.083454, "relative": 0.922367 }, { "name": "depth_confidence", "l2": 9.192564, "mean_abs": 0.170951, "relative": 0.000305 }, { "name": "video_fisheye_cam0", "l2": 5.179431, "mean_abs": 0.108719, "relative": 0.937884 }, { "name": "video_fisheye_cam1", "l2": 5.253865, "mean_abs": 0.109818, "relative": 0.868157 }, { "name": "video_fisheye_cam2", "l2": 5.544309, "mean_abs": 0.112492, "relative": 0.885066 }, { "name": "video_fisheye_cam3", "l2": 5.505312, "mean_abs": 0.110801, "relative": 0.921798 }, { "name": "video_stereo_left", "l2": 5.132506, "mean_abs": 0.111879, "relative": 0.834042 }, { "name": "video_stereo_right", "l2": 5.957757, "mean_abs": 0.133387, "relative": 0.839919 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.064415, "mean_abs": 0.078255, "relative": 0.738823 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 194, "start_frame": 970, "end_frame": 989, "center_frame": 979, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.152455, "mean_abs": 0.269854, "relative": 0.940757 }, { "name": "hand_right_joints", "l2": 11.297665, "mean_abs": 0.291638, "relative": 0.890558 }, { "name": "body_joints", "l2": 6.22356, "mean_abs": 0.090792, "relative": 0.840611 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.0119, "mean_abs": 0.001841, "relative": 0.028082 }, { "name": "camera_rotation_matrix", "l2": 3.000072, "mean_abs": 0.169436, "relative": 0.921273 }, { "name": "imu_accel_gyro", "l2": 17.048927, "mean_abs": 1.084966, "relative": 0.919848 }, { "name": "depth_confidence", "l2": 9.143683, "mean_abs": 0.169902, "relative": 0.000303 }, { "name": "video_fisheye_cam0", "l2": 5.163229, "mean_abs": 0.108055, "relative": 0.93495 }, { "name": "video_fisheye_cam1", "l2": 5.251939, "mean_abs": 0.109722, "relative": 0.867839 }, { "name": "video_fisheye_cam2", "l2": 5.549383, "mean_abs": 0.112388, "relative": 0.885876 }, { "name": "video_fisheye_cam3", "l2": 5.4885, "mean_abs": 0.110352, "relative": 0.918983 }, { "name": "video_stereo_left", "l2": 5.143311, "mean_abs": 0.111548, "relative": 0.835798 }, { "name": "video_stereo_right", "l2": 5.984402, "mean_abs": 0.133722, "relative": 0.843675 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.109483, "mean_abs": 0.081044, "relative": 0.754952 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 195, "start_frame": 975, "end_frame": 994, "center_frame": 984, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.127152, "mean_abs": 0.269634, "relative": 0.938623 }, { "name": "hand_right_joints", "l2": 11.244972, "mean_abs": 0.290447, "relative": 0.886405 }, { "name": "body_joints", "l2": 6.270137, "mean_abs": 0.090758, "relative": 0.846902 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01045, "mean_abs": 0.001739, "relative": 0.02466 }, { "name": "camera_rotation_matrix", "l2": 2.999875, "mean_abs": 0.168967, "relative": 0.921213 }, { "name": "imu_accel_gyro", "l2": 17.067472, "mean_abs": 1.089436, "relative": 0.920849 }, { "name": "depth_confidence", "l2": 9.138516, "mean_abs": 0.169448, "relative": 0.000303 }, { "name": "video_fisheye_cam0", "l2": 5.159714, "mean_abs": 0.108079, "relative": 0.934314 }, { "name": "video_fisheye_cam1", "l2": 5.257582, "mean_abs": 0.109731, "relative": 0.868771 }, { "name": "video_fisheye_cam2", "l2": 5.556021, "mean_abs": 0.112256, "relative": 0.886936 }, { "name": "video_fisheye_cam3", "l2": 5.479255, "mean_abs": 0.110177, "relative": 0.917435 }, { "name": "video_stereo_left", "l2": 5.155553, "mean_abs": 0.11113, "relative": 0.837788 }, { "name": "video_stereo_right", "l2": 6.012753, "mean_abs": 0.133533, "relative": 0.847672 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.121672, "mean_abs": 0.081533, "relative": 0.759315 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 196, "start_frame": 980, "end_frame": 999, "center_frame": 989, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.077174, "mean_abs": 0.268447, "relative": 0.934407 }, { "name": "hand_right_joints", "l2": 11.188031, "mean_abs": 0.289515, "relative": 0.881916 }, { "name": "body_joints", "l2": 6.312119, "mean_abs": 0.089789, "relative": 0.852573 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013432, "mean_abs": 0.002061, "relative": 0.031697 }, { "name": "camera_rotation_matrix", "l2": 3.000048, "mean_abs": 0.169725, "relative": 0.921266 }, { "name": "imu_accel_gyro", "l2": 17.045786, "mean_abs": 1.085834, "relative": 0.919679 }, { "name": "depth_confidence", "l2": 9.178305, "mean_abs": 0.170116, "relative": 0.000305 }, { "name": "video_fisheye_cam0", "l2": 5.138434, "mean_abs": 0.107331, "relative": 0.930461 }, { "name": "video_fisheye_cam1", "l2": 5.267233, "mean_abs": 0.11021, "relative": 0.870366 }, { "name": "video_fisheye_cam2", "l2": 5.560324, "mean_abs": 0.112597, "relative": 0.887623 }, { "name": "video_fisheye_cam3", "l2": 5.50073, "mean_abs": 0.110463, "relative": 0.921031 }, { "name": "video_stereo_left", "l2": 5.165545, "mean_abs": 0.111271, "relative": 0.839411 }, { "name": "video_stereo_right", "l2": 6.040418, "mean_abs": 0.134222, "relative": 0.851572 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.095528, "mean_abs": 0.08354, "relative": 0.749958 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010482, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 197, "start_frame": 985, "end_frame": 1004, "center_frame": 994, "action": "Move kettle", "subtask": "Move kettle", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table", "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.024748, "mean_abs": 0.265894, "relative": 0.929985 }, { "name": "hand_right_joints", "l2": 11.086791, "mean_abs": 0.284546, "relative": 0.873936 }, { "name": "body_joints", "l2": 6.382304, "mean_abs": 0.089129, "relative": 0.862053 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081671, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030346, "mean_abs": 0.004458, "relative": 0.071612 }, { "name": "camera_rotation_matrix", "l2": 3.000415, "mean_abs": 0.172244, "relative": 0.921379 }, { "name": "imu_accel_gyro", "l2": 17.092505, "mean_abs": 1.096946, "relative": 0.922199 }, { "name": "depth_confidence", "l2": 9.213312, "mean_abs": 0.171833, "relative": 0.000306 }, { "name": "video_fisheye_cam0", "l2": 5.131622, "mean_abs": 0.10751, "relative": 0.929227 }, { "name": "video_fisheye_cam1", "l2": 5.285172, "mean_abs": 0.111205, "relative": 0.87333 }, { "name": "video_fisheye_cam2", "l2": 5.573421, "mean_abs": 0.113565, "relative": 0.889714 }, { "name": "video_fisheye_cam3", "l2": 5.516993, "mean_abs": 0.111311, "relative": 0.923754 }, { "name": "video_stereo_left", "l2": 5.174448, "mean_abs": 0.112449, "relative": 0.840858 }, { "name": "video_stereo_right", "l2": 6.06264, "mean_abs": 0.136227, "relative": 0.854705 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.107184, "mean_abs": 0.082443, "relative": 0.75413 }, { "name": "caption_objects_interaction_text", "l2": 2.177815, "mean_abs": 0.019727, "relative": 1.0 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 198, "start_frame": 990, "end_frame": 1009, "center_frame": 999, "action": "", "subtask": "", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table", "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.909909, "mean_abs": 0.263446, "relative": 0.920297 }, { "name": "hand_right_joints", "l2": 10.993065, "mean_abs": 0.284381, "relative": 0.866548 }, { "name": "body_joints", "l2": 6.443512, "mean_abs": 0.090244, "relative": 0.87032 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.061887, "mean_abs": 0.009087, "relative": 0.146045 }, { "name": "camera_rotation_matrix", "l2": 3.001684, "mean_abs": 0.17643, "relative": 0.921768 }, { "name": "imu_accel_gyro", "l2": 17.091572, "mean_abs": 1.107507, "relative": 0.922149 }, { "name": "depth_confidence", "l2": 9.299604, "mean_abs": 0.175323, "relative": 0.000309 }, { "name": "video_fisheye_cam0", "l2": 5.121994, "mean_abs": 0.108114, "relative": 0.927484 }, { "name": "video_fisheye_cam1", "l2": 5.321418, "mean_abs": 0.112596, "relative": 0.879319 }, { "name": "video_fisheye_cam2", "l2": 5.589468, "mean_abs": 0.114659, "relative": 0.892275 }, { "name": "video_fisheye_cam3", "l2": 5.550137, "mean_abs": 0.112528, "relative": 0.929304 }, { "name": "video_stereo_left", "l2": 5.18291, "mean_abs": 0.114184, "relative": 0.842233 }, { "name": "video_stereo_right", "l2": 6.11108, "mean_abs": 0.139063, "relative": 0.861534 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.997721, "mean_abs": 0.076538, "relative": 0.714954 }, { "name": "caption_objects_interaction_text", "l2": 2.177815, "mean_abs": 0.020017, "relative": 1.0 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 199, "start_frame": 995, "end_frame": 1014, "center_frame": 1004, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "kettle", "dripper", "scale", "coffee filter", "table", "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.840755, "mean_abs": 0.261736, "relative": 0.914464 }, { "name": "hand_right_joints", "l2": 10.863957, "mean_abs": 0.281805, "relative": 0.85637 }, { "name": "body_joints", "l2": 6.487967, "mean_abs": 0.091388, "relative": 0.876324 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.10435, "mean_abs": 0.015313, "relative": 0.24625 }, { "name": "camera_rotation_matrix", "l2": 3.004817, "mean_abs": 0.183188, "relative": 0.922731 }, { "name": "imu_accel_gyro", "l2": 17.099575, "mean_abs": 1.129642, "relative": 0.922581 }, { "name": "depth_confidence", "l2": 9.417048, "mean_abs": 0.179107, "relative": 0.000313 }, { "name": "video_fisheye_cam0", "l2": 5.118521, "mean_abs": 0.108953, "relative": 0.926855 }, { "name": "video_fisheye_cam1", "l2": 5.352282, "mean_abs": 0.11465, "relative": 0.884419 }, { "name": "video_fisheye_cam2", "l2": 5.624209, "mean_abs": 0.116899, "relative": 0.897821 }, { "name": "video_fisheye_cam3", "l2": 5.589641, "mean_abs": 0.113864, "relative": 0.935918 }, { "name": "video_stereo_left", "l2": 5.210231, "mean_abs": 0.116988, "relative": 0.846673 }, { "name": "video_stereo_right", "l2": 6.16966, "mean_abs": 0.142266, "relative": 0.869793 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.956733, "mean_abs": 0.076577, "relative": 0.700285 }, { "name": "caption_objects_interaction_text", "l2": 2.177815, "mean_abs": 0.019663, "relative": 1.0 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 200, "start_frame": 1000, "end_frame": 1019, "center_frame": 1009, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.769513, "mean_abs": 0.257446, "relative": 0.908454 }, { "name": "hand_right_joints", "l2": 10.785174, "mean_abs": 0.278931, "relative": 0.85016 }, { "name": "body_joints", "l2": 6.510791, "mean_abs": 0.09252, "relative": 0.879407 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.157141, "mean_abs": 0.022734, "relative": 0.370831 }, { "name": "camera_rotation_matrix", "l2": 3.011064, "mean_abs": 0.191698, "relative": 0.924649 }, { "name": "imu_accel_gyro", "l2": 17.085438, "mean_abs": 1.153129, "relative": 0.921818 }, { "name": "depth_confidence", "l2": 9.624547, "mean_abs": 0.185071, "relative": 0.000319 }, { "name": "video_fisheye_cam0", "l2": 5.131581, "mean_abs": 0.110155, "relative": 0.92922 }, { "name": "video_fisheye_cam1", "l2": 5.394511, "mean_abs": 0.118025, "relative": 0.891397 }, { "name": "video_fisheye_cam2", "l2": 5.682678, "mean_abs": 0.119477, "relative": 0.907155 }, { "name": "video_fisheye_cam3", "l2": 5.613706, "mean_abs": 0.114852, "relative": 0.939947 }, { "name": "video_stereo_left", "l2": 5.279116, "mean_abs": 0.120429, "relative": 0.857867 }, { "name": "video_stereo_right", "l2": 6.217035, "mean_abs": 0.145058, "relative": 0.876471 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.070402, "mean_abs": 0.080058, "relative": 0.740966 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 201, "start_frame": 1005, "end_frame": 1024, "center_frame": 1014, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.689535, "mean_abs": 0.252536, "relative": 0.901708 }, { "name": "hand_right_joints", "l2": 10.705958, "mean_abs": 0.274282, "relative": 0.843916 }, { "name": "body_joints", "l2": 6.526864, "mean_abs": 0.093804, "relative": 0.881578 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.198204, "mean_abs": 0.027576, "relative": 0.467734 }, { "name": "camera_rotation_matrix", "l2": 3.016456, "mean_abs": 0.199057, "relative": 0.926305 }, { "name": "imu_accel_gyro", "l2": 17.089037, "mean_abs": 1.146955, "relative": 0.922012 }, { "name": "depth_confidence", "l2": 9.822595, "mean_abs": 0.189752, "relative": 0.000326 }, { "name": "video_fisheye_cam0", "l2": 5.133359, "mean_abs": 0.110189, "relative": 0.929542 }, { "name": "video_fisheye_cam1", "l2": 5.440187, "mean_abs": 0.120722, "relative": 0.898945 }, { "name": "video_fisheye_cam2", "l2": 5.746737, "mean_abs": 0.121377, "relative": 0.917381 }, { "name": "video_fisheye_cam3", "l2": 5.639814, "mean_abs": 0.115936, "relative": 0.944319 }, { "name": "video_stereo_left", "l2": 5.345253, "mean_abs": 0.123123, "relative": 0.868614 }, { "name": "video_stereo_right", "l2": 6.28292, "mean_abs": 0.148152, "relative": 0.88576 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.073503, "mean_abs": 0.080604, "relative": 0.742076 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 202, "start_frame": 1010, "end_frame": 1029, "center_frame": 1019, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.718187, "mean_abs": 0.251318, "relative": 0.904125 }, { "name": "hand_right_joints", "l2": 10.734783, "mean_abs": 0.275773, "relative": 0.846188 }, { "name": "body_joints", "l2": 6.518675, "mean_abs": 0.09544, "relative": 0.880472 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.226208, "mean_abs": 0.029163, "relative": 0.533818 }, { "name": "camera_rotation_matrix", "l2": 3.019854, "mean_abs": 0.203957, "relative": 0.927348 }, { "name": "imu_accel_gyro", "l2": 17.052008, "mean_abs": 1.123601, "relative": 0.920015 }, { "name": "depth_confidence", "l2": 9.960506, "mean_abs": 0.192319, "relative": 0.000331 }, { "name": "video_fisheye_cam0", "l2": 5.130682, "mean_abs": 0.109271, "relative": 0.929057 }, { "name": "video_fisheye_cam1", "l2": 5.484622, "mean_abs": 0.122272, "relative": 0.906287 }, { "name": "video_fisheye_cam2", "l2": 5.797299, "mean_abs": 0.122501, "relative": 0.925453 }, { "name": "video_fisheye_cam3", "l2": 5.698115, "mean_abs": 0.117953, "relative": 0.954081 }, { "name": "video_stereo_left", "l2": 5.384218, "mean_abs": 0.124656, "relative": 0.874946 }, { "name": "video_stereo_right", "l2": 6.38269, "mean_abs": 0.150763, "relative": 0.899825 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.085997, "mean_abs": 0.081842, "relative": 0.746547 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 203, "start_frame": 1015, "end_frame": 1034, "center_frame": 1024, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.766933, "mean_abs": 0.250446, "relative": 0.908237 }, { "name": "hand_right_joints", "l2": 10.808187, "mean_abs": 0.275144, "relative": 0.851974 }, { "name": "body_joints", "l2": 6.537272, "mean_abs": 0.097671, "relative": 0.882984 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.224703, "mean_abs": 0.025735, "relative": 0.530267 }, { "name": "camera_rotation_matrix", "l2": 3.013313, "mean_abs": 0.20343, "relative": 0.92534 }, { "name": "imu_accel_gyro", "l2": 16.974199, "mean_abs": 1.094077, "relative": 0.915816 }, { "name": "depth_confidence", "l2": 10.039503, "mean_abs": 0.19149, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.116484, "mean_abs": 0.107885, "relative": 0.926486 }, { "name": "video_fisheye_cam1", "l2": 5.483947, "mean_abs": 0.12126, "relative": 0.906176 }, { "name": "video_fisheye_cam2", "l2": 5.803147, "mean_abs": 0.122052, "relative": 0.926386 }, { "name": "video_fisheye_cam3", "l2": 5.697606, "mean_abs": 0.118991, "relative": 0.953995 }, { "name": "video_stereo_left", "l2": 5.387697, "mean_abs": 0.123394, "relative": 0.875512 }, { "name": "video_stereo_right", "l2": 6.407177, "mean_abs": 0.150379, "relative": 0.903278 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.164112, "mean_abs": 0.083182, "relative": 0.774503 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 204, "start_frame": 1020, "end_frame": 1039, "center_frame": 1029, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.821261, "mean_abs": 0.247487, "relative": 0.91282 }, { "name": "hand_right_joints", "l2": 10.864, "mean_abs": 0.271157, "relative": 0.856374 }, { "name": "body_joints", "l2": 6.544496, "mean_abs": 0.099449, "relative": 0.88396 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.20985, "mean_abs": 0.027631, "relative": 0.495216 }, { "name": "camera_rotation_matrix", "l2": 3.004291, "mean_abs": 0.200917, "relative": 0.922569 }, { "name": "imu_accel_gyro", "l2": 17.00362, "mean_abs": 1.088329, "relative": 0.917404 }, { "name": "depth_confidence", "l2": 10.031879, "mean_abs": 0.191107, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.119227, "mean_abs": 0.106921, "relative": 0.926983 }, { "name": "video_fisheye_cam1", "l2": 5.45111, "mean_abs": 0.11912, "relative": 0.90075 }, { "name": "video_fisheye_cam2", "l2": 5.810898, "mean_abs": 0.120762, "relative": 0.927623 }, { "name": "video_fisheye_cam3", "l2": 5.729033, "mean_abs": 0.120434, "relative": 0.959258 }, { "name": "video_stereo_left", "l2": 5.401588, "mean_abs": 0.121363, "relative": 0.877769 }, { "name": "video_stereo_right", "l2": 6.378794, "mean_abs": 0.14622, "relative": 0.899276 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.021945, "mean_abs": 0.078723, "relative": 0.723624 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 205, "start_frame": 1025, "end_frame": 1044, "center_frame": 1034, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.90982, "mean_abs": 0.244696, "relative": 0.92029 }, { "name": "hand_right_joints", "l2": 10.939257, "mean_abs": 0.267454, "relative": 0.862306 }, { "name": "body_joints", "l2": 6.551918, "mean_abs": 0.100305, "relative": 0.884962 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.199554, "mean_abs": 0.03057, "relative": 0.470919 }, { "name": "camera_rotation_matrix", "l2": 2.999914, "mean_abs": 0.198052, "relative": 0.921225 }, { "name": "imu_accel_gyro", "l2": 17.085733, "mean_abs": 1.090628, "relative": 0.921834 }, { "name": "depth_confidence", "l2": 9.907502, "mean_abs": 0.18962, "relative": 0.000329 }, { "name": "video_fisheye_cam0", "l2": 5.127072, "mean_abs": 0.106803, "relative": 0.928403 }, { "name": "video_fisheye_cam1", "l2": 5.424919, "mean_abs": 0.117287, "relative": 0.896422 }, { "name": "video_fisheye_cam2", "l2": 5.801949, "mean_abs": 0.119456, "relative": 0.926195 }, { "name": "video_fisheye_cam3", "l2": 5.746819, "mean_abs": 0.121255, "relative": 0.962236 }, { "name": "video_stereo_left", "l2": 5.404251, "mean_abs": 0.120718, "relative": 0.878201 }, { "name": "video_stereo_right", "l2": 6.397783, "mean_abs": 0.144318, "relative": 0.901953 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.953065, "mean_abs": 0.073726, "relative": 0.698972 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 206, "start_frame": 1030, "end_frame": 1049, "center_frame": 1039, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.063375, "mean_abs": 0.246672, "relative": 0.933243 }, { "name": "hand_right_joints", "l2": 11.064495, "mean_abs": 0.267177, "relative": 0.872178 }, { "name": "body_joints", "l2": 6.608679, "mean_abs": 0.101679, "relative": 0.892629 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.182087, "mean_abs": 0.028401, "relative": 0.4297 }, { "name": "camera_rotation_matrix", "l2": 3.004413, "mean_abs": 0.196577, "relative": 0.922607 }, { "name": "imu_accel_gyro", "l2": 17.147289, "mean_abs": 1.08512, "relative": 0.925155 }, { "name": "depth_confidence", "l2": 9.801542, "mean_abs": 0.187978, "relative": 0.000325 }, { "name": "video_fisheye_cam0", "l2": 5.127644, "mean_abs": 0.106648, "relative": 0.928507 }, { "name": "video_fisheye_cam1", "l2": 5.384873, "mean_abs": 0.115921, "relative": 0.889805 }, { "name": "video_fisheye_cam2", "l2": 5.772095, "mean_abs": 0.118711, "relative": 0.921429 }, { "name": "video_fisheye_cam3", "l2": 5.694597, "mean_abs": 0.120771, "relative": 0.953492 }, { "name": "video_stereo_left", "l2": 5.383705, "mean_abs": 0.120217, "relative": 0.874863 }, { "name": "video_stereo_right", "l2": 6.395146, "mean_abs": 0.145014, "relative": 0.901581 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.029205, "mean_abs": 0.080938, "relative": 0.726222 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 207, "start_frame": 1035, "end_frame": 1054, "center_frame": 1044, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.152738, "mean_abs": 0.250131, "relative": 0.940781 }, { "name": "hand_right_joints", "l2": 11.138975, "mean_abs": 0.26928, "relative": 0.878049 }, { "name": "body_joints", "l2": 6.625499, "mean_abs": 0.101918, "relative": 0.894901 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.132768, "mean_abs": 0.020208, "relative": 0.313314 }, { "name": "camera_rotation_matrix", "l2": 3.002945, "mean_abs": 0.193964, "relative": 0.922156 }, { "name": "imu_accel_gyro", "l2": 17.192686, "mean_abs": 1.086288, "relative": 0.927605 }, { "name": "depth_confidence", "l2": 9.65232, "mean_abs": 0.185286, "relative": 0.00032 }, { "name": "video_fisheye_cam0", "l2": 5.13025, "mean_abs": 0.106502, "relative": 0.928978 }, { "name": "video_fisheye_cam1", "l2": 5.345206, "mean_abs": 0.114678, "relative": 0.88325 }, { "name": "video_fisheye_cam2", "l2": 5.746387, "mean_abs": 0.11772, "relative": 0.917325 }, { "name": "video_fisheye_cam3", "l2": 5.657827, "mean_abs": 0.119666, "relative": 0.947335 }, { "name": "video_stereo_left", "l2": 5.356618, "mean_abs": 0.118597, "relative": 0.870461 }, { "name": "video_stereo_right", "l2": 6.35043, "mean_abs": 0.143687, "relative": 0.895277 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.970473, "mean_abs": 0.078375, "relative": 0.705203 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 208, "start_frame": 1040, "end_frame": 1059, "center_frame": 1049, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.203609, "mean_abs": 0.250015, "relative": 0.945072 }, { "name": "hand_right_joints", "l2": 11.169306, "mean_abs": 0.270418, "relative": 0.88044 }, { "name": "body_joints", "l2": 6.612325, "mean_abs": 0.101918, "relative": 0.893121 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.074375, "mean_abs": 0.011028, "relative": 0.175515 }, { "name": "camera_rotation_matrix", "l2": 3.001396, "mean_abs": 0.1896, "relative": 0.92168 }, { "name": "imu_accel_gyro", "l2": 17.162933, "mean_abs": 1.076103, "relative": 0.925999 }, { "name": "depth_confidence", "l2": 9.514817, "mean_abs": 0.180667, "relative": 0.000316 }, { "name": "video_fisheye_cam0", "l2": 5.126967, "mean_abs": 0.106, "relative": 0.928384 }, { "name": "video_fisheye_cam1", "l2": 5.293226, "mean_abs": 0.11271, "relative": 0.874661 }, { "name": "video_fisheye_cam2", "l2": 5.707245, "mean_abs": 0.115696, "relative": 0.911077 }, { "name": "video_fisheye_cam3", "l2": 5.613481, "mean_abs": 0.117265, "relative": 0.93991 }, { "name": "video_stereo_left", "l2": 5.309317, "mean_abs": 0.115453, "relative": 0.862775 }, { "name": "video_stereo_right", "l2": 6.278772, "mean_abs": 0.140441, "relative": 0.885175 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.950415, "mean_abs": 0.075961, "relative": 0.698024 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 209, "start_frame": 1045, "end_frame": 1064, "center_frame": 1054, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.253433, "mean_abs": 0.249202, "relative": 0.949275 }, { "name": "hand_right_joints", "l2": 11.178959, "mean_abs": 0.269916, "relative": 0.881201 }, { "name": "body_joints", "l2": 6.60038, "mean_abs": 0.101721, "relative": 0.891508 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.031175, "mean_abs": 0.004722, "relative": 0.073569 }, { "name": "camera_rotation_matrix", "l2": 3.001222, "mean_abs": 0.186618, "relative": 0.921626 }, { "name": "imu_accel_gyro", "l2": 17.180494, "mean_abs": 1.06819, "relative": 0.926947 }, { "name": "depth_confidence", "l2": 9.450134, "mean_abs": 0.177968, "relative": 0.000314 }, { "name": "video_fisheye_cam0", "l2": 5.126235, "mean_abs": 0.105845, "relative": 0.928252 }, { "name": "video_fisheye_cam1", "l2": 5.269691, "mean_abs": 0.111641, "relative": 0.870772 }, { "name": "video_fisheye_cam2", "l2": 5.692213, "mean_abs": 0.1149, "relative": 0.908677 }, { "name": "video_fisheye_cam3", "l2": 5.601194, "mean_abs": 0.115506, "relative": 0.937852 }, { "name": "video_stereo_left", "l2": 5.308441, "mean_abs": 0.114403, "relative": 0.862632 }, { "name": "video_stereo_right", "l2": 6.262606, "mean_abs": 0.1387, "relative": 0.882896 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.987884, "mean_abs": 0.078304, "relative": 0.711434 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 210, "start_frame": 1050, "end_frame": 1069, "center_frame": 1059, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.231986, "mean_abs": 0.248543, "relative": 0.947466 }, { "name": "hand_right_joints", "l2": 11.211972, "mean_abs": 0.27413, "relative": 0.883803 }, { "name": "body_joints", "l2": 6.578078, "mean_abs": 0.101194, "relative": 0.888496 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020653, "mean_abs": 0.002966, "relative": 0.048739 }, { "name": "camera_rotation_matrix", "l2": 3.000637, "mean_abs": 0.185696, "relative": 0.921447 }, { "name": "imu_accel_gyro", "l2": 17.159815, "mean_abs": 1.058979, "relative": 0.925831 }, { "name": "depth_confidence", "l2": 9.420632, "mean_abs": 0.176555, "relative": 0.000313 }, { "name": "video_fisheye_cam0", "l2": 5.121995, "mean_abs": 0.105814, "relative": 0.927484 }, { "name": "video_fisheye_cam1", "l2": 5.256218, "mean_abs": 0.111168, "relative": 0.868546 }, { "name": "video_fisheye_cam2", "l2": 5.701567, "mean_abs": 0.11492, "relative": 0.91017 }, { "name": "video_fisheye_cam3", "l2": 5.533643, "mean_abs": 0.114035, "relative": 0.926542 }, { "name": "video_stereo_left", "l2": 5.332792, "mean_abs": 0.114305, "relative": 0.866589 }, { "name": "video_stereo_right", "l2": 6.253373, "mean_abs": 0.138221, "relative": 0.881594 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.976583, "mean_abs": 0.078581, "relative": 0.707389 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 211, "start_frame": 1055, "end_frame": 1074, "center_frame": 1064, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.080716, "mean_abs": 0.249552, "relative": 0.934706 }, { "name": "hand_right_joints", "l2": 11.174219, "mean_abs": 0.276476, "relative": 0.880827 }, { "name": "body_joints", "l2": 6.490018, "mean_abs": 0.101142, "relative": 0.876602 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.050661, "mean_abs": 0.007012, "relative": 0.119553 }, { "name": "camera_rotation_matrix", "l2": 3.000704, "mean_abs": 0.185502, "relative": 0.921467 }, { "name": "imu_accel_gyro", "l2": 17.093515, "mean_abs": 1.057043, "relative": 0.922254 }, { "name": "depth_confidence", "l2": 9.442374, "mean_abs": 0.177256, "relative": 0.000313 }, { "name": "video_fisheye_cam0", "l2": 5.118913, "mean_abs": 0.106042, "relative": 0.926926 }, { "name": "video_fisheye_cam1", "l2": 5.257329, "mean_abs": 0.111863, "relative": 0.868729 }, { "name": "video_fisheye_cam2", "l2": 5.713684, "mean_abs": 0.115526, "relative": 0.912105 }, { "name": "video_fisheye_cam3", "l2": 5.529419, "mean_abs": 0.114107, "relative": 0.925835 }, { "name": "video_stereo_left", "l2": 5.359402, "mean_abs": 0.115517, "relative": 0.870913 }, { "name": "video_stereo_right", "l2": 6.257002, "mean_abs": 0.140004, "relative": 0.882106 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.014789, "mean_abs": 0.07884, "relative": 0.721063 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 212, "start_frame": 1060, "end_frame": 1079, "center_frame": 1069, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.973539, "mean_abs": 0.252558, "relative": 0.925665 }, { "name": "hand_right_joints", "l2": 11.404532, "mean_abs": 0.284475, "relative": 0.898982 }, { "name": "body_joints", "l2": 6.319231, "mean_abs": 0.1032, "relative": 0.853533 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.074224, "mean_abs": 0.01026, "relative": 0.175159 }, { "name": "camera_rotation_matrix", "l2": 3.001074, "mean_abs": 0.186409, "relative": 0.921581 }, { "name": "imu_accel_gyro", "l2": 17.090082, "mean_abs": 1.078051, "relative": 0.922069 }, { "name": "depth_confidence", "l2": 9.437522, "mean_abs": 0.179377, "relative": 0.000313 }, { "name": "video_fisheye_cam0", "l2": 5.116608, "mean_abs": 0.106419, "relative": 0.926508 }, { "name": "video_fisheye_cam1", "l2": 5.277419, "mean_abs": 0.113023, "relative": 0.872049 }, { "name": "video_fisheye_cam2", "l2": 5.740511, "mean_abs": 0.116629, "relative": 0.916387 }, { "name": "video_fisheye_cam3", "l2": 5.535168, "mean_abs": 0.114887, "relative": 0.926797 }, { "name": "video_stereo_left", "l2": 5.417417, "mean_abs": 0.118591, "relative": 0.880341 }, { "name": "video_stereo_right", "l2": 6.323041, "mean_abs": 0.144312, "relative": 0.891416 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.174227, "mean_abs": 0.084323, "relative": 0.778123 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 213, "start_frame": 1065, "end_frame": 1084, "center_frame": 1074, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.908114, "mean_abs": 0.258088, "relative": 0.920146 }, { "name": "hand_right_joints", "l2": 11.598902, "mean_abs": 0.292657, "relative": 0.914304 }, { "name": "body_joints", "l2": 6.243013, "mean_abs": 0.104631, "relative": 0.843239 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.084444, "mean_abs": 0.011424, "relative": 0.199276 }, { "name": "camera_rotation_matrix", "l2": 3.00123, "mean_abs": 0.186615, "relative": 0.921629 }, { "name": "imu_accel_gyro", "l2": 17.09688, "mean_abs": 1.100473, "relative": 0.922436 }, { "name": "depth_confidence", "l2": 9.437357, "mean_abs": 0.180282, "relative": 0.000313 }, { "name": "video_fisheye_cam0", "l2": 5.109188, "mean_abs": 0.106333, "relative": 0.925165 }, { "name": "video_fisheye_cam1", "l2": 5.303685, "mean_abs": 0.113947, "relative": 0.876389 }, { "name": "video_fisheye_cam2", "l2": 5.76544, "mean_abs": 0.117455, "relative": 0.920367 }, { "name": "video_fisheye_cam3", "l2": 5.50663, "mean_abs": 0.11418, "relative": 0.922019 }, { "name": "video_stereo_left", "l2": 5.472381, "mean_abs": 0.120763, "relative": 0.889273 }, { "name": "video_stereo_right", "l2": 6.383661, "mean_abs": 0.146476, "relative": 0.899962 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.086191, "mean_abs": 0.081402, "relative": 0.746616 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 214, "start_frame": 1070, "end_frame": 1089, "center_frame": 1079, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.802157, "mean_abs": 0.256227, "relative": 0.911208 }, { "name": "hand_right_joints", "l2": 11.744157, "mean_abs": 0.297056, "relative": 0.925754 }, { "name": "body_joints", "l2": 6.162855, "mean_abs": 0.103645, "relative": 0.832412 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.072294, "mean_abs": 0.009969, "relative": 0.170604 }, { "name": "camera_rotation_matrix", "l2": 3.000494, "mean_abs": 0.184739, "relative": 0.921403 }, { "name": "imu_accel_gyro", "l2": 16.998051, "mean_abs": 1.09774, "relative": 0.917103 }, { "name": "depth_confidence", "l2": 9.452337, "mean_abs": 0.179481, "relative": 0.000314 }, { "name": "video_fisheye_cam0", "l2": 5.098331, "mean_abs": 0.105755, "relative": 0.923199 }, { "name": "video_fisheye_cam1", "l2": 5.331143, "mean_abs": 0.114143, "relative": 0.880926 }, { "name": "video_fisheye_cam2", "l2": 5.787776, "mean_abs": 0.117715, "relative": 0.923932 }, { "name": "video_fisheye_cam3", "l2": 5.519601, "mean_abs": 0.114482, "relative": 0.924191 }, { "name": "video_stereo_left", "l2": 5.501271, "mean_abs": 0.120829, "relative": 0.893967 }, { "name": "video_stereo_right", "l2": 6.397554, "mean_abs": 0.145428, "relative": 0.901921 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.084978, "mean_abs": 0.082085, "relative": 0.746182 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 215, "start_frame": 1075, "end_frame": 1094, "center_frame": 1084, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.668243, "mean_abs": 0.248939, "relative": 0.899912 }, { "name": "hand_right_joints", "l2": 11.859858, "mean_abs": 0.291463, "relative": 0.934874 }, { "name": "body_joints", "l2": 5.927044, "mean_abs": 0.09843, "relative": 0.800561 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.051564, "mean_abs": 0.007533, "relative": 0.121683 }, { "name": "camera_rotation_matrix", "l2": 3.000077, "mean_abs": 0.183444, "relative": 0.921275 }, { "name": "imu_accel_gyro", "l2": 17.032164, "mean_abs": 1.088839, "relative": 0.918944 }, { "name": "depth_confidence", "l2": 9.463835, "mean_abs": 0.178095, "relative": 0.000314 }, { "name": "video_fisheye_cam0", "l2": 5.09411, "mean_abs": 0.105161, "relative": 0.922434 }, { "name": "video_fisheye_cam1", "l2": 5.357375, "mean_abs": 0.114212, "relative": 0.885261 }, { "name": "video_fisheye_cam2", "l2": 5.80498, "mean_abs": 0.117743, "relative": 0.926679 }, { "name": "video_fisheye_cam3", "l2": 5.52825, "mean_abs": 0.11415, "relative": 0.925639 }, { "name": "video_stereo_left", "l2": 5.524683, "mean_abs": 0.11975, "relative": 0.897772 }, { "name": "video_stereo_right", "l2": 6.432981, "mean_abs": 0.144409, "relative": 0.906915 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.033665, "mean_abs": 0.078451, "relative": 0.727818 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 216, "start_frame": 1080, "end_frame": 1099, "center_frame": 1089, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.657224, "mean_abs": 0.244866, "relative": 0.898982 }, { "name": "hand_right_joints", "l2": 12.184105, "mean_abs": 0.287157, "relative": 0.960433 }, { "name": "body_joints", "l2": 5.755909, "mean_abs": 0.09368, "relative": 0.777446 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.040049, "mean_abs": 0.005938, "relative": 0.094511 }, { "name": "camera_rotation_matrix", "l2": 3.000247, "mean_abs": 0.184432, "relative": 0.921327 }, { "name": "imu_accel_gyro", "l2": 17.089405, "mean_abs": 1.094793, "relative": 0.922032 }, { "name": "depth_confidence", "l2": 9.486773, "mean_abs": 0.176669, "relative": 0.000315 }, { "name": "video_fisheye_cam0", "l2": 5.096543, "mean_abs": 0.105043, "relative": 0.922875 }, { "name": "video_fisheye_cam1", "l2": 5.397748, "mean_abs": 0.114241, "relative": 0.891932 }, { "name": "video_fisheye_cam2", "l2": 5.82968, "mean_abs": 0.117624, "relative": 0.930622 }, { "name": "video_fisheye_cam3", "l2": 5.548223, "mean_abs": 0.113794, "relative": 0.928983 }, { "name": "video_stereo_left", "l2": 5.56896, "mean_abs": 0.119392, "relative": 0.904967 }, { "name": "video_stereo_right", "l2": 6.511844, "mean_abs": 0.143675, "relative": 0.918033 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.013742, "mean_abs": 0.079595, "relative": 0.720688 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 217, "start_frame": 1085, "end_frame": 1104, "center_frame": 1094, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.700838, "mean_abs": 0.243863, "relative": 0.902661 }, { "name": "hand_right_joints", "l2": 12.286076, "mean_abs": 0.289293, "relative": 0.968471 }, { "name": "body_joints", "l2": 5.740428, "mean_abs": 0.09079, "relative": 0.775355 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.040115, "mean_abs": 0.005471, "relative": 0.094667 }, { "name": "camera_rotation_matrix", "l2": 3.000437, "mean_abs": 0.185033, "relative": 0.921386 }, { "name": "imu_accel_gyro", "l2": 17.083263, "mean_abs": 1.091039, "relative": 0.921701 }, { "name": "depth_confidence", "l2": 9.524195, "mean_abs": 0.177449, "relative": 0.000316 }, { "name": "video_fisheye_cam0", "l2": 5.101391, "mean_abs": 0.105216, "relative": 0.923753 }, { "name": "video_fisheye_cam1", "l2": 5.440222, "mean_abs": 0.114859, "relative": 0.898951 }, { "name": "video_fisheye_cam2", "l2": 5.84296, "mean_abs": 0.117478, "relative": 0.932742 }, { "name": "video_fisheye_cam3", "l2": 5.556661, "mean_abs": 0.114293, "relative": 0.930396 }, { "name": "video_stereo_left", "l2": 5.575817, "mean_abs": 0.119345, "relative": 0.906081 }, { "name": "video_stereo_right", "l2": 6.568178, "mean_abs": 0.14451, "relative": 0.925975 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.025769, "mean_abs": 0.081934, "relative": 0.724992 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 218, "start_frame": 1090, "end_frame": 1109, "center_frame": 1099, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.674342, "mean_abs": 0.247813, "relative": 0.900426 }, { "name": "hand_right_joints", "l2": 12.160399, "mean_abs": 0.293967, "relative": 0.958565 }, { "name": "body_joints", "l2": 5.787109, "mean_abs": 0.09094, "relative": 0.78166 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.044488, "mean_abs": 0.00666, "relative": 0.104985 }, { "name": "camera_rotation_matrix", "l2": 3.000305, "mean_abs": 0.185319, "relative": 0.921345 }, { "name": "imu_accel_gyro", "l2": 17.10182, "mean_abs": 1.088578, "relative": 0.922702 }, { "name": "depth_confidence", "l2": 9.587168, "mean_abs": 0.178661, "relative": 0.000318 }, { "name": "video_fisheye_cam0", "l2": 5.105815, "mean_abs": 0.105105, "relative": 0.924554 }, { "name": "video_fisheye_cam1", "l2": 5.465494, "mean_abs": 0.11535, "relative": 0.903127 }, { "name": "video_fisheye_cam2", "l2": 5.844683, "mean_abs": 0.117472, "relative": 0.933017 }, { "name": "video_fisheye_cam3", "l2": 5.535947, "mean_abs": 0.113687, "relative": 0.926928 }, { "name": "video_stereo_left", "l2": 5.572179, "mean_abs": 0.119077, "relative": 0.90549 }, { "name": "video_stereo_right", "l2": 6.592651, "mean_abs": 0.144618, "relative": 0.929425 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.076737, "mean_abs": 0.082883, "relative": 0.743233 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 219, "start_frame": 1095, "end_frame": 1114, "center_frame": 1104, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.626714, "mean_abs": 0.250962, "relative": 0.896409 }, { "name": "hand_right_joints", "l2": 12.038918, "mean_abs": 0.299252, "relative": 0.948989 }, { "name": "body_joints", "l2": 5.845528, "mean_abs": 0.093379, "relative": 0.789551 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.056114, "mean_abs": 0.008537, "relative": 0.132421 }, { "name": "camera_rotation_matrix", "l2": 3.000434, "mean_abs": 0.186646, "relative": 0.921385 }, { "name": "imu_accel_gyro", "l2": 17.22048, "mean_abs": 1.095848, "relative": 0.929104 }, { "name": "depth_confidence", "l2": 9.643493, "mean_abs": 0.1815, "relative": 0.00032 }, { "name": "video_fisheye_cam0", "l2": 5.116421, "mean_abs": 0.10554, "relative": 0.926474 }, { "name": "video_fisheye_cam1", "l2": 5.481542, "mean_abs": 0.116204, "relative": 0.905778 }, { "name": "video_fisheye_cam2", "l2": 5.84715, "mean_abs": 0.117865, "relative": 0.93341 }, { "name": "video_fisheye_cam3", "l2": 5.529119, "mean_abs": 0.114032, "relative": 0.925784 }, { "name": "video_stereo_left", "l2": 5.559619, "mean_abs": 0.119945, "relative": 0.903449 }, { "name": "video_stereo_right", "l2": 6.582797, "mean_abs": 0.146354, "relative": 0.928036 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.092408, "mean_abs": 0.082372, "relative": 0.748842 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 220, "start_frame": 1100, "end_frame": 1119, "center_frame": 1109, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.620428, "mean_abs": 0.251656, "relative": 0.895878 }, { "name": "hand_right_joints", "l2": 11.977016, "mean_abs": 0.30512, "relative": 0.944109 }, { "name": "body_joints", "l2": 5.838685, "mean_abs": 0.095457, "relative": 0.788626 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.069867, "mean_abs": 0.010565, "relative": 0.164877 }, { "name": "camera_rotation_matrix", "l2": 3.00063, "mean_abs": 0.188518, "relative": 0.921445 }, { "name": "imu_accel_gyro", "l2": 17.223585, "mean_abs": 1.105385, "relative": 0.929272 }, { "name": "depth_confidence", "l2": 9.689111, "mean_abs": 0.183615, "relative": 0.000322 }, { "name": "video_fisheye_cam0", "l2": 5.12572, "mean_abs": 0.106272, "relative": 0.928158 }, { "name": "video_fisheye_cam1", "l2": 5.493847, "mean_abs": 0.117025, "relative": 0.907812 }, { "name": "video_fisheye_cam2", "l2": 5.852321, "mean_abs": 0.118524, "relative": 0.934236 }, { "name": "video_fisheye_cam3", "l2": 5.511529, "mean_abs": 0.113822, "relative": 0.922839 }, { "name": "video_stereo_left", "l2": 5.553488, "mean_abs": 0.12093, "relative": 0.902453 }, { "name": "video_stereo_right", "l2": 6.584367, "mean_abs": 0.147568, "relative": 0.928258 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.131394, "mean_abs": 0.08221, "relative": 0.762794 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 221, "start_frame": 1105, "end_frame": 1124, "center_frame": 1114, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.583133, "mean_abs": 0.249057, "relative": 0.892732 }, { "name": "hand_right_joints", "l2": 11.879939, "mean_abs": 0.303538, "relative": 0.936457 }, { "name": "body_joints", "l2": 5.836271, "mean_abs": 0.095067, "relative": 0.7883 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.070528, "mean_abs": 0.010736, "relative": 0.166436 }, { "name": "camera_rotation_matrix", "l2": 3.00095, "mean_abs": 0.189692, "relative": 0.921543 }, { "name": "imu_accel_gyro", "l2": 17.100208, "mean_abs": 1.079978, "relative": 0.922615 }, { "name": "depth_confidence", "l2": 9.754352, "mean_abs": 0.184379, "relative": 0.000324 }, { "name": "video_fisheye_cam0", "l2": 5.131281, "mean_abs": 0.106441, "relative": 0.929165 }, { "name": "video_fisheye_cam1", "l2": 5.508876, "mean_abs": 0.117291, "relative": 0.910295 }, { "name": "video_fisheye_cam2", "l2": 5.85231, "mean_abs": 0.118509, "relative": 0.934234 }, { "name": "video_fisheye_cam3", "l2": 5.485551, "mean_abs": 0.113565, "relative": 0.918489 }, { "name": "video_stereo_left", "l2": 5.538413, "mean_abs": 0.120982, "relative": 0.900003 }, { "name": "video_stereo_right", "l2": 6.587065, "mean_abs": 0.147707, "relative": 0.928638 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.165217, "mean_abs": 0.08319, "relative": 0.774899 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 222, "start_frame": 1110, "end_frame": 1129, "center_frame": 1119, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.561499, "mean_abs": 0.247701, "relative": 0.890907 }, { "name": "hand_right_joints", "l2": 11.70315, "mean_abs": 0.293899, "relative": 0.922521 }, { "name": "body_joints", "l2": 5.818568, "mean_abs": 0.092445, "relative": 0.785909 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.079007, "mean_abs": 0.011958, "relative": 0.186445 }, { "name": "camera_rotation_matrix", "l2": 3.00063, "mean_abs": 0.189469, "relative": 0.921445 }, { "name": "imu_accel_gyro", "l2": 17.063408, "mean_abs": 1.069713, "relative": 0.92063 }, { "name": "depth_confidence", "l2": 9.798638, "mean_abs": 0.184169, "relative": 0.000325 }, { "name": "video_fisheye_cam0", "l2": 5.129586, "mean_abs": 0.106544, "relative": 0.928858 }, { "name": "video_fisheye_cam1", "l2": 5.512687, "mean_abs": 0.117084, "relative": 0.910925 }, { "name": "video_fisheye_cam2", "l2": 5.854301, "mean_abs": 0.118582, "relative": 0.934552 }, { "name": "video_fisheye_cam3", "l2": 5.440595, "mean_abs": 0.11278, "relative": 0.910962 }, { "name": "video_stereo_left", "l2": 5.526526, "mean_abs": 0.120454, "relative": 0.898071 }, { "name": "video_stereo_right", "l2": 6.556009, "mean_abs": 0.145614, "relative": 0.92426 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.158157, "mean_abs": 0.083761, "relative": 0.772372 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 223, "start_frame": 1115, "end_frame": 1134, "center_frame": 1124, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.525399, "mean_abs": 0.243134, "relative": 0.887862 }, { "name": "hand_right_joints", "l2": 11.585262, "mean_abs": 0.282728, "relative": 0.913229 }, { "name": "body_joints", "l2": 5.779827, "mean_abs": 0.088998, "relative": 0.780677 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.079018, "mean_abs": 0.011916, "relative": 0.186471 }, { "name": "camera_rotation_matrix", "l2": 3.000989, "mean_abs": 0.189617, "relative": 0.921555 }, { "name": "imu_accel_gyro", "l2": 16.996714, "mean_abs": 1.068816, "relative": 0.917031 }, { "name": "depth_confidence", "l2": 9.827844, "mean_abs": 0.184122, "relative": 0.000326 }, { "name": "video_fisheye_cam0", "l2": 5.133628, "mean_abs": 0.106128, "relative": 0.92959 }, { "name": "video_fisheye_cam1", "l2": 5.536901, "mean_abs": 0.116976, "relative": 0.914926 }, { "name": "video_fisheye_cam2", "l2": 5.848598, "mean_abs": 0.117962, "relative": 0.933642 }, { "name": "video_fisheye_cam3", "l2": 5.434128, "mean_abs": 0.11244, "relative": 0.909879 }, { "name": "video_stereo_left", "l2": 5.510292, "mean_abs": 0.119448, "relative": 0.895433 }, { "name": "video_stereo_right", "l2": 6.565249, "mean_abs": 0.145312, "relative": 0.925562 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.083824, "mean_abs": 0.080377, "relative": 0.745769 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 224, "start_frame": 1120, "end_frame": 1139, "center_frame": 1129, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.502692, "mean_abs": 0.24085, "relative": 0.885947 }, { "name": "hand_right_joints", "l2": 11.568039, "mean_abs": 0.28422, "relative": 0.911871 }, { "name": "body_joints", "l2": 5.780667, "mean_abs": 0.088296, "relative": 0.78079 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.066568, "mean_abs": 0.009783, "relative": 0.157091 }, { "name": "camera_rotation_matrix", "l2": 3.000708, "mean_abs": 0.189973, "relative": 0.921469 }, { "name": "imu_accel_gyro", "l2": 17.037531, "mean_abs": 1.070335, "relative": 0.919233 }, { "name": "depth_confidence", "l2": 9.868249, "mean_abs": 0.184135, "relative": 0.000328 }, { "name": "video_fisheye_cam0", "l2": 5.133945, "mean_abs": 0.106116, "relative": 0.929648 }, { "name": "video_fisheye_cam1", "l2": 5.560923, "mean_abs": 0.11761, "relative": 0.918896 }, { "name": "video_fisheye_cam2", "l2": 5.853906, "mean_abs": 0.117869, "relative": 0.934489 }, { "name": "video_fisheye_cam3", "l2": 5.431513, "mean_abs": 0.112555, "relative": 0.909441 }, { "name": "video_stereo_left", "l2": 5.517529, "mean_abs": 0.119276, "relative": 0.896609 }, { "name": "video_stereo_right", "l2": 6.592287, "mean_abs": 0.145796, "relative": 0.929374 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.096322, "mean_abs": 0.08244, "relative": 0.750242 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 225, "start_frame": 1125, "end_frame": 1144, "center_frame": 1134, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.473976, "mean_abs": 0.240211, "relative": 0.883525 }, { "name": "hand_right_joints", "l2": 11.552858, "mean_abs": 0.285974, "relative": 0.910674 }, { "name": "body_joints", "l2": 5.796565, "mean_abs": 0.088466, "relative": 0.782937 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.056901, "mean_abs": 0.007357, "relative": 0.134278 }, { "name": "camera_rotation_matrix", "l2": 3.000189, "mean_abs": 0.190225, "relative": 0.921309 }, { "name": "imu_accel_gyro", "l2": 17.053875, "mean_abs": 1.063729, "relative": 0.920115 }, { "name": "depth_confidence", "l2": 9.919989, "mean_abs": 0.183941, "relative": 0.000329 }, { "name": "video_fisheye_cam0", "l2": 5.130832, "mean_abs": 0.105971, "relative": 0.929084 }, { "name": "video_fisheye_cam1", "l2": 5.564039, "mean_abs": 0.117328, "relative": 0.91941 }, { "name": "video_fisheye_cam2", "l2": 5.852661, "mean_abs": 0.117746, "relative": 0.93429 }, { "name": "video_fisheye_cam3", "l2": 5.422503, "mean_abs": 0.111856, "relative": 0.907933 }, { "name": "video_stereo_left", "l2": 5.515517, "mean_abs": 0.118721, "relative": 0.896282 }, { "name": "video_stereo_right", "l2": 6.590417, "mean_abs": 0.144894, "relative": 0.929111 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.105363, "mean_abs": 0.083676, "relative": 0.753478 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 226, "start_frame": 1130, "end_frame": 1149, "center_frame": 1139, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.44679, "mean_abs": 0.239396, "relative": 0.881231 }, { "name": "hand_right_joints", "l2": 11.533735, "mean_abs": 0.286615, "relative": 0.909167 }, { "name": "body_joints", "l2": 5.831766, "mean_abs": 0.088972, "relative": 0.787692 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.039346, "mean_abs": 0.00537, "relative": 0.092851 }, { "name": "camera_rotation_matrix", "l2": 3.000079, "mean_abs": 0.190765, "relative": 0.921276 }, { "name": "imu_accel_gyro", "l2": 17.019239, "mean_abs": 1.055493, "relative": 0.918247 }, { "name": "depth_confidence", "l2": 9.961449, "mean_abs": 0.184006, "relative": 0.000331 }, { "name": "video_fisheye_cam0", "l2": 5.132982, "mean_abs": 0.10601, "relative": 0.929473 }, { "name": "video_fisheye_cam1", "l2": 5.578373, "mean_abs": 0.117307, "relative": 0.921779 }, { "name": "video_fisheye_cam2", "l2": 5.853628, "mean_abs": 0.117722, "relative": 0.934445 }, { "name": "video_fisheye_cam3", "l2": 5.439197, "mean_abs": 0.111764, "relative": 0.910728 }, { "name": "video_stereo_left", "l2": 5.515823, "mean_abs": 0.118457, "relative": 0.896332 }, { "name": "video_stereo_right", "l2": 6.608791, "mean_abs": 0.14485, "relative": 0.931701 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.13825, "mean_abs": 0.083119, "relative": 0.765248 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 227, "start_frame": 1135, "end_frame": 1154, "center_frame": 1144, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.412838, "mean_abs": 0.238616, "relative": 0.878367 }, { "name": "hand_right_joints", "l2": 11.491485, "mean_abs": 0.288008, "relative": 0.905836 }, { "name": "body_joints", "l2": 5.870174, "mean_abs": 0.090286, "relative": 0.79288 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.025455, "mean_abs": 0.003507, "relative": 0.060071 }, { "name": "camera_rotation_matrix", "l2": 2.999982, "mean_abs": 0.19047, "relative": 0.921246 }, { "name": "imu_accel_gyro", "l2": 17.026558, "mean_abs": 1.043733, "relative": 0.918641 }, { "name": "depth_confidence", "l2": 9.984816, "mean_abs": 0.184033, "relative": 0.000331 }, { "name": "video_fisheye_cam0", "l2": 5.135194, "mean_abs": 0.106036, "relative": 0.929874 }, { "name": "video_fisheye_cam1", "l2": 5.574895, "mean_abs": 0.117157, "relative": 0.921204 }, { "name": "video_fisheye_cam2", "l2": 5.839289, "mean_abs": 0.117307, "relative": 0.932156 }, { "name": "video_fisheye_cam3", "l2": 5.456046, "mean_abs": 0.112108, "relative": 0.913549 }, { "name": "video_stereo_left", "l2": 5.509671, "mean_abs": 0.118087, "relative": 0.895332 }, { "name": "video_stereo_right", "l2": 6.60414, "mean_abs": 0.144662, "relative": 0.931045 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.168853, "mean_abs": 0.085851, "relative": 0.7762 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 228, "start_frame": 1140, "end_frame": 1159, "center_frame": 1149, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.343314, "mean_abs": 0.238394, "relative": 0.872503 }, { "name": "hand_right_joints", "l2": 11.4101, "mean_abs": 0.288481, "relative": 0.899421 }, { "name": "body_joints", "l2": 5.905554, "mean_abs": 0.093536, "relative": 0.797658 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019501, "mean_abs": 0.002992, "relative": 0.046019 }, { "name": "camera_rotation_matrix", "l2": 3.00116, "mean_abs": 0.193502, "relative": 0.921608 }, { "name": "imu_accel_gyro", "l2": 17.081573, "mean_abs": 1.076389, "relative": 0.92161 }, { "name": "depth_confidence", "l2": 9.9741, "mean_abs": 0.186975, "relative": 0.000331 }, { "name": "video_fisheye_cam0", "l2": 5.13819, "mean_abs": 0.106598, "relative": 0.930416 }, { "name": "video_fisheye_cam1", "l2": 5.550334, "mean_abs": 0.117989, "relative": 0.917146 }, { "name": "video_fisheye_cam2", "l2": 5.796247, "mean_abs": 0.118167, "relative": 0.925285 }, { "name": "video_fisheye_cam3", "l2": 5.497103, "mean_abs": 0.113027, "relative": 0.920424 }, { "name": "video_stereo_left", "l2": 5.458803, "mean_abs": 0.12037, "relative": 0.887066 }, { "name": "video_stereo_right", "l2": 6.520037, "mean_abs": 0.146197, "relative": 0.919188 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.093109, "mean_abs": 0.082802, "relative": 0.749092 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 229, "start_frame": 1145, "end_frame": 1164, "center_frame": 1154, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.263104, "mean_abs": 0.237842, "relative": 0.865737 }, { "name": "hand_right_joints", "l2": 11.363661, "mean_abs": 0.291842, "relative": 0.89576 }, { "name": "body_joints", "l2": 6.037511, "mean_abs": 0.098749, "relative": 0.815482 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.075767, "mean_abs": 0.010759, "relative": 0.178799 }, { "name": "camera_rotation_matrix", "l2": 3.013168, "mean_abs": 0.199306, "relative": 0.925295 }, { "name": "imu_accel_gyro", "l2": 17.119617, "mean_abs": 1.12522, "relative": 0.923662 }, { "name": "depth_confidence", "l2": 9.903814, "mean_abs": 0.190528, "relative": 0.000329 }, { "name": "video_fisheye_cam0", "l2": 5.133502, "mean_abs": 0.108774, "relative": 0.929567 }, { "name": "video_fisheye_cam1", "l2": 5.578696, "mean_abs": 0.122363, "relative": 0.921832 }, { "name": "video_fisheye_cam2", "l2": 5.701347, "mean_abs": 0.120172, "relative": 0.910135 }, { "name": "video_fisheye_cam3", "l2": 5.564585, "mean_abs": 0.11545, "relative": 0.931723 }, { "name": "video_stereo_left", "l2": 5.390152, "mean_abs": 0.124786, "relative": 0.87591 }, { "name": "video_stereo_right", "l2": 6.523768, "mean_abs": 0.153661, "relative": 0.919714 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.105416, "mean_abs": 0.081054, "relative": 0.753497 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 230, "start_frame": 1150, "end_frame": 1169, "center_frame": 1159, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.166657, "mean_abs": 0.232817, "relative": 0.857601 }, { "name": "hand_right_joints", "l2": 11.429378, "mean_abs": 0.306531, "relative": 0.900941 }, { "name": "body_joints", "l2": 6.322975, "mean_abs": 0.103735, "relative": 0.854039 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.138007, "mean_abs": 0.020099, "relative": 0.325676 }, { "name": "camera_rotation_matrix", "l2": 3.022549, "mean_abs": 0.201558, "relative": 0.928176 }, { "name": "imu_accel_gyro", "l2": 17.057997, "mean_abs": 1.112907, "relative": 0.920338 }, { "name": "depth_confidence", "l2": 9.856378, "mean_abs": 0.193096, "relative": 0.000327 }, { "name": "video_fisheye_cam0", "l2": 5.126332, "mean_abs": 0.10978, "relative": 0.928269 }, { "name": "video_fisheye_cam1", "l2": 5.624275, "mean_abs": 0.125198, "relative": 0.929364 }, { "name": "video_fisheye_cam2", "l2": 5.645728, "mean_abs": 0.120467, "relative": 0.901256 }, { "name": "video_fisheye_cam3", "l2": 5.652111, "mean_abs": 0.117764, "relative": 0.946378 }, { "name": "video_stereo_left", "l2": 5.360703, "mean_abs": 0.12745, "relative": 0.871125 }, { "name": "video_stereo_right", "l2": 6.510336, "mean_abs": 0.15664, "relative": 0.917821 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.097718, "mean_abs": 0.08235, "relative": 0.750742 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 231, "start_frame": 1155, "end_frame": 1174, "center_frame": 1164, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.076993, "mean_abs": 0.229956, "relative": 0.850037 }, { "name": "hand_right_joints", "l2": 11.589646, "mean_abs": 0.313877, "relative": 0.913574 }, { "name": "body_joints", "l2": 6.448197, "mean_abs": 0.104677, "relative": 0.870953 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.161315, "mean_abs": 0.023914, "relative": 0.380681 }, { "name": "camera_rotation_matrix", "l2": 3.011157, "mean_abs": 0.195712, "relative": 0.924677 }, { "name": "imu_accel_gyro", "l2": 17.070223, "mean_abs": 1.125284, "relative": 0.920997 }, { "name": "depth_confidence", "l2": 9.790661, "mean_abs": 0.190687, "relative": 0.000325 }, { "name": "video_fisheye_cam0", "l2": 5.111068, "mean_abs": 0.108743, "relative": 0.925505 }, { "name": "video_fisheye_cam1", "l2": 5.611635, "mean_abs": 0.124169, "relative": 0.927275 }, { "name": "video_fisheye_cam2", "l2": 5.580585, "mean_abs": 0.118672, "relative": 0.890857 }, { "name": "video_fisheye_cam3", "l2": 5.639736, "mean_abs": 0.117011, "relative": 0.944306 }, { "name": "video_stereo_left", "l2": 5.271811, "mean_abs": 0.124668, "relative": 0.85668 }, { "name": "video_stereo_right", "l2": 6.490702, "mean_abs": 0.155606, "relative": 0.915053 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.101352, "mean_abs": 0.081889, "relative": 0.752042 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 232, "start_frame": 1160, "end_frame": 1179, "center_frame": 1169, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.972254, "mean_abs": 0.220962, "relative": 0.841202 }, { "name": "hand_right_joints", "l2": 11.621609, "mean_abs": 0.305436, "relative": 0.916094 }, { "name": "body_joints", "l2": 6.456597, "mean_abs": 0.102491, "relative": 0.872087 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.137554, "mean_abs": 0.020513, "relative": 0.324607 }, { "name": "camera_rotation_matrix", "l2": 3.001069, "mean_abs": 0.18689, "relative": 0.92158 }, { "name": "imu_accel_gyro", "l2": 17.279175, "mean_abs": 1.156134, "relative": 0.932271 }, { "name": "depth_confidence", "l2": 9.729583, "mean_abs": 0.186066, "relative": 0.000323 }, { "name": "video_fisheye_cam0", "l2": 5.082362, "mean_abs": 0.106287, "relative": 0.920307 }, { "name": "video_fisheye_cam1", "l2": 5.550592, "mean_abs": 0.12008, "relative": 0.917189 }, { "name": "video_fisheye_cam2", "l2": 5.49016, "mean_abs": 0.113884, "relative": 0.876422 }, { "name": "video_fisheye_cam3", "l2": 5.611597, "mean_abs": 0.117619, "relative": 0.939594 }, { "name": "video_stereo_left", "l2": 5.139236, "mean_abs": 0.117937, "relative": 0.835136 }, { "name": "video_stereo_right", "l2": 6.309856, "mean_abs": 0.147861, "relative": 0.889557 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.174239, "mean_abs": 0.088129, "relative": 0.778127 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 233, "start_frame": 1165, "end_frame": 1184, "center_frame": 1174, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.847648, "mean_abs": 0.209021, "relative": 0.830691 }, { "name": "hand_right_joints", "l2": 11.717156, "mean_abs": 0.308132, "relative": 0.923625 }, { "name": "body_joints", "l2": 6.510279, "mean_abs": 0.098692, "relative": 0.879338 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.070878, "mean_abs": 0.010976, "relative": 0.167261 }, { "name": "camera_rotation_matrix", "l2": 3.017892, "mean_abs": 0.202035, "relative": 0.926746 }, { "name": "imu_accel_gyro", "l2": 17.431627, "mean_abs": 1.190081, "relative": 0.940496 }, { "name": "depth_confidence", "l2": 9.818178, "mean_abs": 0.190918, "relative": 0.000326 }, { "name": "video_fisheye_cam0", "l2": 5.102385, "mean_abs": 0.10829, "relative": 0.923933 }, { "name": "video_fisheye_cam1", "l2": 5.584693, "mean_abs": 0.123842, "relative": 0.922823 }, { "name": "video_fisheye_cam2", "l2": 5.555623, "mean_abs": 0.117446, "relative": 0.886872 }, { "name": "video_fisheye_cam3", "l2": 5.580683, "mean_abs": 0.118399, "relative": 0.934418 }, { "name": "video_stereo_left", "l2": 5.202725, "mean_abs": 0.122513, "relative": 0.845453 }, { "name": "video_stereo_right", "l2": 6.442318, "mean_abs": 0.15444, "relative": 0.908232 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.15422, "mean_abs": 0.08585, "relative": 0.770963 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 234, "start_frame": 1170, "end_frame": 1189, "center_frame": 1179, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.762716, "mean_abs": 0.212121, "relative": 0.823527 }, { "name": "hand_right_joints", "l2": 11.610405, "mean_abs": 0.309217, "relative": 0.91521 }, { "name": "body_joints", "l2": 6.70899, "mean_abs": 0.097498, "relative": 0.906178 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.097163, "mean_abs": 0.015399, "relative": 0.229291 }, { "name": "camera_rotation_matrix", "l2": 3.025798, "mean_abs": 0.208261, "relative": 0.929173 }, { "name": "imu_accel_gyro", "l2": 17.398165, "mean_abs": 1.144485, "relative": 0.938691 }, { "name": "depth_confidence", "l2": 10.056336, "mean_abs": 0.195824, "relative": 0.000334 }, { "name": "video_fisheye_cam0", "l2": 5.127124, "mean_abs": 0.109119, "relative": 0.928412 }, { "name": "video_fisheye_cam1", "l2": 5.593005, "mean_abs": 0.124192, "relative": 0.924197 }, { "name": "video_fisheye_cam2", "l2": 5.607126, "mean_abs": 0.11846, "relative": 0.895094 }, { "name": "video_fisheye_cam3", "l2": 5.500132, "mean_abs": 0.115532, "relative": 0.920931 }, { "name": "video_stereo_left", "l2": 5.246176, "mean_abs": 0.122121, "relative": 0.852514 }, { "name": "video_stereo_right", "l2": 6.473189, "mean_abs": 0.15296, "relative": 0.912584 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.250177, "mean_abs": 0.090693, "relative": 0.805304 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 235, "start_frame": 1175, "end_frame": 1194, "center_frame": 1184, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.796726, "mean_abs": 0.229997, "relative": 0.826396 }, { "name": "hand_right_joints", "l2": 11.259149, "mean_abs": 0.30809, "relative": 0.887522 }, { "name": "body_joints", "l2": 6.826025, "mean_abs": 0.107936, "relative": 0.921986 }, { "name": "body_contacts", "l2": 3.464914, "mean_abs": 0.081711, "relative": 0.999859 }, { "name": "camera_translation", "l2": 0.191521, "mean_abs": 0.027189, "relative": 0.451964 }, { "name": "camera_rotation_matrix", "l2": 3.02122, "mean_abs": 0.212257, "relative": 0.927768 }, { "name": "imu_accel_gyro", "l2": 17.244925, "mean_abs": 1.155734, "relative": 0.930423 }, { "name": "depth_confidence", "l2": 10.324035, "mean_abs": 0.201194, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.127565, "mean_abs": 0.108939, "relative": 0.928492 }, { "name": "video_fisheye_cam1", "l2": 5.601254, "mean_abs": 0.123833, "relative": 0.92556 }, { "name": "video_fisheye_cam2", "l2": 5.688551, "mean_abs": 0.12055, "relative": 0.908093 }, { "name": "video_fisheye_cam3", "l2": 5.450869, "mean_abs": 0.114497, "relative": 0.912682 }, { "name": "video_stereo_left", "l2": 5.330575, "mean_abs": 0.1229, "relative": 0.866229 }, { "name": "video_stereo_right", "l2": 6.543891, "mean_abs": 0.152743, "relative": 0.922551 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.140582, "mean_abs": 0.086766, "relative": 0.766082 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 236, "start_frame": 1180, "end_frame": 1199, "center_frame": 1189, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.835484, "mean_abs": 0.235146, "relative": 0.829665 }, { "name": "hand_right_joints", "l2": 11.017225, "mean_abs": 0.306134, "relative": 0.868452 }, { "name": "body_joints", "l2": 6.814945, "mean_abs": 0.110785, "relative": 0.920489 }, { "name": "body_contacts", "l2": 3.464914, "mean_abs": 0.081716, "relative": 0.999859 }, { "name": "camera_translation", "l2": 0.252749, "mean_abs": 0.032564, "relative": 0.596452 }, { "name": "camera_rotation_matrix", "l2": 3.012804, "mean_abs": 0.213872, "relative": 0.925183 }, { "name": "imu_accel_gyro", "l2": 17.465532, "mean_abs": 1.200046, "relative": 0.942326 }, { "name": "depth_confidence", "l2": 10.577773, "mean_abs": 0.203852, "relative": 0.000351 }, { "name": "video_fisheye_cam0", "l2": 5.121615, "mean_abs": 0.10813, "relative": 0.927415 }, { "name": "video_fisheye_cam1", "l2": 5.609855, "mean_abs": 0.122833, "relative": 0.926981 }, { "name": "video_fisheye_cam2", "l2": 5.765234, "mean_abs": 0.120886, "relative": 0.920334 }, { "name": "video_fisheye_cam3", "l2": 5.463777, "mean_abs": 0.114993, "relative": 0.914843 }, { "name": "video_stereo_left", "l2": 5.407699, "mean_abs": 0.124452, "relative": 0.878762 }, { "name": "video_stereo_right", "l2": 6.595116, "mean_abs": 0.153751, "relative": 0.929773 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.061611, "mean_abs": 0.083309, "relative": 0.73782 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 237, "start_frame": 1185, "end_frame": 1204, "center_frame": 1194, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.839325, "mean_abs": 0.234276, "relative": 0.829989 }, { "name": "hand_right_joints", "l2": 10.60813, "mean_abs": 0.291456, "relative": 0.836204 }, { "name": "body_joints", "l2": 6.699623, "mean_abs": 0.108562, "relative": 0.904913 }, { "name": "body_contacts", "l2": 3.464914, "mean_abs": 0.081719, "relative": 0.999859 }, { "name": "camera_translation", "l2": 0.257381, "mean_abs": 0.032308, "relative": 0.607382 }, { "name": "camera_rotation_matrix", "l2": 3.009311, "mean_abs": 0.214749, "relative": 0.924111 }, { "name": "imu_accel_gyro", "l2": 17.348228, "mean_abs": 1.151197, "relative": 0.935997 }, { "name": "depth_confidence", "l2": 10.745106, "mean_abs": 0.204448, "relative": 0.000357 }, { "name": "video_fisheye_cam0", "l2": 5.122419, "mean_abs": 0.107961, "relative": 0.927561 }, { "name": "video_fisheye_cam1", "l2": 5.648829, "mean_abs": 0.123284, "relative": 0.933421 }, { "name": "video_fisheye_cam2", "l2": 5.793625, "mean_abs": 0.121021, "relative": 0.924866 }, { "name": "video_fisheye_cam3", "l2": 5.507761, "mean_abs": 0.115235, "relative": 0.922208 }, { "name": "video_stereo_left", "l2": 5.445776, "mean_abs": 0.125147, "relative": 0.884949 }, { "name": "video_stereo_right", "l2": 6.620949, "mean_abs": 0.154804, "relative": 0.933415 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.000259, "mean_abs": 0.078864, "relative": 0.715863 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 238, "start_frame": 1190, "end_frame": 1209, "center_frame": 1199, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.874255, "mean_abs": 0.233493, "relative": 0.832936 }, { "name": "hand_right_joints", "l2": 9.94333, "mean_abs": 0.260394, "relative": 0.7838 }, { "name": "body_joints", "l2": 6.522403, "mean_abs": 0.100486, "relative": 0.880976 }, { "name": "body_contacts", "l2": 3.464914, "mean_abs": 0.08173, "relative": 0.999859 }, { "name": "camera_translation", "l2": 0.20942, "mean_abs": 0.025095, "relative": 0.494201 }, { "name": "camera_rotation_matrix", "l2": 3.006326, "mean_abs": 0.215207, "relative": 0.923194 }, { "name": "imu_accel_gyro", "l2": 16.952681, "mean_abs": 1.066323, "relative": 0.914656 }, { "name": "depth_confidence", "l2": 10.951286, "mean_abs": 0.204419, "relative": 0.000363 }, { "name": "video_fisheye_cam0", "l2": 5.10665, "mean_abs": 0.107335, "relative": 0.924705 }, { "name": "video_fisheye_cam1", "l2": 5.692331, "mean_abs": 0.123285, "relative": 0.94061 }, { "name": "video_fisheye_cam2", "l2": 5.803892, "mean_abs": 0.121065, "relative": 0.926505 }, { "name": "video_fisheye_cam3", "l2": 5.497887, "mean_abs": 0.114528, "relative": 0.920555 }, { "name": "video_stereo_left", "l2": 5.465862, "mean_abs": 0.125115, "relative": 0.888213 }, { "name": "video_stereo_right", "l2": 6.671052, "mean_abs": 0.154502, "relative": 0.940478 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.10241, "mean_abs": 0.088122, "relative": 0.752421 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 239, "start_frame": 1195, "end_frame": 1214, "center_frame": 1204, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.961622, "mean_abs": 0.235042, "relative": 0.840305 }, { "name": "hand_right_joints", "l2": 9.645703, "mean_abs": 0.242466, "relative": 0.760339 }, { "name": "body_joints", "l2": 6.340414, "mean_abs": 0.095617, "relative": 0.856395 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.122933, "mean_abs": 0.014985, "relative": 0.290105 }, { "name": "camera_rotation_matrix", "l2": 3.004349, "mean_abs": 0.214739, "relative": 0.922587 }, { "name": "imu_accel_gyro", "l2": 16.998035, "mean_abs": 1.030359, "relative": 0.917103 }, { "name": "depth_confidence", "l2": 11.104833, "mean_abs": 0.203504, "relative": 0.000369 }, { "name": "video_fisheye_cam0", "l2": 5.088027, "mean_abs": 0.106318, "relative": 0.921333 }, { "name": "video_fisheye_cam1", "l2": 5.718548, "mean_abs": 0.122097, "relative": 0.944942 }, { "name": "video_fisheye_cam2", "l2": 5.809278, "mean_abs": 0.120207, "relative": 0.927365 }, { "name": "video_fisheye_cam3", "l2": 5.497635, "mean_abs": 0.113723, "relative": 0.920513 }, { "name": "video_stereo_left", "l2": 5.477689, "mean_abs": 0.122475, "relative": 0.890135 }, { "name": "video_stereo_right", "l2": 6.675497, "mean_abs": 0.151255, "relative": 0.941105 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.03991, "mean_abs": 0.081228, "relative": 0.730053 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 240, "start_frame": 1200, "end_frame": 1219, "center_frame": 1209, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.989251, "mean_abs": 0.23234, "relative": 0.842636 }, { "name": "hand_right_joints", "l2": 9.601192, "mean_abs": 0.235197, "relative": 0.756831 }, { "name": "body_joints", "l2": 6.192675, "mean_abs": 0.092327, "relative": 0.83644 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.070949, "mean_abs": 0.009911, "relative": 0.167429 }, { "name": "camera_rotation_matrix", "l2": 3.004244, "mean_abs": 0.215091, "relative": 0.922555 }, { "name": "imu_accel_gyro", "l2": 17.08633, "mean_abs": 1.029108, "relative": 0.921866 }, { "name": "depth_confidence", "l2": 11.222393, "mean_abs": 0.204348, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.074744, "mean_abs": 0.105845, "relative": 0.918928 }, { "name": "video_fisheye_cam1", "l2": 5.715842, "mean_abs": 0.121243, "relative": 0.944495 }, { "name": "video_fisheye_cam2", "l2": 5.806008, "mean_abs": 0.11974, "relative": 0.926843 }, { "name": "video_fisheye_cam3", "l2": 5.534501, "mean_abs": 0.114219, "relative": 0.926685 }, { "name": "video_stereo_left", "l2": 5.478685, "mean_abs": 0.122659, "relative": 0.890297 }, { "name": "video_stereo_right", "l2": 6.633114, "mean_abs": 0.149871, "relative": 0.93513 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.073838, "mean_abs": 0.084129, "relative": 0.742196 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 241, "start_frame": 1205, "end_frame": 1224, "center_frame": 1214, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.993603, "mean_abs": 0.230474, "relative": 0.843003 }, { "name": "hand_right_joints", "l2": 9.590191, "mean_abs": 0.234392, "relative": 0.755964 }, { "name": "body_joints", "l2": 6.156034, "mean_abs": 0.092624, "relative": 0.831491 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.046004, "mean_abs": 0.007175, "relative": 0.108563 }, { "name": "camera_rotation_matrix", "l2": 3.000525, "mean_abs": 0.216902, "relative": 0.921412 }, { "name": "imu_accel_gyro", "l2": 17.036577, "mean_abs": 1.023687, "relative": 0.919182 }, { "name": "depth_confidence", "l2": 11.366081, "mean_abs": 0.20668, "relative": 0.000377 }, { "name": "video_fisheye_cam0", "l2": 5.072291, "mean_abs": 0.106214, "relative": 0.918483 }, { "name": "video_fisheye_cam1", "l2": 5.694597, "mean_abs": 0.12111, "relative": 0.940984 }, { "name": "video_fisheye_cam2", "l2": 5.796963, "mean_abs": 0.120162, "relative": 0.925399 }, { "name": "video_fisheye_cam3", "l2": 5.55533, "mean_abs": 0.114175, "relative": 0.930173 }, { "name": "video_stereo_left", "l2": 5.476946, "mean_abs": 0.124281, "relative": 0.890015 }, { "name": "video_stereo_right", "l2": 6.605036, "mean_abs": 0.150321, "relative": 0.931171 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.078607, "mean_abs": 0.08696, "relative": 0.743902 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 242, "start_frame": 1210, "end_frame": 1229, "center_frame": 1219, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.920094, "mean_abs": 0.228306, "relative": 0.836802 }, { "name": "hand_right_joints", "l2": 9.570315, "mean_abs": 0.232809, "relative": 0.754397 }, { "name": "body_joints", "l2": 6.148284, "mean_abs": 0.092421, "relative": 0.830444 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.031676, "mean_abs": 0.004754, "relative": 0.074752 }, { "name": "camera_rotation_matrix", "l2": 3.001461, "mean_abs": 0.216034, "relative": 0.9217 }, { "name": "imu_accel_gyro", "l2": 16.971693, "mean_abs": 1.00043, "relative": 0.915681 }, { "name": "depth_confidence", "l2": 11.421677, "mean_abs": 0.205795, "relative": 0.000379 }, { "name": "video_fisheye_cam0", "l2": 5.073837, "mean_abs": 0.105998, "relative": 0.918763 }, { "name": "video_fisheye_cam1", "l2": 5.674129, "mean_abs": 0.120319, "relative": 0.937602 }, { "name": "video_fisheye_cam2", "l2": 5.781033, "mean_abs": 0.119396, "relative": 0.922856 }, { "name": "video_fisheye_cam3", "l2": 5.576032, "mean_abs": 0.113914, "relative": 0.933639 }, { "name": "video_stereo_left", "l2": 5.448037, "mean_abs": 0.122304, "relative": 0.885317 }, { "name": "video_stereo_right", "l2": 6.555053, "mean_abs": 0.147303, "relative": 0.924125 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.033737, "mean_abs": 0.088947, "relative": 0.727844 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 243, "start_frame": 1215, "end_frame": 1234, "center_frame": 1224, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.900322, "mean_abs": 0.225054, "relative": 0.835134 }, { "name": "hand_right_joints", "l2": 9.472392, "mean_abs": 0.231007, "relative": 0.746678 }, { "name": "body_joints", "l2": 6.134795, "mean_abs": 0.090982, "relative": 0.828622 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028502, "mean_abs": 0.003883, "relative": 0.067261 }, { "name": "camera_rotation_matrix", "l2": 3.004007, "mean_abs": 0.214877, "relative": 0.922482 }, { "name": "imu_accel_gyro", "l2": 17.145679, "mean_abs": 1.070545, "relative": 0.925068 }, { "name": "depth_confidence", "l2": 11.474575, "mean_abs": 0.205506, "relative": 0.000381 }, { "name": "video_fisheye_cam0", "l2": 5.080873, "mean_abs": 0.10594, "relative": 0.920038 }, { "name": "video_fisheye_cam1", "l2": 5.673103, "mean_abs": 0.12005, "relative": 0.937432 }, { "name": "video_fisheye_cam2", "l2": 5.771441, "mean_abs": 0.118955, "relative": 0.921325 }, { "name": "video_fisheye_cam3", "l2": 5.598414, "mean_abs": 0.113947, "relative": 0.937387 }, { "name": "video_stereo_left", "l2": 5.431761, "mean_abs": 0.121495, "relative": 0.882672 }, { "name": "video_stereo_right", "l2": 6.564355, "mean_abs": 0.146467, "relative": 0.925436 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.071368, "mean_abs": 0.088924, "relative": 0.741311 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 244, "start_frame": 1220, "end_frame": 1239, "center_frame": 1229, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.913327, "mean_abs": 0.226376, "relative": 0.836231 }, { "name": "hand_right_joints", "l2": 9.426256, "mean_abs": 0.22953, "relative": 0.743041 }, { "name": "body_joints", "l2": 6.12144, "mean_abs": 0.089643, "relative": 0.826818 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015418, "mean_abs": 0.002438, "relative": 0.036385 }, { "name": "camera_rotation_matrix", "l2": 3.001101, "mean_abs": 0.212578, "relative": 0.921589 }, { "name": "imu_accel_gyro", "l2": 17.180458, "mean_abs": 1.07481, "relative": 0.926945 }, { "name": "depth_confidence", "l2": 11.455979, "mean_abs": 0.20548, "relative": 0.00038 }, { "name": "video_fisheye_cam0", "l2": 5.090227, "mean_abs": 0.105729, "relative": 0.921731 }, { "name": "video_fisheye_cam1", "l2": 5.639299, "mean_abs": 0.119166, "relative": 0.931847 }, { "name": "video_fisheye_cam2", "l2": 5.765639, "mean_abs": 0.118191, "relative": 0.920398 }, { "name": "video_fisheye_cam3", "l2": 5.608906, "mean_abs": 0.113552, "relative": 0.939144 }, { "name": "video_stereo_left", "l2": 5.436602, "mean_abs": 0.12161, "relative": 0.883459 }, { "name": "video_stereo_right", "l2": 6.552516, "mean_abs": 0.146012, "relative": 0.923767 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.13219, "mean_abs": 0.091805, "relative": 0.763079 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010954, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 245, "start_frame": 1225, "end_frame": 1244, "center_frame": 1234, "action": "Hold coffee carafe", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.873907, "mean_abs": 0.227507, "relative": 0.832906 }, { "name": "hand_right_joints", "l2": 9.370394, "mean_abs": 0.230872, "relative": 0.738638 }, { "name": "body_joints", "l2": 6.138898, "mean_abs": 0.090494, "relative": 0.829176 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019781, "mean_abs": 0.002932, "relative": 0.04668 }, { "name": "camera_rotation_matrix", "l2": 3.005391, "mean_abs": 0.214792, "relative": 0.922907 }, { "name": "imu_accel_gyro", "l2": 17.185305, "mean_abs": 1.070928, "relative": 0.927206 }, { "name": "depth_confidence", "l2": 11.387598, "mean_abs": 0.206313, "relative": 0.000378 }, { "name": "video_fisheye_cam0", "l2": 5.100819, "mean_abs": 0.106497, "relative": 0.923649 }, { "name": "video_fisheye_cam1", "l2": 5.601793, "mean_abs": 0.118679, "relative": 0.925649 }, { "name": "video_fisheye_cam2", "l2": 5.776646, "mean_abs": 0.119304, "relative": 0.922156 }, { "name": "video_fisheye_cam3", "l2": 5.609267, "mean_abs": 0.113925, "relative": 0.939204 }, { "name": "video_stereo_left", "l2": 5.481995, "mean_abs": 0.124063, "relative": 0.890835 }, { "name": "video_stereo_right", "l2": 6.550713, "mean_abs": 0.147429, "relative": 0.923513 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.272375, "mean_abs": 0.103212, "relative": 0.813249 }, { "name": "caption_objects_interaction_text", "l2": 1.86401, "mean_abs": 0.01415, "relative": 0.855908 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 246, "start_frame": 1230, "end_frame": 1249, "center_frame": 1239, "action": "", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.878204, "mean_abs": 0.22706, "relative": 0.833269 }, { "name": "hand_right_joints", "l2": 9.319416, "mean_abs": 0.228805, "relative": 0.734619 }, { "name": "body_joints", "l2": 6.149366, "mean_abs": 0.091011, "relative": 0.83059 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019246, "mean_abs": 0.002985, "relative": 0.045417 }, { "name": "camera_rotation_matrix", "l2": 3.003084, "mean_abs": 0.21445, "relative": 0.922198 }, { "name": "imu_accel_gyro", "l2": 17.223148, "mean_abs": 1.075431, "relative": 0.929248 }, { "name": "depth_confidence", "l2": 11.281439, "mean_abs": 0.204009, "relative": 0.000374 }, { "name": "video_fisheye_cam0", "l2": 5.116315, "mean_abs": 0.106832, "relative": 0.926455 }, { "name": "video_fisheye_cam1", "l2": 5.587855, "mean_abs": 0.118384, "relative": 0.923346 }, { "name": "video_fisheye_cam2", "l2": 5.785783, "mean_abs": 0.119631, "relative": 0.923614 }, { "name": "video_fisheye_cam3", "l2": 5.615963, "mean_abs": 0.11428, "relative": 0.940325 }, { "name": "video_stereo_left", "l2": 5.528937, "mean_abs": 0.124755, "relative": 0.898463 }, { "name": "video_stereo_right", "l2": 6.591396, "mean_abs": 0.148099, "relative": 0.929249 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.156918, "mean_abs": 0.091998, "relative": 0.771928 }, { "name": "caption_objects_interaction_text", "l2": 1.86401, "mean_abs": 0.014143, "relative": 0.855908 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 247, "start_frame": 1235, "end_frame": 1254, "center_frame": 1244, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "coffee scale", "coffee dripper", "white mug", "bottle", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.877585, "mean_abs": 0.226833, "relative": 0.833216 }, { "name": "hand_right_joints", "l2": 9.268731, "mean_abs": 0.232423, "relative": 0.730624 }, { "name": "body_joints", "l2": 6.15953, "mean_abs": 0.091811, "relative": 0.831963 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081689, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01913, "mean_abs": 0.003016, "relative": 0.045145 }, { "name": "camera_rotation_matrix", "l2": 3.001033, "mean_abs": 0.213728, "relative": 0.921569 }, { "name": "imu_accel_gyro", "l2": 17.081858, "mean_abs": 1.018141, "relative": 0.921625 }, { "name": "depth_confidence", "l2": 11.228395, "mean_abs": 0.202511, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.117476, "mean_abs": 0.106844, "relative": 0.926665 }, { "name": "video_fisheye_cam1", "l2": 5.571005, "mean_abs": 0.117865, "relative": 0.920562 }, { "name": "video_fisheye_cam2", "l2": 5.788176, "mean_abs": 0.119625, "relative": 0.923996 }, { "name": "video_fisheye_cam3", "l2": 5.616743, "mean_abs": 0.114167, "relative": 0.940456 }, { "name": "video_stereo_left", "l2": 5.543108, "mean_abs": 0.124588, "relative": 0.900766 }, { "name": "video_stereo_right", "l2": 6.605268, "mean_abs": 0.146995, "relative": 0.931204 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.163918, "mean_abs": 0.091673, "relative": 0.774434 }, { "name": "caption_objects_interaction_text", "l2": 1.86401, "mean_abs": 0.013822, "relative": 0.855908 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 248, "start_frame": 1240, "end_frame": 1259, "center_frame": 1249, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.781304, "mean_abs": 0.224196, "relative": 0.825095 }, { "name": "hand_right_joints", "l2": 9.251522, "mean_abs": 0.240123, "relative": 0.729267 }, { "name": "body_joints", "l2": 6.267019, "mean_abs": 0.096414, "relative": 0.846481 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08169, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.032321, "mean_abs": 0.005069, "relative": 0.076274 }, { "name": "camera_rotation_matrix", "l2": 3.004155, "mean_abs": 0.212326, "relative": 0.922527 }, { "name": "imu_accel_gyro", "l2": 17.107771, "mean_abs": 1.062743, "relative": 0.923023 }, { "name": "depth_confidence", "l2": 11.141589, "mean_abs": 0.201999, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.120018, "mean_abs": 0.106435, "relative": 0.927126 }, { "name": "video_fisheye_cam1", "l2": 5.522748, "mean_abs": 0.117308, "relative": 0.912587 }, { "name": "video_fisheye_cam2", "l2": 5.767471, "mean_abs": 0.119194, "relative": 0.920691 }, { "name": "video_fisheye_cam3", "l2": 5.613427, "mean_abs": 0.11405, "relative": 0.939901 }, { "name": "video_stereo_left", "l2": 5.522836, "mean_abs": 0.123631, "relative": 0.897472 }, { "name": "video_stereo_right", "l2": 6.51829, "mean_abs": 0.146008, "relative": 0.918942 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.095999, "mean_abs": 0.086355, "relative": 0.750126 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008984, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 249, "start_frame": 1245, "end_frame": 1264, "center_frame": 1254, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.789751, "mean_abs": 0.22625, "relative": 0.825807 }, { "name": "hand_right_joints", "l2": 9.691106, "mean_abs": 0.260301, "relative": 0.763918 }, { "name": "body_joints", "l2": 6.306123, "mean_abs": 0.100739, "relative": 0.851763 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081689, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.097579, "mean_abs": 0.013398, "relative": 0.230272 }, { "name": "camera_rotation_matrix", "l2": 3.012413, "mean_abs": 0.217665, "relative": 0.925063 }, { "name": "imu_accel_gyro", "l2": 17.014406, "mean_abs": 1.077878, "relative": 0.917986 }, { "name": "depth_confidence", "l2": 10.922968, "mean_abs": 0.203012, "relative": 0.000363 }, { "name": "video_fisheye_cam0", "l2": 5.135852, "mean_abs": 0.108351, "relative": 0.929993 }, { "name": "video_fisheye_cam1", "l2": 5.462102, "mean_abs": 0.119632, "relative": 0.902566 }, { "name": "video_fisheye_cam2", "l2": 5.720687, "mean_abs": 0.121656, "relative": 0.913223 }, { "name": "video_fisheye_cam3", "l2": 5.604638, "mean_abs": 0.115566, "relative": 0.938429 }, { "name": "video_stereo_left", "l2": 5.468395, "mean_abs": 0.127636, "relative": 0.888625 }, { "name": "video_stereo_right", "l2": 6.398682, "mean_abs": 0.149374, "relative": 0.90208 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.216644, "mean_abs": 0.088685, "relative": 0.793304 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008984, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 250, "start_frame": 1250, "end_frame": 1269, "center_frame": 1259, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.778105, "mean_abs": 0.225662, "relative": 0.824825 }, { "name": "hand_right_joints", "l2": 10.326035, "mean_abs": 0.282029, "relative": 0.813968 }, { "name": "body_joints", "l2": 6.429362, "mean_abs": 0.106214, "relative": 0.868409 }, { "name": "body_contacts", "l2": 3.463825, "mean_abs": 0.081734, "relative": 0.999545 }, { "name": "camera_translation", "l2": 0.204423, "mean_abs": 0.028207, "relative": 0.48241 }, { "name": "camera_rotation_matrix", "l2": 3.027385, "mean_abs": 0.22037, "relative": 0.929661 }, { "name": "imu_accel_gyro", "l2": 17.010977, "mean_abs": 1.073463, "relative": 0.917801 }, { "name": "depth_confidence", "l2": 10.81389, "mean_abs": 0.206251, "relative": 0.000359 }, { "name": "video_fisheye_cam0", "l2": 5.135907, "mean_abs": 0.109813, "relative": 0.930003 }, { "name": "video_fisheye_cam1", "l2": 5.393926, "mean_abs": 0.120994, "relative": 0.891301 }, { "name": "video_fisheye_cam2", "l2": 5.636761, "mean_abs": 0.122009, "relative": 0.899825 }, { "name": "video_fisheye_cam3", "l2": 5.584908, "mean_abs": 0.116553, "relative": 0.935125 }, { "name": "video_stereo_left", "l2": 5.370479, "mean_abs": 0.128405, "relative": 0.872714 }, { "name": "video_stereo_right", "l2": 6.286539, "mean_abs": 0.150797, "relative": 0.88627 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.193057, "mean_abs": 0.087283, "relative": 0.784862 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008984, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 251, "start_frame": 1255, "end_frame": 1274, "center_frame": 1264, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.707159, "mean_abs": 0.222737, "relative": 0.81884 }, { "name": "hand_right_joints", "l2": 10.859331, "mean_abs": 0.299879, "relative": 0.856006 }, { "name": "body_joints", "l2": 6.652724, "mean_abs": 0.1109, "relative": 0.898578 }, { "name": "body_contacts", "l2": 3.46383, "mean_abs": 0.081707, "relative": 0.999546 }, { "name": "camera_translation", "l2": 0.276262, "mean_abs": 0.039495, "relative": 0.65194 }, { "name": "camera_rotation_matrix", "l2": 3.022627, "mean_abs": 0.21644, "relative": 0.9282 }, { "name": "imu_accel_gyro", "l2": 17.172438, "mean_abs": 1.170386, "relative": 0.926512 }, { "name": "depth_confidence", "l2": 10.592231, "mean_abs": 0.203698, "relative": 0.000352 }, { "name": "video_fisheye_cam0", "l2": 5.136314, "mean_abs": 0.109706, "relative": 0.930077 }, { "name": "video_fisheye_cam1", "l2": 5.331584, "mean_abs": 0.120336, "relative": 0.880999 }, { "name": "video_fisheye_cam2", "l2": 5.591963, "mean_abs": 0.12103, "relative": 0.892674 }, { "name": "video_fisheye_cam3", "l2": 5.595074, "mean_abs": 0.117589, "relative": 0.936828 }, { "name": "video_stereo_left", "l2": 5.296199, "mean_abs": 0.126298, "relative": 0.860643 }, { "name": "video_stereo_right", "l2": 6.174342, "mean_abs": 0.149098, "relative": 0.870453 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.265395, "mean_abs": 0.090271, "relative": 0.810751 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008984, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 252, "start_frame": 1260, "end_frame": 1279, "center_frame": 1269, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.636602, "mean_abs": 0.214885, "relative": 0.812889 }, { "name": "hand_right_joints", "l2": 11.370996, "mean_abs": 0.30873, "relative": 0.896339 }, { "name": "body_joints", "l2": 6.680807, "mean_abs": 0.110582, "relative": 0.902371 }, { "name": "body_contacts", "l2": 3.46383, "mean_abs": 0.081705, "relative": 0.999546 }, { "name": "camera_translation", "l2": 0.284476, "mean_abs": 0.04177, "relative": 0.671322 }, { "name": "camera_rotation_matrix", "l2": 3.00944, "mean_abs": 0.205753, "relative": 0.92415 }, { "name": "imu_accel_gyro", "l2": 17.277962, "mean_abs": 1.224113, "relative": 0.932206 }, { "name": "depth_confidence", "l2": 10.235974, "mean_abs": 0.197374, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.134686, "mean_abs": 0.108624, "relative": 0.929782 }, { "name": "video_fisheye_cam1", "l2": 5.206783, "mean_abs": 0.115437, "relative": 0.860377 }, { "name": "video_fisheye_cam2", "l2": 5.510089, "mean_abs": 0.117116, "relative": 0.879604 }, { "name": "video_fisheye_cam3", "l2": 5.596572, "mean_abs": 0.117473, "relative": 0.937079 }, { "name": "video_stereo_left", "l2": 5.161996, "mean_abs": 0.120775, "relative": 0.838835 }, { "name": "video_stereo_right", "l2": 5.962174, "mean_abs": 0.142092, "relative": 0.840541 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.268492, "mean_abs": 0.094451, "relative": 0.811859 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008984, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 253, "start_frame": 1265, "end_frame": 1284, "center_frame": 1274, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.541139, "mean_abs": 0.213912, "relative": 0.804836 }, { "name": "hand_right_joints", "l2": 11.909872, "mean_abs": 0.319415, "relative": 0.938817 }, { "name": "body_joints", "l2": 6.612441, "mean_abs": 0.109673, "relative": 0.893137 }, { "name": "body_contacts", "l2": 3.46383, "mean_abs": 0.081705, "relative": 0.999546 }, { "name": "camera_translation", "l2": 0.231373, "mean_abs": 0.034271, "relative": 0.546008 }, { "name": "camera_rotation_matrix", "l2": 3.01341, "mean_abs": 0.199862, "relative": 0.925369 }, { "name": "imu_accel_gyro", "l2": 17.297113, "mean_abs": 1.219589, "relative": 0.933239 }, { "name": "depth_confidence", "l2": 9.994017, "mean_abs": 0.192695, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.13324, "mean_abs": 0.108228, "relative": 0.92952 }, { "name": "video_fisheye_cam1", "l2": 5.093194, "mean_abs": 0.112226, "relative": 0.841607 }, { "name": "video_fisheye_cam2", "l2": 5.429927, "mean_abs": 0.113924, "relative": 0.866807 }, { "name": "video_fisheye_cam3", "l2": 5.579472, "mean_abs": 0.117826, "relative": 0.934215 }, { "name": "video_stereo_left", "l2": 5.080521, "mean_abs": 0.118199, "relative": 0.825595 }, { "name": "video_stereo_right", "l2": 5.819896, "mean_abs": 0.137798, "relative": 0.820483 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.160076, "mean_abs": 0.090196, "relative": 0.773059 }, { "name": "caption_objects_interaction_text", "l2": 1.952101, "mean_abs": 0.014947, "relative": 0.896357 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 254, "start_frame": 1270, "end_frame": 1289, "center_frame": 1279, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.478716, "mean_abs": 0.215693, "relative": 0.79957 }, { "name": "hand_right_joints", "l2": 12.373788, "mean_abs": 0.325223, "relative": 0.975385 }, { "name": "body_joints", "l2": 6.654895, "mean_abs": 0.109441, "relative": 0.898871 }, { "name": "body_contacts", "l2": 3.46399, "mean_abs": 0.08173, "relative": 0.999593 }, { "name": "camera_translation", "l2": 0.139914, "mean_abs": 0.019112, "relative": 0.330178 }, { "name": "camera_rotation_matrix", "l2": 3.015712, "mean_abs": 0.199605, "relative": 0.926076 }, { "name": "imu_accel_gyro", "l2": 17.372982, "mean_abs": 1.197799, "relative": 0.937332 }, { "name": "depth_confidence", "l2": 9.6658, "mean_abs": 0.187565, "relative": 0.000321 }, { "name": "video_fisheye_cam0", "l2": 5.134687, "mean_abs": 0.108618, "relative": 0.929782 }, { "name": "video_fisheye_cam1", "l2": 5.0618, "mean_abs": 0.111414, "relative": 0.83642 }, { "name": "video_fisheye_cam2", "l2": 5.446107, "mean_abs": 0.11439, "relative": 0.86939 }, { "name": "video_fisheye_cam3", "l2": 5.54211, "mean_abs": 0.116916, "relative": 0.92796 }, { "name": "video_stereo_left", "l2": 5.093757, "mean_abs": 0.117763, "relative": 0.827746 }, { "name": "video_stereo_right", "l2": 5.815905, "mean_abs": 0.137435, "relative": 0.819921 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.225611, "mean_abs": 0.091521, "relative": 0.796513 }, { "name": "caption_objects_interaction_text", "l2": 1.952101, "mean_abs": 0.015251, "relative": 0.896357 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 255, "start_frame": 1275, "end_frame": 1294, "center_frame": 1284, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.434993, "mean_abs": 0.216072, "relative": 0.795882 }, { "name": "hand_right_joints", "l2": 12.378541, "mean_abs": 0.326, "relative": 0.97576 }, { "name": "body_joints", "l2": 6.615876, "mean_abs": 0.10589, "relative": 0.893601 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.065846, "mean_abs": 0.010184, "relative": 0.155387 }, { "name": "camera_rotation_matrix", "l2": 3.002681, "mean_abs": 0.188574, "relative": 0.922074 }, { "name": "imu_accel_gyro", "l2": 17.432789, "mean_abs": 1.152169, "relative": 0.940559 }, { "name": "depth_confidence", "l2": 9.362717, "mean_abs": 0.177932, "relative": 0.000311 }, { "name": "video_fisheye_cam0", "l2": 5.138528, "mean_abs": 0.106984, "relative": 0.930478 }, { "name": "video_fisheye_cam1", "l2": 5.023563, "mean_abs": 0.108495, "relative": 0.830101 }, { "name": "video_fisheye_cam2", "l2": 5.475766, "mean_abs": 0.112615, "relative": 0.874125 }, { "name": "video_fisheye_cam3", "l2": 5.49253, "mean_abs": 0.11367, "relative": 0.919658 }, { "name": "video_stereo_left", "l2": 5.100732, "mean_abs": 0.114748, "relative": 0.828879 }, { "name": "video_stereo_right", "l2": 5.820128, "mean_abs": 0.134492, "relative": 0.820516 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.270416, "mean_abs": 0.095879, "relative": 0.812548 }, { "name": "caption_objects_interaction_text", "l2": 1.952101, "mean_abs": 0.015123, "relative": 0.896357 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 256, "start_frame": 1280, "end_frame": 1299, "center_frame": 1289, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.449207, "mean_abs": 0.216758, "relative": 0.797081 }, { "name": "hand_right_joints", "l2": 12.375159, "mean_abs": 0.32207, "relative": 0.975494 }, { "name": "body_joints", "l2": 6.482892, "mean_abs": 0.106914, "relative": 0.875639 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.05526, "mean_abs": 0.007825, "relative": 0.130407 }, { "name": "camera_rotation_matrix", "l2": 3.000624, "mean_abs": 0.183369, "relative": 0.921443 }, { "name": "imu_accel_gyro", "l2": 17.227127, "mean_abs": 1.104522, "relative": 0.929463 }, { "name": "depth_confidence", "l2": 9.243289, "mean_abs": 0.173479, "relative": 0.000307 }, { "name": "video_fisheye_cam0", "l2": 5.147374, "mean_abs": 0.106266, "relative": 0.932079 }, { "name": "video_fisheye_cam1", "l2": 5.085196, "mean_abs": 0.109696, "relative": 0.840286 }, { "name": "video_fisheye_cam2", "l2": 5.5196, "mean_abs": 0.112523, "relative": 0.881122 }, { "name": "video_fisheye_cam3", "l2": 5.509727, "mean_abs": 0.113185, "relative": 0.922537 }, { "name": "video_stereo_left", "l2": 5.141102, "mean_abs": 0.113747, "relative": 0.835439 }, { "name": "video_stereo_right", "l2": 5.889125, "mean_abs": 0.134687, "relative": 0.830243 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.080253, "mean_abs": 0.082971, "relative": 0.744491 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 257, "start_frame": 1285, "end_frame": 1304, "center_frame": 1294, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.514488, "mean_abs": 0.222633, "relative": 0.802588 }, { "name": "hand_right_joints", "l2": 12.387111, "mean_abs": 0.322063, "relative": 0.976436 }, { "name": "body_joints", "l2": 6.43018, "mean_abs": 0.109438, "relative": 0.868519 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.078878, "mean_abs": 0.011378, "relative": 0.186141 }, { "name": "camera_rotation_matrix", "l2": 3.000085, "mean_abs": 0.182611, "relative": 0.921277 }, { "name": "imu_accel_gyro", "l2": 17.1985, "mean_abs": 1.10197, "relative": 0.927918 }, { "name": "depth_confidence", "l2": 9.298081, "mean_abs": 0.175293, "relative": 0.000309 }, { "name": "video_fisheye_cam0", "l2": 5.150323, "mean_abs": 0.106586, "relative": 0.932613 }, { "name": "video_fisheye_cam1", "l2": 5.187929, "mean_abs": 0.111551, "relative": 0.857261 }, { "name": "video_fisheye_cam2", "l2": 5.575107, "mean_abs": 0.114074, "relative": 0.889983 }, { "name": "video_fisheye_cam3", "l2": 5.555253, "mean_abs": 0.113996, "relative": 0.93016 }, { "name": "video_stereo_left", "l2": 5.227739, "mean_abs": 0.115791, "relative": 0.849518 }, { "name": "video_stereo_right", "l2": 6.037361, "mean_abs": 0.137778, "relative": 0.851141 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.04043, "mean_abs": 0.078914, "relative": 0.730239 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 258, "start_frame": 1290, "end_frame": 1309, "center_frame": 1299, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.586159, "mean_abs": 0.226828, "relative": 0.808633 }, { "name": "hand_right_joints", "l2": 12.371486, "mean_abs": 0.320068, "relative": 0.975204 }, { "name": "body_joints", "l2": 6.337898, "mean_abs": 0.108076, "relative": 0.856055 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.097323, "mean_abs": 0.01405, "relative": 0.229669 }, { "name": "camera_rotation_matrix", "l2": 3.00018, "mean_abs": 0.184352, "relative": 0.921307 }, { "name": "imu_accel_gyro", "l2": 17.202736, "mean_abs": 1.119614, "relative": 0.928147 }, { "name": "depth_confidence", "l2": 9.349297, "mean_abs": 0.176399, "relative": 0.00031 }, { "name": "video_fisheye_cam0", "l2": 5.157839, "mean_abs": 0.106869, "relative": 0.933974 }, { "name": "video_fisheye_cam1", "l2": 5.253092, "mean_abs": 0.112982, "relative": 0.868029 }, { "name": "video_fisheye_cam2", "l2": 5.612454, "mean_abs": 0.114894, "relative": 0.895945 }, { "name": "video_fisheye_cam3", "l2": 5.57327, "mean_abs": 0.114483, "relative": 0.933177 }, { "name": "video_stereo_left", "l2": 5.262705, "mean_abs": 0.116497, "relative": 0.8552 }, { "name": "video_stereo_right", "l2": 6.094127, "mean_abs": 0.139166, "relative": 0.859144 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.023164, "mean_abs": 0.080651, "relative": 0.72406 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 259, "start_frame": 1295, "end_frame": 1314, "center_frame": 1304, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.660929, "mean_abs": 0.227753, "relative": 0.814941 }, { "name": "hand_right_joints", "l2": 12.385801, "mean_abs": 0.31166, "relative": 0.976332 }, { "name": "body_joints", "l2": 6.042235, "mean_abs": 0.100056, "relative": 0.81612 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.078569, "mean_abs": 0.011129, "relative": 0.185413 }, { "name": "camera_rotation_matrix", "l2": 3.001487, "mean_abs": 0.186805, "relative": 0.921708 }, { "name": "imu_accel_gyro", "l2": 17.062765, "mean_abs": 1.055544, "relative": 0.920595 }, { "name": "depth_confidence", "l2": 9.413695, "mean_abs": 0.176933, "relative": 0.000312 }, { "name": "video_fisheye_cam0", "l2": 5.163112, "mean_abs": 0.107028, "relative": 0.934929 }, { "name": "video_fisheye_cam1", "l2": 5.310174, "mean_abs": 0.113717, "relative": 0.877461 }, { "name": "video_fisheye_cam2", "l2": 5.643098, "mean_abs": 0.115429, "relative": 0.900837 }, { "name": "video_fisheye_cam3", "l2": 5.609699, "mean_abs": 0.114774, "relative": 0.939277 }, { "name": "video_stereo_left", "l2": 5.289921, "mean_abs": 0.116638, "relative": 0.859623 }, { "name": "video_stereo_right", "l2": 6.151204, "mean_abs": 0.139799, "relative": 0.867191 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.090146, "mean_abs": 0.085001, "relative": 0.748032 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 260, "start_frame": 1300, "end_frame": 1319, "center_frame": 1309, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.73984, "mean_abs": 0.225506, "relative": 0.821597 }, { "name": "hand_right_joints", "l2": 12.46409, "mean_abs": 0.31091, "relative": 0.982504 }, { "name": "body_joints", "l2": 5.917259, "mean_abs": 0.094922, "relative": 0.799239 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.072738, "mean_abs": 0.010833, "relative": 0.171652 }, { "name": "camera_rotation_matrix", "l2": 3.002048, "mean_abs": 0.190028, "relative": 0.92188 }, { "name": "imu_accel_gyro", "l2": 17.187424, "mean_abs": 1.077994, "relative": 0.927321 }, { "name": "depth_confidence", "l2": 9.547033, "mean_abs": 0.179693, "relative": 0.000317 }, { "name": "video_fisheye_cam0", "l2": 5.16256, "mean_abs": 0.107189, "relative": 0.934829 }, { "name": "video_fisheye_cam1", "l2": 5.365513, "mean_abs": 0.114705, "relative": 0.886606 }, { "name": "video_fisheye_cam2", "l2": 5.670115, "mean_abs": 0.115624, "relative": 0.905149 }, { "name": "video_fisheye_cam3", "l2": 5.636624, "mean_abs": 0.116146, "relative": 0.943785 }, { "name": "video_stereo_left", "l2": 5.321648, "mean_abs": 0.116918, "relative": 0.864778 }, { "name": "video_stereo_right", "l2": 6.197265, "mean_abs": 0.140104, "relative": 0.873684 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.169107, "mean_abs": 0.084945, "relative": 0.776291 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 261, "start_frame": 1305, "end_frame": 1324, "center_frame": 1314, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.785208, "mean_abs": 0.22351, "relative": 0.825424 }, { "name": "hand_right_joints", "l2": 12.396891, "mean_abs": 0.306875, "relative": 0.977207 }, { "name": "body_joints", "l2": 5.906397, "mean_abs": 0.092981, "relative": 0.797772 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.087536, "mean_abs": 0.012443, "relative": 0.206574 }, { "name": "camera_rotation_matrix", "l2": 3.004757, "mean_abs": 0.194338, "relative": 0.922712 }, { "name": "imu_accel_gyro", "l2": 17.210873, "mean_abs": 1.107766, "relative": 0.928586 }, { "name": "depth_confidence", "l2": 9.725552, "mean_abs": 0.18463, "relative": 0.000323 }, { "name": "video_fisheye_cam0", "l2": 5.156757, "mean_abs": 0.107288, "relative": 0.933778 }, { "name": "video_fisheye_cam1", "l2": 5.424231, "mean_abs": 0.115918, "relative": 0.896308 }, { "name": "video_fisheye_cam2", "l2": 5.676021, "mean_abs": 0.116175, "relative": 0.906092 }, { "name": "video_fisheye_cam3", "l2": 5.649329, "mean_abs": 0.117442, "relative": 0.945912 }, { "name": "video_stereo_left", "l2": 5.331929, "mean_abs": 0.119328, "relative": 0.866449 }, { "name": "video_stereo_right", "l2": 6.225231, "mean_abs": 0.142558, "relative": 0.877627 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.107533, "mean_abs": 0.084514, "relative": 0.754255 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 262, "start_frame": 1310, "end_frame": 1329, "center_frame": 1319, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.841968, "mean_abs": 0.224517, "relative": 0.830212 }, { "name": "hand_right_joints", "l2": 12.310146, "mean_abs": 0.305107, "relative": 0.970369 }, { "name": "body_joints", "l2": 5.918764, "mean_abs": 0.093264, "relative": 0.799443 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.11431, "mean_abs": 0.01519, "relative": 0.269756 }, { "name": "camera_rotation_matrix", "l2": 3.007139, "mean_abs": 0.198824, "relative": 0.923443 }, { "name": "imu_accel_gyro", "l2": 17.18523, "mean_abs": 1.121629, "relative": 0.927202 }, { "name": "depth_confidence", "l2": 9.871449, "mean_abs": 0.187645, "relative": 0.000328 }, { "name": "video_fisheye_cam0", "l2": 5.151885, "mean_abs": 0.107368, "relative": 0.932896 }, { "name": "video_fisheye_cam1", "l2": 5.47876, "mean_abs": 0.117599, "relative": 0.905319 }, { "name": "video_fisheye_cam2", "l2": 5.686017, "mean_abs": 0.117093, "relative": 0.907688 }, { "name": "video_fisheye_cam3", "l2": 5.670909, "mean_abs": 0.117924, "relative": 0.949525 }, { "name": "video_stereo_left", "l2": 5.321738, "mean_abs": 0.120401, "relative": 0.864793 }, { "name": "video_stereo_right", "l2": 6.2415, "mean_abs": 0.144192, "relative": 0.879921 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.156115, "mean_abs": 0.085555, "relative": 0.771641 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 263, "start_frame": 1315, "end_frame": 1334, "center_frame": 1324, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.868116, "mean_abs": 0.225327, "relative": 0.832418 }, { "name": "hand_right_joints", "l2": 12.079427, "mean_abs": 0.301582, "relative": 0.952182 }, { "name": "body_joints", "l2": 5.935419, "mean_abs": 0.093815, "relative": 0.801692 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.128264, "mean_abs": 0.016785, "relative": 0.302686 }, { "name": "camera_rotation_matrix", "l2": 3.009589, "mean_abs": 0.20263, "relative": 0.924196 }, { "name": "imu_accel_gyro", "l2": 17.211487, "mean_abs": 1.133461, "relative": 0.928619 }, { "name": "depth_confidence", "l2": 10.140328, "mean_abs": 0.192974, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.142214, "mean_abs": 0.107079, "relative": 0.931145 }, { "name": "video_fisheye_cam1", "l2": 5.529938, "mean_abs": 0.118796, "relative": 0.913776 }, { "name": "video_fisheye_cam2", "l2": 5.685579, "mean_abs": 0.117438, "relative": 0.907618 }, { "name": "video_fisheye_cam3", "l2": 5.682619, "mean_abs": 0.117737, "relative": 0.951486 }, { "name": "video_stereo_left", "l2": 5.301515, "mean_abs": 0.12052, "relative": 0.861507 }, { "name": "video_stereo_right", "l2": 6.270311, "mean_abs": 0.145812, "relative": 0.883982 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.083425, "mean_abs": 0.080963, "relative": 0.745626 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 264, "start_frame": 1320, "end_frame": 1339, "center_frame": 1329, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.867139, "mean_abs": 0.223948, "relative": 0.832335 }, { "name": "hand_right_joints", "l2": 11.686692, "mean_abs": 0.299218, "relative": 0.921224 }, { "name": "body_joints", "l2": 6.008201, "mean_abs": 0.096126, "relative": 0.811523 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.126131, "mean_abs": 0.017596, "relative": 0.29765 }, { "name": "camera_rotation_matrix", "l2": 3.011553, "mean_abs": 0.204474, "relative": 0.924799 }, { "name": "imu_accel_gyro", "l2": 17.174603, "mean_abs": 1.111582, "relative": 0.926629 }, { "name": "depth_confidence", "l2": 10.453372, "mean_abs": 0.197818, "relative": 0.000347 }, { "name": "video_fisheye_cam0", "l2": 5.121843, "mean_abs": 0.106458, "relative": 0.927456 }, { "name": "video_fisheye_cam1", "l2": 5.564846, "mean_abs": 0.119335, "relative": 0.919544 }, { "name": "video_fisheye_cam2", "l2": 5.683633, "mean_abs": 0.117295, "relative": 0.907308 }, { "name": "video_fisheye_cam3", "l2": 5.644312, "mean_abs": 0.116893, "relative": 0.945072 }, { "name": "video_stereo_left", "l2": 5.301373, "mean_abs": 0.120928, "relative": 0.861484 }, { "name": "video_stereo_right", "l2": 6.281557, "mean_abs": 0.145512, "relative": 0.885568 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.112245, "mean_abs": 0.085468, "relative": 0.755941 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 265, "start_frame": 1325, "end_frame": 1344, "center_frame": 1334, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.867474, "mean_abs": 0.223438, "relative": 0.832363 }, { "name": "hand_right_joints", "l2": 11.307817, "mean_abs": 0.294765, "relative": 0.891358 }, { "name": "body_joints", "l2": 5.98263, "mean_abs": 0.097095, "relative": 0.808069 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.118852, "mean_abs": 0.017386, "relative": 0.280473 }, { "name": "camera_rotation_matrix", "l2": 3.008236, "mean_abs": 0.208509, "relative": 0.923781 }, { "name": "imu_accel_gyro", "l2": 17.094812, "mean_abs": 1.108896, "relative": 0.922324 }, { "name": "depth_confidence", "l2": 10.626255, "mean_abs": 0.200343, "relative": 0.000353 }, { "name": "video_fisheye_cam0", "l2": 5.117187, "mean_abs": 0.10655, "relative": 0.926613 }, { "name": "video_fisheye_cam1", "l2": 5.600273, "mean_abs": 0.119711, "relative": 0.925398 }, { "name": "video_fisheye_cam2", "l2": 5.722514, "mean_abs": 0.118373, "relative": 0.913514 }, { "name": "video_fisheye_cam3", "l2": 5.612263, "mean_abs": 0.116048, "relative": 0.939706 }, { "name": "video_stereo_left", "l2": 5.356602, "mean_abs": 0.122778, "relative": 0.870458 }, { "name": "video_stereo_right", "l2": 6.324555, "mean_abs": 0.14645, "relative": 0.89163 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.267577, "mean_abs": 0.098206, "relative": 0.811532 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 266, "start_frame": 1330, "end_frame": 1349, "center_frame": 1339, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.845497, "mean_abs": 0.22191, "relative": 0.83051 }, { "name": "hand_right_joints", "l2": 10.402488, "mean_abs": 0.31449, "relative": 0.819994 }, { "name": "body_joints", "l2": 5.89317, "mean_abs": 0.096545, "relative": 0.795986 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.098833, "mean_abs": 0.015043, "relative": 0.233232 }, { "name": "camera_rotation_matrix", "l2": 3.005287, "mean_abs": 0.213389, "relative": 0.922875 }, { "name": "imu_accel_gyro", "l2": 17.139284, "mean_abs": 1.110226, "relative": 0.924723 }, { "name": "depth_confidence", "l2": 10.889594, "mean_abs": 0.203999, "relative": 0.000361 }, { "name": "video_fisheye_cam0", "l2": 5.110808, "mean_abs": 0.107394, "relative": 0.925458 }, { "name": "video_fisheye_cam1", "l2": 5.633189, "mean_abs": 0.121372, "relative": 0.930837 }, { "name": "video_fisheye_cam2", "l2": 5.806594, "mean_abs": 0.12105, "relative": 0.926936 }, { "name": "video_fisheye_cam3", "l2": 5.578787, "mean_abs": 0.115267, "relative": 0.934101 }, { "name": "video_stereo_left", "l2": 5.462272, "mean_abs": 0.126073, "relative": 0.88763 }, { "name": "video_stereo_right", "l2": 6.451666, "mean_abs": 0.150913, "relative": 0.90955 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.227786, "mean_abs": 0.097056, "relative": 0.797291 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 267, "start_frame": 1335, "end_frame": 1354, "center_frame": 1344, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.807753, "mean_abs": 0.220335, "relative": 0.827326 }, { "name": "hand_right_joints", "l2": 9.649081, "mean_abs": 0.297695, "relative": 0.760606 }, { "name": "body_joints", "l2": 5.84012, "mean_abs": 0.095054, "relative": 0.78882 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.07293, "mean_abs": 0.010938, "relative": 0.172105 }, { "name": "camera_rotation_matrix", "l2": 3.006645, "mean_abs": 0.215967, "relative": 0.923292 }, { "name": "imu_accel_gyro", "l2": 17.115597, "mean_abs": 1.089667, "relative": 0.923445 }, { "name": "depth_confidence", "l2": 11.031257, "mean_abs": 0.205587, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.115026, "mean_abs": 0.107745, "relative": 0.926222 }, { "name": "video_fisheye_cam1", "l2": 5.627151, "mean_abs": 0.121976, "relative": 0.929839 }, { "name": "video_fisheye_cam2", "l2": 5.837128, "mean_abs": 0.121862, "relative": 0.931811 }, { "name": "video_fisheye_cam3", "l2": 5.535225, "mean_abs": 0.114277, "relative": 0.926807 }, { "name": "video_stereo_left", "l2": 5.47819, "mean_abs": 0.125994, "relative": 0.890217 }, { "name": "video_stereo_right", "l2": 6.477101, "mean_abs": 0.151393, "relative": 0.913135 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.476894, "mean_abs": 0.108638, "relative": 0.886443 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 268, "start_frame": 1340, "end_frame": 1359, "center_frame": 1349, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.81062, "mean_abs": 0.221928, "relative": 0.827568 }, { "name": "hand_right_joints", "l2": 8.557318, "mean_abs": 0.247931, "relative": 0.674546 }, { "name": "body_joints", "l2": 5.907347, "mean_abs": 0.096463, "relative": 0.7979 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.067916, "mean_abs": 0.009912, "relative": 0.160272 }, { "name": "camera_rotation_matrix", "l2": 3.012719, "mean_abs": 0.22108, "relative": 0.925157 }, { "name": "imu_accel_gyro", "l2": 17.102577, "mean_abs": 1.054799, "relative": 0.922743 }, { "name": "depth_confidence", "l2": 11.141166, "mean_abs": 0.208158, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.132281, "mean_abs": 0.108638, "relative": 0.929346 }, { "name": "video_fisheye_cam1", "l2": 5.635266, "mean_abs": 0.122792, "relative": 0.93118 }, { "name": "video_fisheye_cam2", "l2": 5.894652, "mean_abs": 0.123836, "relative": 0.940994 }, { "name": "video_fisheye_cam3", "l2": 5.514066, "mean_abs": 0.114384, "relative": 0.923264 }, { "name": "video_stereo_left", "l2": 5.537902, "mean_abs": 0.127965, "relative": 0.89992 }, { "name": "video_stereo_right", "l2": 6.603481, "mean_abs": 0.156375, "relative": 0.930952 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.701795, "mean_abs": 0.132951, "relative": 0.966932 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 269, "start_frame": 1345, "end_frame": 1364, "center_frame": 1354, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.766421, "mean_abs": 0.224818, "relative": 0.823839 }, { "name": "hand_right_joints", "l2": 7.905758, "mean_abs": 0.231831, "relative": 0.623185 }, { "name": "body_joints", "l2": 5.884226, "mean_abs": 0.095535, "relative": 0.794778 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.069863, "mean_abs": 0.010076, "relative": 0.164868 }, { "name": "camera_rotation_matrix", "l2": 3.004916, "mean_abs": 0.218658, "relative": 0.922761 }, { "name": "imu_accel_gyro", "l2": 17.061951, "mean_abs": 1.045317, "relative": 0.920551 }, { "name": "depth_confidence", "l2": 11.237066, "mean_abs": 0.20994, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.124424, "mean_abs": 0.10789, "relative": 0.927924 }, { "name": "video_fisheye_cam1", "l2": 5.62851, "mean_abs": 0.122254, "relative": 0.930064 }, { "name": "video_fisheye_cam2", "l2": 5.941651, "mean_abs": 0.123491, "relative": 0.948496 }, { "name": "video_fisheye_cam3", "l2": 5.533525, "mean_abs": 0.114202, "relative": 0.926522 }, { "name": "video_stereo_left", "l2": 5.54064, "mean_abs": 0.126848, "relative": 0.900365 }, { "name": "video_stereo_right", "l2": 6.734215, "mean_abs": 0.158215, "relative": 0.949383 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.520771, "mean_abs": 0.113055, "relative": 0.902146 }, { "name": "caption_objects_interaction_text", "l2": 1.83535, "mean_abs": 0.013488, "relative": 0.842748 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 270, "start_frame": 1350, "end_frame": 1369, "center_frame": 1359, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.740163, "mean_abs": 0.225098, "relative": 0.821624 }, { "name": "hand_right_joints", "l2": 8.185724, "mean_abs": 0.238898, "relative": 0.645254 }, { "name": "body_joints", "l2": 5.901322, "mean_abs": 0.095968, "relative": 0.797087 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.087351, "mean_abs": 0.011733, "relative": 0.206135 }, { "name": "camera_rotation_matrix", "l2": 3.00386, "mean_abs": 0.220579, "relative": 0.922437 }, { "name": "imu_accel_gyro", "l2": 17.035261, "mean_abs": 1.045781, "relative": 0.919111 }, { "name": "depth_confidence", "l2": 11.341412, "mean_abs": 0.210501, "relative": 0.000376 }, { "name": "video_fisheye_cam0", "l2": 5.110243, "mean_abs": 0.107189, "relative": 0.925356 }, { "name": "video_fisheye_cam1", "l2": 5.63323, "mean_abs": 0.122165, "relative": 0.930844 }, { "name": "video_fisheye_cam2", "l2": 5.934647, "mean_abs": 0.123272, "relative": 0.947378 }, { "name": "video_fisheye_cam3", "l2": 5.574488, "mean_abs": 0.115824, "relative": 0.933381 }, { "name": "video_stereo_left", "l2": 5.531375, "mean_abs": 0.125306, "relative": 0.89886 }, { "name": "video_stereo_right", "l2": 6.730637, "mean_abs": 0.156243, "relative": 0.948879 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.794194, "mean_abs": 0.131625, "relative": 1.0 }, { "name": "caption_objects_interaction_text", "l2": 1.83535, "mean_abs": 0.01369, "relative": 0.842748 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 271, "start_frame": 1355, "end_frame": 1374, "center_frame": 1364, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.723441, "mean_abs": 0.224013, "relative": 0.820214 }, { "name": "hand_right_joints", "l2": 8.484209, "mean_abs": 0.242911, "relative": 0.668783 }, { "name": "body_joints", "l2": 5.953827, "mean_abs": 0.09598, "relative": 0.804179 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.099307, "mean_abs": 0.01326, "relative": 0.23435 }, { "name": "camera_rotation_matrix", "l2": 3.008031, "mean_abs": 0.225425, "relative": 0.923718 }, { "name": "imu_accel_gyro", "l2": 17.006449, "mean_abs": 1.066981, "relative": 0.917556 }, { "name": "depth_confidence", "l2": 11.480615, "mean_abs": 0.213008, "relative": 0.000381 }, { "name": "video_fisheye_cam0", "l2": 5.106675, "mean_abs": 0.107489, "relative": 0.92471 }, { "name": "video_fisheye_cam1", "l2": 5.629728, "mean_abs": 0.122723, "relative": 0.930265 }, { "name": "video_fisheye_cam2", "l2": 5.914225, "mean_abs": 0.123068, "relative": 0.944118 }, { "name": "video_fisheye_cam3", "l2": 5.61111, "mean_abs": 0.116876, "relative": 0.939513 }, { "name": "video_stereo_left", "l2": 5.50066, "mean_abs": 0.124582, "relative": 0.893868 }, { "name": "video_stereo_right", "l2": 6.704618, "mean_abs": 0.155106, "relative": 0.94521 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.329211, "mean_abs": 0.107011, "relative": 0.833589 }, { "name": "caption_objects_interaction_text", "l2": 1.83535, "mean_abs": 0.013655, "relative": 0.842748 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 272, "start_frame": 1360, "end_frame": 1379, "center_frame": 1369, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.71282, "mean_abs": 0.225544, "relative": 0.819318 }, { "name": "hand_right_joints", "l2": 9.217376, "mean_abs": 0.239947, "relative": 0.726576 }, { "name": "body_joints", "l2": 6.030596, "mean_abs": 0.09668, "relative": 0.814548 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.084196, "mean_abs": 0.011489, "relative": 0.198691 }, { "name": "camera_rotation_matrix", "l2": 3.006934, "mean_abs": 0.225466, "relative": 0.923381 }, { "name": "imu_accel_gyro", "l2": 17.067947, "mean_abs": 1.092029, "relative": 0.920875 }, { "name": "depth_confidence", "l2": 11.51981, "mean_abs": 0.213322, "relative": 0.000382 }, { "name": "video_fisheye_cam0", "l2": 5.112833, "mean_abs": 0.107601, "relative": 0.925825 }, { "name": "video_fisheye_cam1", "l2": 5.62787, "mean_abs": 0.122127, "relative": 0.929958 }, { "name": "video_fisheye_cam2", "l2": 5.876487, "mean_abs": 0.122244, "relative": 0.938094 }, { "name": "video_fisheye_cam3", "l2": 5.682094, "mean_abs": 0.117986, "relative": 0.951398 }, { "name": "video_stereo_left", "l2": 5.470065, "mean_abs": 0.124238, "relative": 0.888896 }, { "name": "video_stereo_right", "l2": 6.716779, "mean_abs": 0.153632, "relative": 0.946925 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.368168, "mean_abs": 0.106734, "relative": 0.847532 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011049, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 273, "start_frame": 1365, "end_frame": 1384, "center_frame": 1374, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.717905, "mean_abs": 0.226089, "relative": 0.819747 }, { "name": "hand_right_joints", "l2": 9.253459, "mean_abs": 0.237063, "relative": 0.72942 }, { "name": "body_joints", "l2": 6.074941, "mean_abs": 0.097595, "relative": 0.820537 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.066372, "mean_abs": 0.009753, "relative": 0.15663 }, { "name": "camera_rotation_matrix", "l2": 3.003492, "mean_abs": 0.22244, "relative": 0.922324 }, { "name": "imu_accel_gyro", "l2": 17.088377, "mean_abs": 1.040358, "relative": 0.921977 }, { "name": "depth_confidence", "l2": 11.558331, "mean_abs": 0.212228, "relative": 0.000384 }, { "name": "video_fisheye_cam0", "l2": 5.113801, "mean_abs": 0.107312, "relative": 0.926 }, { "name": "video_fisheye_cam1", "l2": 5.611091, "mean_abs": 0.121194, "relative": 0.927185 }, { "name": "video_fisheye_cam2", "l2": 5.827698, "mean_abs": 0.12078, "relative": 0.930305 }, { "name": "video_fisheye_cam3", "l2": 5.693594, "mean_abs": 0.118087, "relative": 0.953324 }, { "name": "video_stereo_left", "l2": 5.432694, "mean_abs": 0.12256, "relative": 0.882824 }, { "name": "video_stereo_right", "l2": 6.685812, "mean_abs": 0.153375, "relative": 0.942559 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.282928, "mean_abs": 0.096977, "relative": 0.817025 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011049, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 274, "start_frame": 1370, "end_frame": 1389, "center_frame": 1379, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.739698, "mean_abs": 0.224253, "relative": 0.821585 }, { "name": "hand_right_joints", "l2": 9.354241, "mean_abs": 0.234843, "relative": 0.737364 }, { "name": "body_joints", "l2": 6.088922, "mean_abs": 0.097046, "relative": 0.822426 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.045565, "mean_abs": 0.006897, "relative": 0.107528 }, { "name": "camera_rotation_matrix", "l2": 3.004776, "mean_abs": 0.221567, "relative": 0.922718 }, { "name": "imu_accel_gyro", "l2": 17.098032, "mean_abs": 1.042067, "relative": 0.922498 }, { "name": "depth_confidence", "l2": 11.526098, "mean_abs": 0.210937, "relative": 0.000383 }, { "name": "video_fisheye_cam0", "l2": 5.113384, "mean_abs": 0.107112, "relative": 0.925925 }, { "name": "video_fisheye_cam1", "l2": 5.588846, "mean_abs": 0.119851, "relative": 0.92351 }, { "name": "video_fisheye_cam2", "l2": 5.775044, "mean_abs": 0.119273, "relative": 0.9219 }, { "name": "video_fisheye_cam3", "l2": 5.702458, "mean_abs": 0.117035, "relative": 0.954808 }, { "name": "video_stereo_left", "l2": 5.377152, "mean_abs": 0.120495, "relative": 0.873798 }, { "name": "video_stereo_right", "l2": 6.632159, "mean_abs": 0.151561, "relative": 0.934995 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.07964, "mean_abs": 0.081867, "relative": 0.744272 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011049, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 275, "start_frame": 1375, "end_frame": 1394, "center_frame": 1384, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.691611, "mean_abs": 0.221924, "relative": 0.817529 }, { "name": "hand_right_joints", "l2": 9.407961, "mean_abs": 0.233132, "relative": 0.741599 }, { "name": "body_joints", "l2": 6.088827, "mean_abs": 0.095916, "relative": 0.822413 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.035733, "mean_abs": 0.005665, "relative": 0.084325 }, { "name": "camera_rotation_matrix", "l2": 3.003441, "mean_abs": 0.221789, "relative": 0.922308 }, { "name": "imu_accel_gyro", "l2": 17.138916, "mean_abs": 1.049538, "relative": 0.924704 }, { "name": "depth_confidence", "l2": 11.451585, "mean_abs": 0.207639, "relative": 0.00038 }, { "name": "video_fisheye_cam0", "l2": 5.116918, "mean_abs": 0.106928, "relative": 0.926564 }, { "name": "video_fisheye_cam1", "l2": 5.547627, "mean_abs": 0.118678, "relative": 0.916699 }, { "name": "video_fisheye_cam2", "l2": 5.728006, "mean_abs": 0.118167, "relative": 0.914391 }, { "name": "video_fisheye_cam3", "l2": 5.710208, "mean_abs": 0.116178, "relative": 0.956105 }, { "name": "video_stereo_left", "l2": 5.356171, "mean_abs": 0.118253, "relative": 0.870388 }, { "name": "video_stereo_right", "l2": 6.552032, "mean_abs": 0.147077, "relative": 0.923699 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.068039, "mean_abs": 0.082005, "relative": 0.74012 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011049, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 276, "start_frame": 1380, "end_frame": 1399, "center_frame": 1389, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.686098, "mean_abs": 0.221569, "relative": 0.817064 }, { "name": "hand_right_joints", "l2": 9.45128, "mean_abs": 0.232825, "relative": 0.745014 }, { "name": "body_joints", "l2": 6.104777, "mean_abs": 0.094877, "relative": 0.824567 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.027904, "mean_abs": 0.004429, "relative": 0.065849 }, { "name": "camera_rotation_matrix", "l2": 3.001738, "mean_abs": 0.219219, "relative": 0.921785 }, { "name": "imu_accel_gyro", "l2": 17.04347, "mean_abs": 1.020536, "relative": 0.919554 }, { "name": "depth_confidence", "l2": 11.42866, "mean_abs": 0.205473, "relative": 0.000379 }, { "name": "video_fisheye_cam0", "l2": 5.123639, "mean_abs": 0.106939, "relative": 0.927781 }, { "name": "video_fisheye_cam1", "l2": 5.522712, "mean_abs": 0.117522, "relative": 0.912582 }, { "name": "video_fisheye_cam2", "l2": 5.712838, "mean_abs": 0.117633, "relative": 0.91197 }, { "name": "video_fisheye_cam3", "l2": 5.713005, "mean_abs": 0.116003, "relative": 0.956574 }, { "name": "video_stereo_left", "l2": 5.359693, "mean_abs": 0.117857, "relative": 0.870961 }, { "name": "video_stereo_right", "l2": 6.510868, "mean_abs": 0.145111, "relative": 0.917896 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.130994, "mean_abs": 0.085653, "relative": 0.762651 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011049, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 277, "start_frame": 1385, "end_frame": 1404, "center_frame": 1394, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop", "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.668424, "mean_abs": 0.220355, "relative": 0.815573 }, { "name": "hand_right_joints", "l2": 9.468506, "mean_abs": 0.231971, "relative": 0.746372 }, { "name": "body_joints", "l2": 6.125956, "mean_abs": 0.092833, "relative": 0.827428 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.027426, "mean_abs": 0.003826, "relative": 0.06472 }, { "name": "camera_rotation_matrix", "l2": 3.000312, "mean_abs": 0.215056, "relative": 0.921347 }, { "name": "imu_accel_gyro", "l2": 17.080666, "mean_abs": 1.030794, "relative": 0.921561 }, { "name": "depth_confidence", "l2": 11.512229, "mean_abs": 0.204469, "relative": 0.000382 }, { "name": "video_fisheye_cam0", "l2": 5.134038, "mean_abs": 0.106557, "relative": 0.929664 }, { "name": "video_fisheye_cam1", "l2": 5.518915, "mean_abs": 0.11613, "relative": 0.911954 }, { "name": "video_fisheye_cam2", "l2": 5.701348, "mean_abs": 0.116521, "relative": 0.910135 }, { "name": "video_fisheye_cam3", "l2": 5.708743, "mean_abs": 0.115638, "relative": 0.95586 }, { "name": "video_stereo_left", "l2": 5.349969, "mean_abs": 0.116457, "relative": 0.869381 }, { "name": "video_stereo_right", "l2": 6.50153, "mean_abs": 0.142658, "relative": 0.916579 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.972101, "mean_abs": 0.078957, "relative": 0.705785 }, { "name": "caption_objects_interaction_text", "l2": 2.063714, "mean_abs": 0.017121, "relative": 0.947608 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 278, "start_frame": 1390, "end_frame": 1409, "center_frame": 1399, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop", "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.666091, "mean_abs": 0.217748, "relative": 0.815376 }, { "name": "hand_right_joints", "l2": 9.451266, "mean_abs": 0.230793, "relative": 0.745013 }, { "name": "body_joints", "l2": 6.152938, "mean_abs": 0.092052, "relative": 0.831072 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020417, "mean_abs": 0.002971, "relative": 0.04818 }, { "name": "camera_rotation_matrix", "l2": 3.000129, "mean_abs": 0.213166, "relative": 0.921291 }, { "name": "imu_accel_gyro", "l2": 17.056705, "mean_abs": 1.028448, "relative": 0.920268 }, { "name": "depth_confidence", "l2": 11.492735, "mean_abs": 0.203722, "relative": 0.000381 }, { "name": "video_fisheye_cam0", "l2": 5.145528, "mean_abs": 0.106679, "relative": 0.931745 }, { "name": "video_fisheye_cam1", "l2": 5.524562, "mean_abs": 0.116094, "relative": 0.912887 }, { "name": "video_fisheye_cam2", "l2": 5.695579, "mean_abs": 0.116352, "relative": 0.909214 }, { "name": "video_fisheye_cam3", "l2": 5.701101, "mean_abs": 0.115787, "relative": 0.954581 }, { "name": "video_stereo_left", "l2": 5.343034, "mean_abs": 0.116268, "relative": 0.868254 }, { "name": "video_stereo_right", "l2": 6.500441, "mean_abs": 0.14238, "relative": 0.916426 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.019168, "mean_abs": 0.081585, "relative": 0.72263 }, { "name": "caption_objects_interaction_text", "l2": 2.063714, "mean_abs": 0.017193, "relative": 0.947608 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 279, "start_frame": 1395, "end_frame": 1414, "center_frame": 1404, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "glass carafe", "wooden scoop", "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.691092, "mean_abs": 0.21717, "relative": 0.817485 }, { "name": "hand_right_joints", "l2": 9.487256, "mean_abs": 0.230324, "relative": 0.74785 }, { "name": "body_joints", "l2": 6.146681, "mean_abs": 0.091854, "relative": 0.830227 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.03745, "mean_abs": 0.005526, "relative": 0.088376 }, { "name": "camera_rotation_matrix", "l2": 3.003189, "mean_abs": 0.215893, "relative": 0.92223 }, { "name": "imu_accel_gyro", "l2": 16.996386, "mean_abs": 1.040934, "relative": 0.917014 }, { "name": "depth_confidence", "l2": 11.470298, "mean_abs": 0.205746, "relative": 0.000381 }, { "name": "video_fisheye_cam0", "l2": 5.167089, "mean_abs": 0.108171, "relative": 0.935649 }, { "name": "video_fisheye_cam1", "l2": 5.519633, "mean_abs": 0.117566, "relative": 0.912073 }, { "name": "video_fisheye_cam2", "l2": 5.682761, "mean_abs": 0.117821, "relative": 0.907168 }, { "name": "video_fisheye_cam3", "l2": 5.681666, "mean_abs": 0.115821, "relative": 0.951326 }, { "name": "video_stereo_left", "l2": 5.330901, "mean_abs": 0.119232, "relative": 0.866282 }, { "name": "video_stereo_right", "l2": 6.481227, "mean_abs": 0.145102, "relative": 0.913717 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.999169, "mean_abs": 0.078043, "relative": 0.715472 }, { "name": "caption_objects_interaction_text", "l2": 2.063714, "mean_abs": 0.016707, "relative": 0.947608 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 280, "start_frame": 1400, "end_frame": 1419, "center_frame": 1409, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.7455, "mean_abs": 0.218807, "relative": 0.822074 }, { "name": "hand_right_joints", "l2": 9.504, "mean_abs": 0.231919, "relative": 0.749169 }, { "name": "body_joints", "l2": 6.15955, "mean_abs": 0.093234, "relative": 0.831965 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.094143, "mean_abs": 0.013732, "relative": 0.222165 }, { "name": "camera_rotation_matrix", "l2": 3.009568, "mean_abs": 0.21937, "relative": 0.924189 }, { "name": "imu_accel_gyro", "l2": 17.025038, "mean_abs": 1.047582, "relative": 0.918559 }, { "name": "depth_confidence", "l2": 11.390788, "mean_abs": 0.208405, "relative": 0.000378 }, { "name": "video_fisheye_cam0", "l2": 5.186265, "mean_abs": 0.109439, "relative": 0.939122 }, { "name": "video_fisheye_cam1", "l2": 5.494538, "mean_abs": 0.119011, "relative": 0.907926 }, { "name": "video_fisheye_cam2", "l2": 5.660444, "mean_abs": 0.1193, "relative": 0.903606 }, { "name": "video_fisheye_cam3", "l2": 5.685555, "mean_abs": 0.117204, "relative": 0.951978 }, { "name": "video_stereo_left", "l2": 5.318007, "mean_abs": 0.121974, "relative": 0.864187 }, { "name": "video_stereo_right", "l2": 6.432597, "mean_abs": 0.146927, "relative": 0.906861 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.045841, "mean_abs": 0.07791, "relative": 0.732176 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 281, "start_frame": 1405, "end_frame": 1424, "center_frame": 1414, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.794153, "mean_abs": 0.220371, "relative": 0.826179 }, { "name": "hand_right_joints", "l2": 9.542591, "mean_abs": 0.231952, "relative": 0.752211 }, { "name": "body_joints", "l2": 6.194845, "mean_abs": 0.09425, "relative": 0.836733 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.144053, "mean_abs": 0.020036, "relative": 0.339944 }, { "name": "camera_rotation_matrix", "l2": 3.009491, "mean_abs": 0.218651, "relative": 0.924166 }, { "name": "imu_accel_gyro", "l2": 16.997881, "mean_abs": 1.057345, "relative": 0.917094 }, { "name": "depth_confidence", "l2": 11.206403, "mean_abs": 0.207264, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.195566, "mean_abs": 0.109738, "relative": 0.940806 }, { "name": "video_fisheye_cam1", "l2": 5.471405, "mean_abs": 0.118855, "relative": 0.904103 }, { "name": "video_fisheye_cam2", "l2": 5.639978, "mean_abs": 0.119234, "relative": 0.900339 }, { "name": "video_fisheye_cam3", "l2": 5.728127, "mean_abs": 0.118729, "relative": 0.959106 }, { "name": "video_stereo_left", "l2": 5.294962, "mean_abs": 0.122047, "relative": 0.860442 }, { "name": "video_stereo_right", "l2": 6.386898, "mean_abs": 0.146904, "relative": 0.900419 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.102166, "mean_abs": 0.082894, "relative": 0.752334 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 282, "start_frame": 1410, "end_frame": 1429, "center_frame": 1419, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.822307, "mean_abs": 0.219224, "relative": 0.828553 }, { "name": "hand_right_joints", "l2": 9.556839, "mean_abs": 0.229778, "relative": 0.753335 }, { "name": "body_joints", "l2": 6.279507, "mean_abs": 0.09546, "relative": 0.848168 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.196372, "mean_abs": 0.025779, "relative": 0.463411 }, { "name": "camera_rotation_matrix", "l2": 3.003756, "mean_abs": 0.213665, "relative": 0.922405 }, { "name": "imu_accel_gyro", "l2": 17.037596, "mean_abs": 1.092914, "relative": 0.919237 }, { "name": "depth_confidence", "l2": 10.976587, "mean_abs": 0.205514, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.199275, "mean_abs": 0.109041, "relative": 0.941478 }, { "name": "video_fisheye_cam1", "l2": 5.41987, "mean_abs": 0.11714, "relative": 0.895588 }, { "name": "video_fisheye_cam2", "l2": 5.600326, "mean_abs": 0.117594, "relative": 0.894009 }, { "name": "video_fisheye_cam3", "l2": 5.763286, "mean_abs": 0.12042, "relative": 0.964993 }, { "name": "video_stereo_left", "l2": 5.223972, "mean_abs": 0.119374, "relative": 0.848906 }, { "name": "video_stereo_right", "l2": 6.290254, "mean_abs": 0.143669, "relative": 0.886794 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.114558, "mean_abs": 0.082758, "relative": 0.756769 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 283, "start_frame": 1415, "end_frame": 1434, "center_frame": 1424, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.845635, "mean_abs": 0.214828, "relative": 0.830521 }, { "name": "hand_right_joints", "l2": 9.554102, "mean_abs": 0.229296, "relative": 0.753119 }, { "name": "body_joints", "l2": 6.387003, "mean_abs": 0.096351, "relative": 0.862687 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.21474, "mean_abs": 0.02625, "relative": 0.506756 }, { "name": "camera_rotation_matrix", "l2": 2.999615, "mean_abs": 0.205056, "relative": 0.921133 }, { "name": "imu_accel_gyro", "l2": 17.114035, "mean_abs": 1.114767, "relative": 0.923361 }, { "name": "depth_confidence", "l2": 10.75598, "mean_abs": 0.202692, "relative": 0.000357 }, { "name": "video_fisheye_cam0", "l2": 5.216812, "mean_abs": 0.108299, "relative": 0.944653 }, { "name": "video_fisheye_cam1", "l2": 5.335657, "mean_abs": 0.115057, "relative": 0.881672 }, { "name": "video_fisheye_cam2", "l2": 5.553376, "mean_abs": 0.114894, "relative": 0.886514 }, { "name": "video_fisheye_cam3", "l2": 5.789643, "mean_abs": 0.122627, "relative": 0.969406 }, { "name": "video_stereo_left", "l2": 5.16534, "mean_abs": 0.117005, "relative": 0.839378 }, { "name": "video_stereo_right", "l2": 6.210155, "mean_abs": 0.142088, "relative": 0.875502 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.072659, "mean_abs": 0.079514, "relative": 0.741774 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 284, "start_frame": 1420, "end_frame": 1439, "center_frame": 1429, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.753404, "mean_abs": 0.210753, "relative": 0.822741 }, { "name": "hand_right_joints", "l2": 9.714846, "mean_abs": 0.24405, "relative": 0.76579 }, { "name": "body_joints", "l2": 6.539196, "mean_abs": 0.098743, "relative": 0.883244 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.202539, "mean_abs": 0.026273, "relative": 0.477964 }, { "name": "camera_rotation_matrix", "l2": 3.010113, "mean_abs": 0.207902, "relative": 0.924357 }, { "name": "imu_accel_gyro", "l2": 17.324682, "mean_abs": 1.172726, "relative": 0.934726 }, { "name": "depth_confidence", "l2": 10.52336, "mean_abs": 0.201157, "relative": 0.000349 }, { "name": "video_fisheye_cam0", "l2": 5.227386, "mean_abs": 0.109135, "relative": 0.946568 }, { "name": "video_fisheye_cam1", "l2": 5.246757, "mean_abs": 0.11469, "relative": 0.866982 }, { "name": "video_fisheye_cam2", "l2": 5.544516, "mean_abs": 0.115222, "relative": 0.885099 }, { "name": "video_fisheye_cam3", "l2": 5.802199, "mean_abs": 0.12438, "relative": 0.971508 }, { "name": "video_stereo_left", "l2": 5.194909, "mean_abs": 0.119476, "relative": 0.844183 }, { "name": "video_stereo_right", "l2": 6.163766, "mean_abs": 0.142518, "relative": 0.868962 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.069608, "mean_abs": 0.080266, "relative": 0.740682 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 285, "start_frame": 1425, "end_frame": 1444, "center_frame": 1434, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.726887, "mean_abs": 0.209842, "relative": 0.820504 }, { "name": "hand_right_joints", "l2": 9.952864, "mean_abs": 0.259375, "relative": 0.784552 }, { "name": "body_joints", "l2": 6.668925, "mean_abs": 0.100199, "relative": 0.900766 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.169528, "mean_abs": 0.022913, "relative": 0.400063 }, { "name": "camera_rotation_matrix", "l2": 3.013717, "mean_abs": 0.208175, "relative": 0.925464 }, { "name": "imu_accel_gyro", "l2": 17.244646, "mean_abs": 1.156929, "relative": 0.930408 }, { "name": "depth_confidence", "l2": 10.36924, "mean_abs": 0.200249, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.228427, "mean_abs": 0.109493, "relative": 0.946756 }, { "name": "video_fisheye_cam1", "l2": 5.18499, "mean_abs": 0.114049, "relative": 0.856776 }, { "name": "video_fisheye_cam2", "l2": 5.556552, "mean_abs": 0.115512, "relative": 0.887021 }, { "name": "video_fisheye_cam3", "l2": 5.753373, "mean_abs": 0.123523, "relative": 0.963333 }, { "name": "video_stereo_left", "l2": 5.235591, "mean_abs": 0.121095, "relative": 0.850794 }, { "name": "video_stereo_right", "l2": 6.152121, "mean_abs": 0.142972, "relative": 0.86732 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.068583, "mean_abs": 0.0792, "relative": 0.740315 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 286, "start_frame": 1430, "end_frame": 1449, "center_frame": 1439, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.699574, "mean_abs": 0.211043, "relative": 0.8182 }, { "name": "hand_right_joints", "l2": 10.081513, "mean_abs": 0.265169, "relative": 0.794693 }, { "name": "body_joints", "l2": 6.750375, "mean_abs": 0.0999, "relative": 0.911768 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.108897, "mean_abs": 0.014916, "relative": 0.256982 }, { "name": "camera_rotation_matrix", "l2": 3.010144, "mean_abs": 0.203802, "relative": 0.924366 }, { "name": "imu_accel_gyro", "l2": 17.208544, "mean_abs": 1.120621, "relative": 0.92846 }, { "name": "depth_confidence", "l2": 10.163003, "mean_abs": 0.1961, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.223038, "mean_abs": 0.108941, "relative": 0.945781 }, { "name": "video_fisheye_cam1", "l2": 5.113304, "mean_abs": 0.111592, "relative": 0.84493 }, { "name": "video_fisheye_cam2", "l2": 5.560722, "mean_abs": 0.11469, "relative": 0.887687 }, { "name": "video_fisheye_cam3", "l2": 5.694725, "mean_abs": 0.121109, "relative": 0.953513 }, { "name": "video_stereo_left", "l2": 5.253093, "mean_abs": 0.120109, "relative": 0.853638 }, { "name": "video_stereo_right", "l2": 6.137268, "mean_abs": 0.141391, "relative": 0.865226 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.061397, "mean_abs": 0.079898, "relative": 0.737743 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 287, "start_frame": 1435, "end_frame": 1454, "center_frame": 1444, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.603027, "mean_abs": 0.20876, "relative": 0.810056 }, { "name": "hand_right_joints", "l2": 10.228284, "mean_abs": 0.267496, "relative": 0.806262 }, { "name": "body_joints", "l2": 6.816494, "mean_abs": 0.097492, "relative": 0.920698 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08169, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.060713, "mean_abs": 0.009167, "relative": 0.143274 }, { "name": "camera_rotation_matrix", "l2": 3.00432, "mean_abs": 0.197723, "relative": 0.922578 }, { "name": "imu_accel_gyro", "l2": 17.195116, "mean_abs": 1.082169, "relative": 0.927736 }, { "name": "depth_confidence", "l2": 9.920102, "mean_abs": 0.189936, "relative": 0.000329 }, { "name": "video_fisheye_cam0", "l2": 5.220853, "mean_abs": 0.1082, "relative": 0.945385 }, { "name": "video_fisheye_cam1", "l2": 5.06482, "mean_abs": 0.109189, "relative": 0.836919 }, { "name": "video_fisheye_cam2", "l2": 5.569496, "mean_abs": 0.114012, "relative": 0.889087 }, { "name": "video_fisheye_cam3", "l2": 5.587923, "mean_abs": 0.116512, "relative": 0.93563 }, { "name": "video_stereo_left", "l2": 5.269536, "mean_abs": 0.118143, "relative": 0.85631 }, { "name": "video_stereo_right", "l2": 6.137001, "mean_abs": 0.138937, "relative": 0.865188 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.073169, "mean_abs": 0.080974, "relative": 0.741956 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 288, "start_frame": 1440, "end_frame": 1459, "center_frame": 1449, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.569592, "mean_abs": 0.208988, "relative": 0.807236 }, { "name": "hand_right_joints", "l2": 10.443089, "mean_abs": 0.267235, "relative": 0.823195 }, { "name": "body_joints", "l2": 6.853278, "mean_abs": 0.093891, "relative": 0.925667 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081689, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.035968, "mean_abs": 0.005051, "relative": 0.084879 }, { "name": "camera_rotation_matrix", "l2": 3.000892, "mean_abs": 0.191244, "relative": 0.921525 }, { "name": "imu_accel_gyro", "l2": 17.154562, "mean_abs": 1.054409, "relative": 0.925548 }, { "name": "depth_confidence", "l2": 9.701561, "mean_abs": 0.182058, "relative": 0.000322 }, { "name": "video_fisheye_cam0", "l2": 5.216549, "mean_abs": 0.10769, "relative": 0.944606 }, { "name": "video_fisheye_cam1", "l2": 5.052258, "mean_abs": 0.107935, "relative": 0.834843 }, { "name": "video_fisheye_cam2", "l2": 5.573712, "mean_abs": 0.113159, "relative": 0.88976 }, { "name": "video_fisheye_cam3", "l2": 5.577678, "mean_abs": 0.115156, "relative": 0.933915 }, { "name": "video_stereo_left", "l2": 5.303373, "mean_abs": 0.115764, "relative": 0.861809 }, { "name": "video_stereo_right", "l2": 6.162776, "mean_abs": 0.13645, "relative": 0.868822 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.036004, "mean_abs": 0.081617, "relative": 0.728655 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 289, "start_frame": 1445, "end_frame": 1464, "center_frame": 1454, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.516574, "mean_abs": 0.209132, "relative": 0.802764 }, { "name": "hand_right_joints", "l2": 10.506244, "mean_abs": 0.268177, "relative": 0.828173 }, { "name": "body_joints", "l2": 6.852848, "mean_abs": 0.09314, "relative": 0.925609 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081689, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030651, "mean_abs": 0.004488, "relative": 0.072332 }, { "name": "camera_rotation_matrix", "l2": 3.000277, "mean_abs": 0.189263, "relative": 0.921336 }, { "name": "imu_accel_gyro", "l2": 17.064613, "mean_abs": 1.034059, "relative": 0.920695 }, { "name": "depth_confidence", "l2": 9.661827, "mean_abs": 0.180038, "relative": 0.000321 }, { "name": "video_fisheye_cam0", "l2": 5.21207, "mean_abs": 0.107358, "relative": 0.943794 }, { "name": "video_fisheye_cam1", "l2": 5.051494, "mean_abs": 0.107744, "relative": 0.834717 }, { "name": "video_fisheye_cam2", "l2": 5.566387, "mean_abs": 0.112709, "relative": 0.888591 }, { "name": "video_fisheye_cam3", "l2": 5.56848, "mean_abs": 0.114291, "relative": 0.932375 }, { "name": "video_stereo_left", "l2": 5.304652, "mean_abs": 0.114674, "relative": 0.862016 }, { "name": "video_stereo_right", "l2": 6.178004, "mean_abs": 0.135928, "relative": 0.870969 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.0713, "mean_abs": 0.082748, "relative": 0.741287 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 290, "start_frame": 1450, "end_frame": 1469, "center_frame": 1459, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.510904, "mean_abs": 0.209048, "relative": 0.802285 }, { "name": "hand_right_joints", "l2": 10.502724, "mean_abs": 0.268318, "relative": 0.827896 }, { "name": "body_joints", "l2": 6.85216, "mean_abs": 0.092839, "relative": 0.925516 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021124, "mean_abs": 0.003061, "relative": 0.04985 }, { "name": "camera_rotation_matrix", "l2": 3.000407, "mean_abs": 0.18761, "relative": 0.921376 }, { "name": "imu_accel_gyro", "l2": 17.058535, "mean_abs": 1.035756, "relative": 0.920367 }, { "name": "depth_confidence", "l2": 9.621367, "mean_abs": 0.178743, "relative": 0.000319 }, { "name": "video_fisheye_cam0", "l2": 5.21025, "mean_abs": 0.107287, "relative": 0.943465 }, { "name": "video_fisheye_cam1", "l2": 5.080344, "mean_abs": 0.108355, "relative": 0.839484 }, { "name": "video_fisheye_cam2", "l2": 5.560247, "mean_abs": 0.112287, "relative": 0.887611 }, { "name": "video_fisheye_cam3", "l2": 5.569367, "mean_abs": 0.114513, "relative": 0.932523 }, { "name": "video_stereo_left", "l2": 5.312565, "mean_abs": 0.113951, "relative": 0.863302 }, { "name": "video_stereo_right", "l2": 6.225675, "mean_abs": 0.136564, "relative": 0.87769 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.005365, "mean_abs": 0.077278, "relative": 0.71769 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 291, "start_frame": 1455, "end_frame": 1474, "center_frame": 1464, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.513605, "mean_abs": 0.210181, "relative": 0.802513 }, { "name": "hand_right_joints", "l2": 10.474496, "mean_abs": 0.268407, "relative": 0.82567 }, { "name": "body_joints", "l2": 6.839692, "mean_abs": 0.093159, "relative": 0.923832 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009743, "mean_abs": 0.001578, "relative": 0.022993 }, { "name": "camera_rotation_matrix", "l2": 3.000269, "mean_abs": 0.187494, "relative": 0.921334 }, { "name": "imu_accel_gyro", "l2": 17.066019, "mean_abs": 1.036246, "relative": 0.92077 }, { "name": "depth_confidence", "l2": 9.619804, "mean_abs": 0.17841, "relative": 0.000319 }, { "name": "video_fisheye_cam0", "l2": 5.208581, "mean_abs": 0.107056, "relative": 0.943163 }, { "name": "video_fisheye_cam1", "l2": 5.15471, "mean_abs": 0.110237, "relative": 0.851772 }, { "name": "video_fisheye_cam2", "l2": 5.554585, "mean_abs": 0.112244, "relative": 0.886707 }, { "name": "video_fisheye_cam3", "l2": 5.578822, "mean_abs": 0.114525, "relative": 0.934106 }, { "name": "video_stereo_left", "l2": 5.312404, "mean_abs": 0.113911, "relative": 0.863276 }, { "name": "video_stereo_right", "l2": 6.309216, "mean_abs": 0.138947, "relative": 0.889467 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.999748, "mean_abs": 0.078805, "relative": 0.71568 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 292, "start_frame": 1460, "end_frame": 1479, "center_frame": 1469, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.5159, "mean_abs": 0.213286, "relative": 0.802707 }, { "name": "hand_right_joints", "l2": 10.425947, "mean_abs": 0.270691, "relative": 0.821844 }, { "name": "body_joints", "l2": 6.845068, "mean_abs": 0.095306, "relative": 0.924558 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018481, "mean_abs": 0.002525, "relative": 0.043613 }, { "name": "camera_rotation_matrix", "l2": 3.000326, "mean_abs": 0.188336, "relative": 0.921351 }, { "name": "imu_accel_gyro", "l2": 17.09387, "mean_abs": 1.03483, "relative": 0.922273 }, { "name": "depth_confidence", "l2": 9.623096, "mean_abs": 0.179064, "relative": 0.000319 }, { "name": "video_fisheye_cam0", "l2": 5.199942, "mean_abs": 0.106986, "relative": 0.941598 }, { "name": "video_fisheye_cam1", "l2": 5.190936, "mean_abs": 0.110736, "relative": 0.857758 }, { "name": "video_fisheye_cam2", "l2": 5.543932, "mean_abs": 0.112286, "relative": 0.885006 }, { "name": "video_fisheye_cam3", "l2": 5.544773, "mean_abs": 0.114469, "relative": 0.928405 }, { "name": "video_stereo_left", "l2": 5.282261, "mean_abs": 0.114273, "relative": 0.858378 }, { "name": "video_stereo_right", "l2": 6.339127, "mean_abs": 0.139416, "relative": 0.893684 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.049743, "mean_abs": 0.081858, "relative": 0.733572 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008566, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 293, "start_frame": 1465, "end_frame": 1484, "center_frame": 1474, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.549164, "mean_abs": 0.216339, "relative": 0.805513 }, { "name": "hand_right_joints", "l2": 10.098849, "mean_abs": 0.273437, "relative": 0.796059 }, { "name": "body_joints", "l2": 6.813721, "mean_abs": 0.099798, "relative": 0.920324 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081689, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.035363, "mean_abs": 0.005268, "relative": 0.083452 }, { "name": "camera_rotation_matrix", "l2": 3.00104, "mean_abs": 0.189509, "relative": 0.921571 }, { "name": "imu_accel_gyro", "l2": 17.087336, "mean_abs": 1.057711, "relative": 0.921921 }, { "name": "depth_confidence", "l2": 9.684683, "mean_abs": 0.182614, "relative": 0.000321 }, { "name": "video_fisheye_cam0", "l2": 5.202796, "mean_abs": 0.107202, "relative": 0.942115 }, { "name": "video_fisheye_cam1", "l2": 5.230546, "mean_abs": 0.111489, "relative": 0.864303 }, { "name": "video_fisheye_cam2", "l2": 5.549239, "mean_abs": 0.112591, "relative": 0.885853 }, { "name": "video_fisheye_cam3", "l2": 5.532889, "mean_abs": 0.114594, "relative": 0.926416 }, { "name": "video_stereo_left", "l2": 5.283081, "mean_abs": 0.115492, "relative": 0.858511 }, { "name": "video_stereo_right", "l2": 6.376971, "mean_abs": 0.141594, "relative": 0.899019 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.069704, "mean_abs": 0.083128, "relative": 0.740716 }, { "name": "caption_objects_interaction_text", "l2": 1.87573, "mean_abs": 0.012695, "relative": 0.86129 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 294, "start_frame": 1470, "end_frame": 1489, "center_frame": 1479, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.602576, "mean_abs": 0.223096, "relative": 0.810018 }, { "name": "hand_right_joints", "l2": 9.881525, "mean_abs": 0.272314, "relative": 0.778929 }, { "name": "body_joints", "l2": 6.763934, "mean_abs": 0.105374, "relative": 0.913599 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081689, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.06331, "mean_abs": 0.009159, "relative": 0.149402 }, { "name": "camera_rotation_matrix", "l2": 3.001225, "mean_abs": 0.195137, "relative": 0.921627 }, { "name": "imu_accel_gyro", "l2": 17.098225, "mean_abs": 1.069912, "relative": 0.922508 }, { "name": "depth_confidence", "l2": 9.828326, "mean_abs": 0.187669, "relative": 0.000326 }, { "name": "video_fisheye_cam0", "l2": 5.204686, "mean_abs": 0.10823, "relative": 0.942457 }, { "name": "video_fisheye_cam1", "l2": 5.262464, "mean_abs": 0.113685, "relative": 0.869578 }, { "name": "video_fisheye_cam2", "l2": 5.579107, "mean_abs": 0.114276, "relative": 0.890621 }, { "name": "video_fisheye_cam3", "l2": 5.512398, "mean_abs": 0.114946, "relative": 0.922985 }, { "name": "video_stereo_left", "l2": 5.318676, "mean_abs": 0.11762, "relative": 0.864295 }, { "name": "video_stereo_right", "l2": 6.45524, "mean_abs": 0.145968, "relative": 0.910053 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.051955, "mean_abs": 0.080442, "relative": 0.734364 }, { "name": "caption_objects_interaction_text", "l2": 1.87573, "mean_abs": 0.012812, "relative": 0.86129 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 295, "start_frame": 1475, "end_frame": 1494, "center_frame": 1484, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.65082, "mean_abs": 0.227849, "relative": 0.814088 }, { "name": "hand_right_joints", "l2": 9.77028, "mean_abs": 0.267665, "relative": 0.770159 }, { "name": "body_joints", "l2": 6.762012, "mean_abs": 0.109191, "relative": 0.91334 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.084177, "mean_abs": 0.012188, "relative": 0.198645 }, { "name": "camera_rotation_matrix", "l2": 3.001354, "mean_abs": 0.196198, "relative": 0.921667 }, { "name": "imu_accel_gyro", "l2": 17.112318, "mean_abs": 1.060752, "relative": 0.923268 }, { "name": "depth_confidence", "l2": 9.884764, "mean_abs": 0.188636, "relative": 0.000328 }, { "name": "video_fisheye_cam0", "l2": 5.202083, "mean_abs": 0.108283, "relative": 0.941986 }, { "name": "video_fisheye_cam1", "l2": 5.287461, "mean_abs": 0.114279, "relative": 0.873708 }, { "name": "video_fisheye_cam2", "l2": 5.581779, "mean_abs": 0.114648, "relative": 0.891048 }, { "name": "video_fisheye_cam3", "l2": 5.471572, "mean_abs": 0.113697, "relative": 0.916149 }, { "name": "video_stereo_left", "l2": 5.32245, "mean_abs": 0.117876, "relative": 0.864909 }, { "name": "video_stereo_right", "l2": 6.500834, "mean_abs": 0.146828, "relative": 0.916481 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.027236, "mean_abs": 0.079596, "relative": 0.725517 }, { "name": "caption_objects_interaction_text", "l2": 1.87573, "mean_abs": 0.012582, "relative": 0.86129 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 296, "start_frame": 1480, "end_frame": 1499, "center_frame": 1489, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.717709, "mean_abs": 0.233061, "relative": 0.81973 }, { "name": "hand_right_joints", "l2": 9.567458, "mean_abs": 0.257352, "relative": 0.754172 }, { "name": "body_joints", "l2": 6.723404, "mean_abs": 0.111536, "relative": 0.908125 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.095181, "mean_abs": 0.013349, "relative": 0.224613 }, { "name": "camera_rotation_matrix", "l2": 3.003084, "mean_abs": 0.198251, "relative": 0.922198 }, { "name": "imu_accel_gyro", "l2": 17.069056, "mean_abs": 1.056575, "relative": 0.920934 }, { "name": "depth_confidence", "l2": 9.996264, "mean_abs": 0.190604, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.205228, "mean_abs": 0.108368, "relative": 0.942556 }, { "name": "video_fisheye_cam1", "l2": 5.321284, "mean_abs": 0.115353, "relative": 0.879297 }, { "name": "video_fisheye_cam2", "l2": 5.590297, "mean_abs": 0.114912, "relative": 0.892408 }, { "name": "video_fisheye_cam3", "l2": 5.466699, "mean_abs": 0.113793, "relative": 0.915333 }, { "name": "video_stereo_left", "l2": 5.322904, "mean_abs": 0.118053, "relative": 0.864982 }, { "name": "video_stereo_right", "l2": 6.537544, "mean_abs": 0.147482, "relative": 0.921657 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.080377, "mean_abs": 0.082021, "relative": 0.744536 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 297, "start_frame": 1485, "end_frame": 1504, "center_frame": 1494, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.757787, "mean_abs": 0.235103, "relative": 0.823111 }, { "name": "hand_right_joints", "l2": 9.369354, "mean_abs": 0.245488, "relative": 0.738556 }, { "name": "body_joints", "l2": 6.570498, "mean_abs": 0.108925, "relative": 0.887472 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.078371, "mean_abs": 0.010986, "relative": 0.184944 }, { "name": "camera_rotation_matrix", "l2": 3.00225, "mean_abs": 0.197164, "relative": 0.921942 }, { "name": "imu_accel_gyro", "l2": 17.064737, "mean_abs": 1.071357, "relative": 0.920701 }, { "name": "depth_confidence", "l2": 10.135685, "mean_abs": 0.190932, "relative": 0.000336 }, { "name": "video_fisheye_cam0", "l2": 5.201008, "mean_abs": 0.107693, "relative": 0.941791 }, { "name": "video_fisheye_cam1", "l2": 5.341707, "mean_abs": 0.114871, "relative": 0.882672 }, { "name": "video_fisheye_cam2", "l2": 5.607103, "mean_abs": 0.114173, "relative": 0.895091 }, { "name": "video_fisheye_cam3", "l2": 5.452203, "mean_abs": 0.113238, "relative": 0.912906 }, { "name": "video_stereo_left", "l2": 5.339283, "mean_abs": 0.116331, "relative": 0.867644 }, { "name": "video_stereo_right", "l2": 6.575767, "mean_abs": 0.146122, "relative": 0.927045 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.115361, "mean_abs": 0.086165, "relative": 0.757056 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 298, "start_frame": 1490, "end_frame": 1509, "center_frame": 1499, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.8143, "mean_abs": 0.235257, "relative": 0.827878 }, { "name": "hand_right_joints", "l2": 9.514458, "mean_abs": 0.245929, "relative": 0.749994 }, { "name": "body_joints", "l2": 6.462941, "mean_abs": 0.105626, "relative": 0.872944 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.044904, "mean_abs": 0.006648, "relative": 0.105968 }, { "name": "camera_rotation_matrix", "l2": 3.000974, "mean_abs": 0.197757, "relative": 0.92155 }, { "name": "imu_accel_gyro", "l2": 16.968208, "mean_abs": 1.030158, "relative": 0.915493 }, { "name": "depth_confidence", "l2": 10.215199, "mean_abs": 0.191427, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.182941, "mean_abs": 0.107128, "relative": 0.93852 }, { "name": "video_fisheye_cam1", "l2": 5.371021, "mean_abs": 0.115119, "relative": 0.887516 }, { "name": "video_fisheye_cam2", "l2": 5.604987, "mean_abs": 0.113992, "relative": 0.894753 }, { "name": "video_fisheye_cam3", "l2": 5.417381, "mean_abs": 0.11241, "relative": 0.907075 }, { "name": "video_stereo_left", "l2": 5.33057, "mean_abs": 0.116241, "relative": 0.866228 }, { "name": "video_stereo_right", "l2": 6.605362, "mean_abs": 0.146411, "relative": 0.931217 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.147024, "mean_abs": 0.085819, "relative": 0.768388 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 299, "start_frame": 1495, "end_frame": 1514, "center_frame": 1504, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.887262, "mean_abs": 0.233321, "relative": 0.834033 }, { "name": "hand_right_joints", "l2": 9.6571, "mean_abs": 0.245595, "relative": 0.761238 }, { "name": "body_joints", "l2": 6.439976, "mean_abs": 0.101916, "relative": 0.869842 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.02097, "mean_abs": 0.002849, "relative": 0.049485 }, { "name": "camera_rotation_matrix", "l2": 2.999914, "mean_abs": 0.196185, "relative": 0.921225 }, { "name": "imu_accel_gyro", "l2": 16.945004, "mean_abs": 0.99883, "relative": 0.914241 }, { "name": "depth_confidence", "l2": 10.256418, "mean_abs": 0.190912, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.179863, "mean_abs": 0.106979, "relative": 0.937962 }, { "name": "video_fisheye_cam1", "l2": 5.407963, "mean_abs": 0.114643, "relative": 0.89362 }, { "name": "video_fisheye_cam2", "l2": 5.599537, "mean_abs": 0.11355, "relative": 0.893883 }, { "name": "video_fisheye_cam3", "l2": 5.43733, "mean_abs": 0.111613, "relative": 0.910415 }, { "name": "video_stereo_left", "l2": 5.310851, "mean_abs": 0.11526, "relative": 0.863024 }, { "name": "video_stereo_right", "l2": 6.626569, "mean_abs": 0.145327, "relative": 0.934207 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.145011, "mean_abs": 0.08417, "relative": 0.767667 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 300, "start_frame": 1500, "end_frame": 1519, "center_frame": 1509, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.89012, "mean_abs": 0.232658, "relative": 0.834274 }, { "name": "hand_right_joints", "l2": 9.723059, "mean_abs": 0.245391, "relative": 0.766437 }, { "name": "body_joints", "l2": 6.430896, "mean_abs": 0.101846, "relative": 0.868616 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00434, "mean_abs": 0.000704, "relative": 0.010242 }, { "name": "camera_rotation_matrix", "l2": 3.000324, "mean_abs": 0.196078, "relative": 0.921351 }, { "name": "imu_accel_gyro", "l2": 17.07151, "mean_abs": 0.997425, "relative": 0.921067 }, { "name": "depth_confidence", "l2": 10.310719, "mean_abs": 0.191054, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.173269, "mean_abs": 0.106708, "relative": 0.936768 }, { "name": "video_fisheye_cam1", "l2": 5.415574, "mean_abs": 0.114454, "relative": 0.894878 }, { "name": "video_fisheye_cam2", "l2": 5.59398, "mean_abs": 0.113239, "relative": 0.892996 }, { "name": "video_fisheye_cam3", "l2": 5.430921, "mean_abs": 0.110666, "relative": 0.909342 }, { "name": "video_stereo_left", "l2": 5.290257, "mean_abs": 0.114481, "relative": 0.859677 }, { "name": "video_stereo_right", "l2": 6.602024, "mean_abs": 0.144102, "relative": 0.930747 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.047815, "mean_abs": 0.082197, "relative": 0.732882 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 301, "start_frame": 1505, "end_frame": 1524, "center_frame": 1514, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.890753, "mean_abs": 0.232928, "relative": 0.834327 }, { "name": "hand_right_joints", "l2": 9.769999, "mean_abs": 0.245825, "relative": 0.770137 }, { "name": "body_joints", "l2": 6.412717, "mean_abs": 0.101456, "relative": 0.86616 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009181, "mean_abs": 0.001238, "relative": 0.021666 }, { "name": "camera_rotation_matrix", "l2": 3.000337, "mean_abs": 0.195554, "relative": 0.921355 }, { "name": "imu_accel_gyro", "l2": 17.054707, "mean_abs": 1.002684, "relative": 0.92016 }, { "name": "depth_confidence", "l2": 10.331401, "mean_abs": 0.190876, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.171387, "mean_abs": 0.106504, "relative": 0.936428 }, { "name": "video_fisheye_cam1", "l2": 5.419071, "mean_abs": 0.114247, "relative": 0.895456 }, { "name": "video_fisheye_cam2", "l2": 5.586034, "mean_abs": 0.112967, "relative": 0.891727 }, { "name": "video_fisheye_cam3", "l2": 5.437085, "mean_abs": 0.110482, "relative": 0.910374 }, { "name": "video_stereo_left", "l2": 5.2663, "mean_abs": 0.113767, "relative": 0.855784 }, { "name": "video_stereo_right", "l2": 6.579762, "mean_abs": 0.143326, "relative": 0.927608 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.086585, "mean_abs": 0.082327, "relative": 0.746757 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 302, "start_frame": 1510, "end_frame": 1529, "center_frame": 1519, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.876197, "mean_abs": 0.231335, "relative": 0.833099 }, { "name": "hand_right_joints", "l2": 9.809093, "mean_abs": 0.244891, "relative": 0.773219 }, { "name": "body_joints", "l2": 6.370675, "mean_abs": 0.100401, "relative": 0.860482 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016351, "mean_abs": 0.002083, "relative": 0.038586 }, { "name": "camera_rotation_matrix", "l2": 3.000298, "mean_abs": 0.194715, "relative": 0.921343 }, { "name": "imu_accel_gyro", "l2": 17.028555, "mean_abs": 0.997197, "relative": 0.918749 }, { "name": "depth_confidence", "l2": 10.3286, "mean_abs": 0.190563, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.16881, "mean_abs": 0.106359, "relative": 0.935961 }, { "name": "video_fisheye_cam1", "l2": 5.413745, "mean_abs": 0.113873, "relative": 0.894576 }, { "name": "video_fisheye_cam2", "l2": 5.57755, "mean_abs": 0.112664, "relative": 0.890373 }, { "name": "video_fisheye_cam3", "l2": 5.441962, "mean_abs": 0.110459, "relative": 0.911191 }, { "name": "video_stereo_left", "l2": 5.246047, "mean_abs": 0.113088, "relative": 0.852493 }, { "name": "video_stereo_right", "l2": 6.551882, "mean_abs": 0.142769, "relative": 0.923678 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.107117, "mean_abs": 0.08326, "relative": 0.754106 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 303, "start_frame": 1515, "end_frame": 1534, "center_frame": 1524, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.909333, "mean_abs": 0.232084, "relative": 0.835895 }, { "name": "hand_right_joints", "l2": 9.819897, "mean_abs": 0.245149, "relative": 0.774071 }, { "name": "body_joints", "l2": 6.339997, "mean_abs": 0.099474, "relative": 0.856338 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.025469, "mean_abs": 0.003294, "relative": 0.060104 }, { "name": "camera_rotation_matrix", "l2": 3.00029, "mean_abs": 0.19452, "relative": 0.92134 }, { "name": "imu_accel_gyro", "l2": 17.01115, "mean_abs": 0.999275, "relative": 0.91781 }, { "name": "depth_confidence", "l2": 10.307796, "mean_abs": 0.190494, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.161975, "mean_abs": 0.106296, "relative": 0.934723 }, { "name": "video_fisheye_cam1", "l2": 5.392605, "mean_abs": 0.113619, "relative": 0.891083 }, { "name": "video_fisheye_cam2", "l2": 5.563418, "mean_abs": 0.112433, "relative": 0.888117 }, { "name": "video_fisheye_cam3", "l2": 5.43728, "mean_abs": 0.110883, "relative": 0.910407 }, { "name": "video_stereo_left", "l2": 5.223446, "mean_abs": 0.113087, "relative": 0.84882 }, { "name": "video_stereo_right", "l2": 6.502651, "mean_abs": 0.142355, "relative": 0.916737 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.092117, "mean_abs": 0.083833, "relative": 0.748737 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 304, "start_frame": 1520, "end_frame": 1539, "center_frame": 1529, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.95063, "mean_abs": 0.231937, "relative": 0.839378 }, { "name": "hand_right_joints", "l2": 9.80668, "mean_abs": 0.243132, "relative": 0.773029 }, { "name": "body_joints", "l2": 6.339563, "mean_abs": 0.099796, "relative": 0.85628 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.036553, "mean_abs": 0.004286, "relative": 0.086261 }, { "name": "camera_rotation_matrix", "l2": 3.000135, "mean_abs": 0.194917, "relative": 0.921293 }, { "name": "imu_accel_gyro", "l2": 17.010483, "mean_abs": 0.99713, "relative": 0.917774 }, { "name": "depth_confidence", "l2": 10.284783, "mean_abs": 0.190524, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.160562, "mean_abs": 0.106443, "relative": 0.934467 }, { "name": "video_fisheye_cam1", "l2": 5.38557, "mean_abs": 0.114, "relative": 0.88992 }, { "name": "video_fisheye_cam2", "l2": 5.550074, "mean_abs": 0.112421, "relative": 0.885987 }, { "name": "video_fisheye_cam3", "l2": 5.444437, "mean_abs": 0.111133, "relative": 0.911605 }, { "name": "video_stereo_left", "l2": 5.204961, "mean_abs": 0.113608, "relative": 0.845816 }, { "name": "video_stereo_right", "l2": 6.474418, "mean_abs": 0.14286, "relative": 0.912757 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.03396, "mean_abs": 0.081785, "relative": 0.727924 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 305, "start_frame": 1525, "end_frame": 1544, "center_frame": 1534, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.9515, "mean_abs": 0.229254, "relative": 0.839451 }, { "name": "hand_right_joints", "l2": 9.938949, "mean_abs": 0.250101, "relative": 0.783455 }, { "name": "body_joints", "l2": 6.359292, "mean_abs": 0.100128, "relative": 0.858944 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.054259, "mean_abs": 0.006556, "relative": 0.128044 }, { "name": "camera_rotation_matrix", "l2": 3.000659, "mean_abs": 0.195862, "relative": 0.921454 }, { "name": "imu_accel_gyro", "l2": 16.981121, "mean_abs": 1.003404, "relative": 0.91619 }, { "name": "depth_confidence", "l2": 10.214327, "mean_abs": 0.190263, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.158792, "mean_abs": 0.106669, "relative": 0.934147 }, { "name": "video_fisheye_cam1", "l2": 5.37136, "mean_abs": 0.114115, "relative": 0.887572 }, { "name": "video_fisheye_cam2", "l2": 5.531938, "mean_abs": 0.112717, "relative": 0.883092 }, { "name": "video_fisheye_cam3", "l2": 5.451113, "mean_abs": 0.111848, "relative": 0.912723 }, { "name": "video_stereo_left", "l2": 5.183411, "mean_abs": 0.114255, "relative": 0.842315 }, { "name": "video_stereo_right", "l2": 6.432311, "mean_abs": 0.143152, "relative": 0.906821 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.076747, "mean_abs": 0.080264, "relative": 0.743237 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 306, "start_frame": 1530, "end_frame": 1549, "center_frame": 1539, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.966111, "mean_abs": 0.227864, "relative": 0.840684 }, { "name": "hand_right_joints", "l2": 10.284102, "mean_abs": 0.26708, "relative": 0.810662 }, { "name": "body_joints", "l2": 6.390523, "mean_abs": 0.101037, "relative": 0.863163 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.066277, "mean_abs": 0.008485, "relative": 0.156405 }, { "name": "camera_rotation_matrix", "l2": 3.000291, "mean_abs": 0.19495, "relative": 0.921341 }, { "name": "imu_accel_gyro", "l2": 17.094479, "mean_abs": 1.022045, "relative": 0.922306 }, { "name": "depth_confidence", "l2": 10.100984, "mean_abs": 0.189582, "relative": 0.000335 }, { "name": "video_fisheye_cam0", "l2": 5.158445, "mean_abs": 0.106733, "relative": 0.934084 }, { "name": "video_fisheye_cam1", "l2": 5.352499, "mean_abs": 0.113469, "relative": 0.884455 }, { "name": "video_fisheye_cam2", "l2": 5.520586, "mean_abs": 0.112446, "relative": 0.881279 }, { "name": "video_fisheye_cam3", "l2": 5.480371, "mean_abs": 0.11353, "relative": 0.917622 }, { "name": "video_stereo_left", "l2": 5.169675, "mean_abs": 0.113409, "relative": 0.840083 }, { "name": "video_stereo_right", "l2": 6.390741, "mean_abs": 0.141936, "relative": 0.90096 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.020074, "mean_abs": 0.076873, "relative": 0.722954 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 307, "start_frame": 1535, "end_frame": 1554, "center_frame": 1544, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.956464, "mean_abs": 0.223976, "relative": 0.83987 }, { "name": "hand_right_joints", "l2": 10.410007, "mean_abs": 0.271647, "relative": 0.820587 }, { "name": "body_joints", "l2": 6.41665, "mean_abs": 0.102516, "relative": 0.866692 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.064928, "mean_abs": 0.008968, "relative": 0.153222 }, { "name": "camera_rotation_matrix", "l2": 2.999709, "mean_abs": 0.193835, "relative": 0.921162 }, { "name": "imu_accel_gyro", "l2": 17.105158, "mean_abs": 1.044668, "relative": 0.922882 }, { "name": "depth_confidence", "l2": 10.053478, "mean_abs": 0.189241, "relative": 0.000334 }, { "name": "video_fisheye_cam0", "l2": 5.158911, "mean_abs": 0.106492, "relative": 0.934168 }, { "name": "video_fisheye_cam1", "l2": 5.333092, "mean_abs": 0.113, "relative": 0.881248 }, { "name": "video_fisheye_cam2", "l2": 5.514067, "mean_abs": 0.111888, "relative": 0.880239 }, { "name": "video_fisheye_cam3", "l2": 5.487322, "mean_abs": 0.11404, "relative": 0.918786 }, { "name": "video_stereo_left", "l2": 5.164422, "mean_abs": 0.112831, "relative": 0.839229 }, { "name": "video_stereo_right", "l2": 6.357343, "mean_abs": 0.140972, "relative": 0.896252 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.059683, "mean_abs": 0.081041, "relative": 0.73713 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 308, "start_frame": 1540, "end_frame": 1559, "center_frame": 1549, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.927793, "mean_abs": 0.223968, "relative": 0.837452 }, { "name": "hand_right_joints", "l2": 10.594308, "mean_abs": 0.27793, "relative": 0.835115 }, { "name": "body_joints", "l2": 6.434978, "mean_abs": 0.103708, "relative": 0.869167 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.052311, "mean_abs": 0.007697, "relative": 0.123446 }, { "name": "camera_rotation_matrix", "l2": 3.001521, "mean_abs": 0.193697, "relative": 0.921718 }, { "name": "imu_accel_gyro", "l2": 17.089779, "mean_abs": 1.045145, "relative": 0.922052 }, { "name": "depth_confidence", "l2": 10.011315, "mean_abs": 0.188334, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.164029, "mean_abs": 0.106511, "relative": 0.935095 }, { "name": "video_fisheye_cam1", "l2": 5.325355, "mean_abs": 0.112793, "relative": 0.87997 }, { "name": "video_fisheye_cam2", "l2": 5.517522, "mean_abs": 0.112034, "relative": 0.88079 }, { "name": "video_fisheye_cam3", "l2": 5.506252, "mean_abs": 0.114455, "relative": 0.921955 }, { "name": "video_stereo_left", "l2": 5.185356, "mean_abs": 0.1137, "relative": 0.842631 }, { "name": "video_stereo_right", "l2": 6.361077, "mean_abs": 0.141111, "relative": 0.896778 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.019516, "mean_abs": 0.078032, "relative": 0.722754 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 309, "start_frame": 1545, "end_frame": 1564, "center_frame": 1554, "action": "Grasp coffee scoop", "subtask": "Prepare coffee equipment and scoop grounds", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.943562, "mean_abs": 0.224757, "relative": 0.838782 }, { "name": "hand_right_joints", "l2": 10.838521, "mean_abs": 0.280017, "relative": 0.854365 }, { "name": "body_joints", "l2": 6.395409, "mean_abs": 0.101687, "relative": 0.863823 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.029906, "mean_abs": 0.004715, "relative": 0.070574 }, { "name": "camera_rotation_matrix", "l2": 3.001377, "mean_abs": 0.193938, "relative": 0.921674 }, { "name": "imu_accel_gyro", "l2": 17.099279, "mean_abs": 1.046422, "relative": 0.922565 }, { "name": "depth_confidence", "l2": 9.945483, "mean_abs": 0.18641, "relative": 0.00033 }, { "name": "video_fisheye_cam0", "l2": 5.168274, "mean_abs": 0.106787, "relative": 0.935864 }, { "name": "video_fisheye_cam1", "l2": 5.307956, "mean_abs": 0.112612, "relative": 0.877095 }, { "name": "video_fisheye_cam2", "l2": 5.528749, "mean_abs": 0.112579, "relative": 0.882583 }, { "name": "video_fisheye_cam3", "l2": 5.516975, "mean_abs": 0.11427, "relative": 0.923751 }, { "name": "video_stereo_left", "l2": 5.212846, "mean_abs": 0.114572, "relative": 0.847098 }, { "name": "video_stereo_right", "l2": 6.355554, "mean_abs": 0.140807, "relative": 0.896 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.005167, "mean_abs": 0.07918, "relative": 0.717619 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 310, "start_frame": 1550, "end_frame": 1569, "center_frame": 1559, "action": "", "subtask": "", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.965857, "mean_abs": 0.223404, "relative": 0.840663 }, { "name": "hand_right_joints", "l2": 11.007881, "mean_abs": 0.279432, "relative": 0.867716 }, { "name": "body_joints", "l2": 6.345152, "mean_abs": 0.09909, "relative": 0.857035 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017323, "mean_abs": 0.002666, "relative": 0.04088 }, { "name": "camera_rotation_matrix", "l2": 3.000128, "mean_abs": 0.191226, "relative": 0.92129 }, { "name": "imu_accel_gyro", "l2": 17.095135, "mean_abs": 1.036839, "relative": 0.922341 }, { "name": "depth_confidence", "l2": 9.926948, "mean_abs": 0.184401, "relative": 0.000329 }, { "name": "video_fisheye_cam0", "l2": 5.171772, "mean_abs": 0.106617, "relative": 0.936497 }, { "name": "video_fisheye_cam1", "l2": 5.303821, "mean_abs": 0.111978, "relative": 0.876412 }, { "name": "video_fisheye_cam2", "l2": 5.548245, "mean_abs": 0.112631, "relative": 0.885695 }, { "name": "video_fisheye_cam3", "l2": 5.526028, "mean_abs": 0.113076, "relative": 0.925267 }, { "name": "video_stereo_left", "l2": 5.241952, "mean_abs": 0.113917, "relative": 0.851828 }, { "name": "video_stereo_right", "l2": 6.370038, "mean_abs": 0.139139, "relative": 0.898042 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.06622, "mean_abs": 0.082078, "relative": 0.739469 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 311, "start_frame": 1555, "end_frame": 1574, "center_frame": 1564, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.016285, "mean_abs": 0.221349, "relative": 0.844916 }, { "name": "hand_right_joints", "l2": 11.008274, "mean_abs": 0.275454, "relative": 0.867746 }, { "name": "body_joints", "l2": 6.280374, "mean_abs": 0.096579, "relative": 0.848285 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.025432, "mean_abs": 0.003633, "relative": 0.060016 }, { "name": "camera_rotation_matrix", "l2": 3.000521, "mean_abs": 0.191801, "relative": 0.921411 }, { "name": "imu_accel_gyro", "l2": 17.099344, "mean_abs": 1.020759, "relative": 0.922569 }, { "name": "depth_confidence", "l2": 9.969714, "mean_abs": 0.185042, "relative": 0.000331 }, { "name": "video_fisheye_cam0", "l2": 5.180669, "mean_abs": 0.106652, "relative": 0.938108 }, { "name": "video_fisheye_cam1", "l2": 5.314883, "mean_abs": 0.112019, "relative": 0.87824 }, { "name": "video_fisheye_cam2", "l2": 5.56513, "mean_abs": 0.112618, "relative": 0.88839 }, { "name": "video_fisheye_cam3", "l2": 5.551154, "mean_abs": 0.113476, "relative": 0.929474 }, { "name": "video_stereo_left", "l2": 5.269165, "mean_abs": 0.11379, "relative": 0.85625 }, { "name": "video_stereo_right", "l2": 6.394454, "mean_abs": 0.139293, "relative": 0.901484 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.09347, "mean_abs": 0.08171, "relative": 0.749221 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 312, "start_frame": 1560, "end_frame": 1579, "center_frame": 1569, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.034134, "mean_abs": 0.220309, "relative": 0.846422 }, { "name": "hand_right_joints", "l2": 10.888855, "mean_abs": 0.271311, "relative": 0.858333 }, { "name": "body_joints", "l2": 6.247761, "mean_abs": 0.096375, "relative": 0.84388 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028747, "mean_abs": 0.00401, "relative": 0.06784 }, { "name": "camera_rotation_matrix", "l2": 3.001107, "mean_abs": 0.192386, "relative": 0.921591 }, { "name": "imu_accel_gyro", "l2": 17.100304, "mean_abs": 1.02676, "relative": 0.92262 }, { "name": "depth_confidence", "l2": 10.000686, "mean_abs": 0.187228, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.180618, "mean_abs": 0.106717, "relative": 0.938099 }, { "name": "video_fisheye_cam1", "l2": 5.32185, "mean_abs": 0.112227, "relative": 0.879391 }, { "name": "video_fisheye_cam2", "l2": 5.579046, "mean_abs": 0.112635, "relative": 0.890612 }, { "name": "video_fisheye_cam3", "l2": 5.54752, "mean_abs": 0.114134, "relative": 0.928865 }, { "name": "video_stereo_left", "l2": 5.273148, "mean_abs": 0.114281, "relative": 0.856897 }, { "name": "video_stereo_right", "l2": 6.390219, "mean_abs": 0.140416, "relative": 0.900887 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.167369, "mean_abs": 0.088473, "relative": 0.775669 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 313, "start_frame": 1565, "end_frame": 1584, "center_frame": 1574, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.04676, "mean_abs": 0.22149, "relative": 0.847487 }, { "name": "hand_right_joints", "l2": 10.673646, "mean_abs": 0.271933, "relative": 0.841369 }, { "name": "body_joints", "l2": 6.2578, "mean_abs": 0.099642, "relative": 0.845236 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.041334, "mean_abs": 0.00565, "relative": 0.097544 }, { "name": "camera_rotation_matrix", "l2": 2.999669, "mean_abs": 0.195236, "relative": 0.92115 }, { "name": "imu_accel_gyro", "l2": 17.098011, "mean_abs": 1.036942, "relative": 0.922497 }, { "name": "depth_confidence", "l2": 10.076504, "mean_abs": 0.188884, "relative": 0.000334 }, { "name": "video_fisheye_cam0", "l2": 5.179888, "mean_abs": 0.106884, "relative": 0.937967 }, { "name": "video_fisheye_cam1", "l2": 5.336765, "mean_abs": 0.113163, "relative": 0.881855 }, { "name": "video_fisheye_cam2", "l2": 5.606157, "mean_abs": 0.113699, "relative": 0.89494 }, { "name": "video_fisheye_cam3", "l2": 5.542771, "mean_abs": 0.114031, "relative": 0.92807 }, { "name": "video_stereo_left", "l2": 5.299361, "mean_abs": 0.115728, "relative": 0.861157 }, { "name": "video_stereo_right", "l2": 6.42015, "mean_abs": 0.142131, "relative": 0.905106 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.161027, "mean_abs": 0.085285, "relative": 0.773399 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 314, "start_frame": 1570, "end_frame": 1589, "center_frame": 1579, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.062403, "mean_abs": 0.223539, "relative": 0.848807 }, { "name": "hand_right_joints", "l2": 10.423875, "mean_abs": 0.270383, "relative": 0.82168 }, { "name": "body_joints", "l2": 6.23438, "mean_abs": 0.101985, "relative": 0.842073 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.058935, "mean_abs": 0.007532, "relative": 0.139077 }, { "name": "camera_rotation_matrix", "l2": 3.003469, "mean_abs": 0.202316, "relative": 0.922317 }, { "name": "imu_accel_gyro", "l2": 17.080685, "mean_abs": 1.04817, "relative": 0.921562 }, { "name": "depth_confidence", "l2": 10.184472, "mean_abs": 0.192944, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.188527, "mean_abs": 0.108249, "relative": 0.939531 }, { "name": "video_fisheye_cam1", "l2": 5.355034, "mean_abs": 0.116366, "relative": 0.884874 }, { "name": "video_fisheye_cam2", "l2": 5.653528, "mean_abs": 0.116536, "relative": 0.902502 }, { "name": "video_fisheye_cam3", "l2": 5.514534, "mean_abs": 0.114067, "relative": 0.923342 }, { "name": "video_stereo_left", "l2": 5.371687, "mean_abs": 0.119357, "relative": 0.87291 }, { "name": "video_stereo_right", "l2": 6.517033, "mean_abs": 0.148129, "relative": 0.918765 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.136534, "mean_abs": 0.084634, "relative": 0.764633 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 315, "start_frame": 1575, "end_frame": 1594, "center_frame": 1584, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.086824, "mean_abs": 0.226632, "relative": 0.850867 }, { "name": "hand_right_joints", "l2": 10.22071, "mean_abs": 0.263521, "relative": 0.805665 }, { "name": "body_joints", "l2": 6.192294, "mean_abs": 0.101358, "relative": 0.836388 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.071566, "mean_abs": 0.008935, "relative": 0.168885 }, { "name": "camera_rotation_matrix", "l2": 3.004954, "mean_abs": 0.20505, "relative": 0.922772 }, { "name": "imu_accel_gyro", "l2": 17.108786, "mean_abs": 1.043908, "relative": 0.923078 }, { "name": "depth_confidence", "l2": 10.268126, "mean_abs": 0.195031, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.192287, "mean_abs": 0.10886, "relative": 0.940212 }, { "name": "video_fisheye_cam1", "l2": 5.36957, "mean_abs": 0.117794, "relative": 0.887276 }, { "name": "video_fisheye_cam2", "l2": 5.674819, "mean_abs": 0.117608, "relative": 0.9059 }, { "name": "video_fisheye_cam3", "l2": 5.491543, "mean_abs": 0.113607, "relative": 0.919493 }, { "name": "video_stereo_left", "l2": 5.404832, "mean_abs": 0.121133, "relative": 0.878296 }, { "name": "video_stereo_right", "l2": 6.571865, "mean_abs": 0.150689, "relative": 0.926495 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.132881, "mean_abs": 0.087589, "relative": 0.763326 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 316, "start_frame": 1580, "end_frame": 1599, "center_frame": 1589, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee jar", "coffee scoop" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.10295, "mean_abs": 0.229142, "relative": 0.852227 }, { "name": "hand_right_joints", "l2": 10.069243, "mean_abs": 0.257164, "relative": 0.793726 }, { "name": "body_joints", "l2": 6.170106, "mean_abs": 0.10072, "relative": 0.833391 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.077076, "mean_abs": 0.009623, "relative": 0.181888 }, { "name": "camera_rotation_matrix", "l2": 3.004536, "mean_abs": 0.205158, "relative": 0.922644 }, { "name": "imu_accel_gyro", "l2": 17.016916, "mean_abs": 1.040769, "relative": 0.918121 }, { "name": "depth_confidence", "l2": 10.351753, "mean_abs": 0.195907, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.194406, "mean_abs": 0.108817, "relative": 0.940596 }, { "name": "video_fisheye_cam1", "l2": 5.379342, "mean_abs": 0.117946, "relative": 0.888891 }, { "name": "video_fisheye_cam2", "l2": 5.687397, "mean_abs": 0.11775, "relative": 0.907908 }, { "name": "video_fisheye_cam3", "l2": 5.48561, "mean_abs": 0.113045, "relative": 0.918499 }, { "name": "video_stereo_left", "l2": 5.427983, "mean_abs": 0.121742, "relative": 0.882058 }, { "name": "video_stereo_right", "l2": 6.610453, "mean_abs": 0.151536, "relative": 0.931935 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.151666, "mean_abs": 0.086293, "relative": 0.770049 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007892, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 317, "start_frame": 1585, "end_frame": 1604, "center_frame": 1594, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.121559, "mean_abs": 0.231978, "relative": 0.853797 }, { "name": "hand_right_joints", "l2": 9.779913, "mean_abs": 0.241728, "relative": 0.770919 }, { "name": "body_joints", "l2": 6.120163, "mean_abs": 0.100224, "relative": 0.826645 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.058323, "mean_abs": 0.00726, "relative": 0.137633 }, { "name": "camera_rotation_matrix", "l2": 3.000958, "mean_abs": 0.201122, "relative": 0.921546 }, { "name": "imu_accel_gyro", "l2": 17.014576, "mean_abs": 1.028716, "relative": 0.917995 }, { "name": "depth_confidence", "l2": 10.401424, "mean_abs": 0.193729, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.191211, "mean_abs": 0.107711, "relative": 0.940017 }, { "name": "video_fisheye_cam1", "l2": 5.381646, "mean_abs": 0.116221, "relative": 0.889271 }, { "name": "video_fisheye_cam2", "l2": 5.715598, "mean_abs": 0.11667, "relative": 0.91241 }, { "name": "video_fisheye_cam3", "l2": 5.474729, "mean_abs": 0.111906, "relative": 0.916677 }, { "name": "video_stereo_left", "l2": 5.478889, "mean_abs": 0.119589, "relative": 0.89033 }, { "name": "video_stereo_right", "l2": 6.640822, "mean_abs": 0.148117, "relative": 0.936217 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.110059, "mean_abs": 0.086139, "relative": 0.755158 }, { "name": "caption_objects_interaction_text", "l2": 1.939118, "mean_abs": 0.012593, "relative": 0.890396 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 318, "start_frame": 1590, "end_frame": 1609, "center_frame": 1599, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.156642, "mean_abs": 0.234835, "relative": 0.856756 }, { "name": "hand_right_joints", "l2": 9.827621, "mean_abs": 0.246606, "relative": 0.774679 }, { "name": "body_joints", "l2": 6.11183, "mean_abs": 0.101, "relative": 0.82552 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.034644, "mean_abs": 0.005257, "relative": 0.081756 }, { "name": "camera_rotation_matrix", "l2": 3.000851, "mean_abs": 0.202351, "relative": 0.921513 }, { "name": "imu_accel_gyro", "l2": 17.033859, "mean_abs": 1.035731, "relative": 0.919035 }, { "name": "depth_confidence", "l2": 10.467659, "mean_abs": 0.194317, "relative": 0.000347 }, { "name": "video_fisheye_cam0", "l2": 5.183431, "mean_abs": 0.107535, "relative": 0.938609 }, { "name": "video_fisheye_cam1", "l2": 5.407347, "mean_abs": 0.11653, "relative": 0.893518 }, { "name": "video_fisheye_cam2", "l2": 5.723925, "mean_abs": 0.116688, "relative": 0.91374 }, { "name": "video_fisheye_cam3", "l2": 5.47927, "mean_abs": 0.113033, "relative": 0.917438 }, { "name": "video_stereo_left", "l2": 5.485277, "mean_abs": 0.119648, "relative": 0.891368 }, { "name": "video_stereo_right", "l2": 6.680797, "mean_abs": 0.14819, "relative": 0.941852 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.137129, "mean_abs": 0.084782, "relative": 0.764846 }, { "name": "caption_objects_interaction_text", "l2": 1.939118, "mean_abs": 0.012763, "relative": 0.890396 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 319, "start_frame": 1595, "end_frame": 1614, "center_frame": 1604, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee jar", "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.114859, "mean_abs": 0.236734, "relative": 0.853231 }, { "name": "hand_right_joints", "l2": 9.960093, "mean_abs": 0.24884, "relative": 0.785122 }, { "name": "body_joints", "l2": 6.181828, "mean_abs": 0.102423, "relative": 0.834974 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.029319, "mean_abs": 0.004066, "relative": 0.069188 }, { "name": "camera_rotation_matrix", "l2": 3.002984, "mean_abs": 0.20458, "relative": 0.922168 }, { "name": "imu_accel_gyro", "l2": 17.016598, "mean_abs": 1.013774, "relative": 0.918104 }, { "name": "depth_confidence", "l2": 10.492785, "mean_abs": 0.196256, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.182158, "mean_abs": 0.108053, "relative": 0.938378 }, { "name": "video_fisheye_cam1", "l2": 5.446675, "mean_abs": 0.118173, "relative": 0.900017 }, { "name": "video_fisheye_cam2", "l2": 5.717557, "mean_abs": 0.117306, "relative": 0.912723 }, { "name": "video_fisheye_cam3", "l2": 5.481941, "mean_abs": 0.11372, "relative": 0.917885 }, { "name": "video_stereo_left", "l2": 5.467828, "mean_abs": 0.120814, "relative": 0.888533 }, { "name": "video_stereo_right", "l2": 6.702965, "mean_abs": 0.150492, "relative": 0.944977 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.146775, "mean_abs": 0.083951, "relative": 0.768299 }, { "name": "caption_objects_interaction_text", "l2": 1.939118, "mean_abs": 0.012558, "relative": 0.890396 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 320, "start_frame": 1600, "end_frame": 1619, "center_frame": 1609, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.100854, "mean_abs": 0.239756, "relative": 0.85205 }, { "name": "hand_right_joints", "l2": 9.995376, "mean_abs": 0.250678, "relative": 0.787903 }, { "name": "body_joints", "l2": 6.251993, "mean_abs": 0.103426, "relative": 0.844452 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.032974, "mean_abs": 0.004697, "relative": 0.077813 }, { "name": "camera_rotation_matrix", "l2": 3.003166, "mean_abs": 0.203751, "relative": 0.922223 }, { "name": "imu_accel_gyro", "l2": 17.027809, "mean_abs": 1.001617, "relative": 0.918709 }, { "name": "depth_confidence", "l2": 10.485123, "mean_abs": 0.195593, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.180057, "mean_abs": 0.10796, "relative": 0.937998 }, { "name": "video_fisheye_cam1", "l2": 5.461693, "mean_abs": 0.118262, "relative": 0.902499 }, { "name": "video_fisheye_cam2", "l2": 5.702826, "mean_abs": 0.11714, "relative": 0.910371 }, { "name": "video_fisheye_cam3", "l2": 5.485002, "mean_abs": 0.113299, "relative": 0.918397 }, { "name": "video_stereo_left", "l2": 5.435689, "mean_abs": 0.120469, "relative": 0.88331 }, { "name": "video_stereo_right", "l2": 6.687142, "mean_abs": 0.150359, "relative": 0.942747 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.14377, "mean_abs": 0.083764, "relative": 0.767223 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 321, "start_frame": 1605, "end_frame": 1624, "center_frame": 1614, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.097185, "mean_abs": 0.240576, "relative": 0.851741 }, { "name": "hand_right_joints", "l2": 10.013625, "mean_abs": 0.252235, "relative": 0.789342 }, { "name": "body_joints", "l2": 6.329465, "mean_abs": 0.102986, "relative": 0.854916 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.035772, "mean_abs": 0.005055, "relative": 0.084417 }, { "name": "camera_rotation_matrix", "l2": 3.002495, "mean_abs": 0.202141, "relative": 0.922017 }, { "name": "imu_accel_gyro", "l2": 17.048487, "mean_abs": 1.003391, "relative": 0.919825 }, { "name": "depth_confidence", "l2": 10.480719, "mean_abs": 0.19529, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.178041, "mean_abs": 0.107728, "relative": 0.937632 }, { "name": "video_fisheye_cam1", "l2": 5.465432, "mean_abs": 0.11769, "relative": 0.903116 }, { "name": "video_fisheye_cam2", "l2": 5.69174, "mean_abs": 0.116671, "relative": 0.908602 }, { "name": "video_fisheye_cam3", "l2": 5.47159, "mean_abs": 0.112394, "relative": 0.916152 }, { "name": "video_stereo_left", "l2": 5.407052, "mean_abs": 0.119604, "relative": 0.878657 }, { "name": "video_stereo_right", "l2": 6.657121, "mean_abs": 0.14901, "relative": 0.938514 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.139526, "mean_abs": 0.083014, "relative": 0.765704 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 322, "start_frame": 1610, "end_frame": 1629, "center_frame": 1619, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.054605, "mean_abs": 0.238669, "relative": 0.848149 }, { "name": "hand_right_joints", "l2": 10.034742, "mean_abs": 0.252723, "relative": 0.791006 }, { "name": "body_joints", "l2": 6.387969, "mean_abs": 0.101526, "relative": 0.862818 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.029379, "mean_abs": 0.004001, "relative": 0.069331 }, { "name": "camera_rotation_matrix", "l2": 3.0012, "mean_abs": 0.199981, "relative": 0.92162 }, { "name": "imu_accel_gyro", "l2": 17.03702, "mean_abs": 0.997982, "relative": 0.919206 }, { "name": "depth_confidence", "l2": 10.478981, "mean_abs": 0.194497, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.175067, "mean_abs": 0.107234, "relative": 0.937094 }, { "name": "video_fisheye_cam1", "l2": 5.466998, "mean_abs": 0.11639, "relative": 0.903375 }, { "name": "video_fisheye_cam2", "l2": 5.678377, "mean_abs": 0.115602, "relative": 0.906468 }, { "name": "video_fisheye_cam3", "l2": 5.450559, "mean_abs": 0.11104, "relative": 0.91263 }, { "name": "video_stereo_left", "l2": 5.36108, "mean_abs": 0.117468, "relative": 0.871186 }, { "name": "video_stereo_right", "l2": 6.610531, "mean_abs": 0.14624, "relative": 0.931946 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.033248, "mean_abs": 0.078233, "relative": 0.727669 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 323, "start_frame": 1615, "end_frame": 1634, "center_frame": 1624, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.051064, "mean_abs": 0.237003, "relative": 0.84785 }, { "name": "hand_right_joints", "l2": 10.022782, "mean_abs": 0.251023, "relative": 0.790063 }, { "name": "body_joints", "l2": 6.435536, "mean_abs": 0.099652, "relative": 0.869243 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017504, "mean_abs": 0.002333, "relative": 0.041307 }, { "name": "camera_rotation_matrix", "l2": 3.000189, "mean_abs": 0.197339, "relative": 0.921309 }, { "name": "imu_accel_gyro", "l2": 17.025349, "mean_abs": 0.99409, "relative": 0.918576 }, { "name": "depth_confidence", "l2": 10.493918, "mean_abs": 0.192729, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.174287, "mean_abs": 0.106671, "relative": 0.936953 }, { "name": "video_fisheye_cam1", "l2": 5.467992, "mean_abs": 0.115155, "relative": 0.903539 }, { "name": "video_fisheye_cam2", "l2": 5.661502, "mean_abs": 0.114445, "relative": 0.903775 }, { "name": "video_fisheye_cam3", "l2": 5.442255, "mean_abs": 0.110156, "relative": 0.91124 }, { "name": "video_stereo_left", "l2": 5.314908, "mean_abs": 0.115021, "relative": 0.863683 }, { "name": "video_stereo_right", "l2": 6.567633, "mean_abs": 0.143199, "relative": 0.925898 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.071422, "mean_abs": 0.078455, "relative": 0.741331 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 324, "start_frame": 1620, "end_frame": 1639, "center_frame": 1629, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.060094, "mean_abs": 0.236383, "relative": 0.848612 }, { "name": "hand_right_joints", "l2": 9.977026, "mean_abs": 0.249751, "relative": 0.786457 }, { "name": "body_joints", "l2": 6.436682, "mean_abs": 0.099102, "relative": 0.869397 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.022522, "mean_abs": 0.0031, "relative": 0.053149 }, { "name": "camera_rotation_matrix", "l2": 3.000417, "mean_abs": 0.196972, "relative": 0.92138 }, { "name": "imu_accel_gyro", "l2": 17.011499, "mean_abs": 0.991476, "relative": 0.917829 }, { "name": "depth_confidence", "l2": 10.469934, "mean_abs": 0.192914, "relative": 0.000347 }, { "name": "video_fisheye_cam0", "l2": 5.171324, "mean_abs": 0.106653, "relative": 0.936416 }, { "name": "video_fisheye_cam1", "l2": 5.449103, "mean_abs": 0.114711, "relative": 0.900418 }, { "name": "video_fisheye_cam2", "l2": 5.644621, "mean_abs": 0.114107, "relative": 0.90108 }, { "name": "video_fisheye_cam3", "l2": 5.426808, "mean_abs": 0.109889, "relative": 0.908654 }, { "name": "video_stereo_left", "l2": 5.287911, "mean_abs": 0.114494, "relative": 0.859296 }, { "name": "video_stereo_right", "l2": 6.522501, "mean_abs": 0.142214, "relative": 0.919536 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.102248, "mean_abs": 0.080071, "relative": 0.752363 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 325, "start_frame": 1625, "end_frame": 1644, "center_frame": 1634, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.036658, "mean_abs": 0.236521, "relative": 0.846635 }, { "name": "hand_right_joints", "l2": 9.954966, "mean_abs": 0.249972, "relative": 0.784718 }, { "name": "body_joints", "l2": 6.424047, "mean_abs": 0.099911, "relative": 0.867691 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.042726, "mean_abs": 0.005516, "relative": 0.100826 }, { "name": "camera_rotation_matrix", "l2": 3.00076, "mean_abs": 0.197254, "relative": 0.921485 }, { "name": "imu_accel_gyro", "l2": 17.053848, "mean_abs": 0.998972, "relative": 0.920114 }, { "name": "depth_confidence", "l2": 10.412109, "mean_abs": 0.193151, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.169271, "mean_abs": 0.10667, "relative": 0.936044 }, { "name": "video_fisheye_cam1", "l2": 5.426406, "mean_abs": 0.114602, "relative": 0.896668 }, { "name": "video_fisheye_cam2", "l2": 5.625192, "mean_abs": 0.114232, "relative": 0.897978 }, { "name": "video_fisheye_cam3", "l2": 5.427104, "mean_abs": 0.110465, "relative": 0.908703 }, { "name": "video_stereo_left", "l2": 5.264919, "mean_abs": 0.11498, "relative": 0.85556 }, { "name": "video_stereo_right", "l2": 6.487132, "mean_abs": 0.142898, "relative": 0.914549 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.123431, "mean_abs": 0.080467, "relative": 0.759944 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 326, "start_frame": 1630, "end_frame": 1649, "center_frame": 1639, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.038606, "mean_abs": 0.235843, "relative": 0.846799 }, { "name": "hand_right_joints", "l2": 9.944457, "mean_abs": 0.248981, "relative": 0.783889 }, { "name": "body_joints", "l2": 6.412119, "mean_abs": 0.099944, "relative": 0.86608 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.063808, "mean_abs": 0.008798, "relative": 0.150579 }, { "name": "camera_rotation_matrix", "l2": 3.000546, "mean_abs": 0.196894, "relative": 0.921419 }, { "name": "imu_accel_gyro", "l2": 17.052307, "mean_abs": 1.01495, "relative": 0.920031 }, { "name": "depth_confidence", "l2": 10.372247, "mean_abs": 0.193246, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.169514, "mean_abs": 0.10676, "relative": 0.936088 }, { "name": "video_fisheye_cam1", "l2": 5.4098, "mean_abs": 0.114614, "relative": 0.893924 }, { "name": "video_fisheye_cam2", "l2": 5.614957, "mean_abs": 0.114093, "relative": 0.896344 }, { "name": "video_fisheye_cam3", "l2": 5.436739, "mean_abs": 0.111585, "relative": 0.910316 }, { "name": "video_stereo_left", "l2": 5.255709, "mean_abs": 0.115121, "relative": 0.854063 }, { "name": "video_stereo_right", "l2": 6.460169, "mean_abs": 0.143216, "relative": 0.910748 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.050643, "mean_abs": 0.07806, "relative": 0.733894 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 327, "start_frame": 1635, "end_frame": 1654, "center_frame": 1644, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.028485, "mean_abs": 0.232567, "relative": 0.845946 }, { "name": "hand_right_joints", "l2": 10.026482, "mean_abs": 0.252471, "relative": 0.790355 }, { "name": "body_joints", "l2": 6.423757, "mean_abs": 0.099793, "relative": 0.867652 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.084833, "mean_abs": 0.011856, "relative": 0.200194 }, { "name": "camera_rotation_matrix", "l2": 3.001143, "mean_abs": 0.196323, "relative": 0.921602 }, { "name": "imu_accel_gyro", "l2": 17.044888, "mean_abs": 1.019957, "relative": 0.91963 }, { "name": "depth_confidence", "l2": 10.261769, "mean_abs": 0.192534, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.170312, "mean_abs": 0.10685, "relative": 0.936233 }, { "name": "video_fisheye_cam1", "l2": 5.37733, "mean_abs": 0.114241, "relative": 0.888558 }, { "name": "video_fisheye_cam2", "l2": 5.601677, "mean_abs": 0.114053, "relative": 0.894224 }, { "name": "video_fisheye_cam3", "l2": 5.450994, "mean_abs": 0.11279, "relative": 0.912703 }, { "name": "video_stereo_left", "l2": 5.251196, "mean_abs": 0.11527, "relative": 0.85333 }, { "name": "video_stereo_right", "l2": 6.419919, "mean_abs": 0.142733, "relative": 0.905074 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.996616, "mean_abs": 0.077378, "relative": 0.714559 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 328, "start_frame": 1640, "end_frame": 1659, "center_frame": 1649, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.0147, "mean_abs": 0.230178, "relative": 0.844783 }, { "name": "hand_right_joints", "l2": 10.391981, "mean_abs": 0.272834, "relative": 0.819166 }, { "name": "body_joints", "l2": 6.463723, "mean_abs": 0.102081, "relative": 0.87305 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.088587, "mean_abs": 0.012609, "relative": 0.209054 }, { "name": "camera_rotation_matrix", "l2": 3.001139, "mean_abs": 0.194297, "relative": 0.921601 }, { "name": "imu_accel_gyro", "l2": 17.089165, "mean_abs": 1.04467, "relative": 0.922019 }, { "name": "depth_confidence", "l2": 10.069775, "mean_abs": 0.189929, "relative": 0.000334 }, { "name": "video_fisheye_cam0", "l2": 5.171799, "mean_abs": 0.107082, "relative": 0.936502 }, { "name": "video_fisheye_cam1", "l2": 5.339915, "mean_abs": 0.113371, "relative": 0.882376 }, { "name": "video_fisheye_cam2", "l2": 5.586359, "mean_abs": 0.113367, "relative": 0.891779 }, { "name": "video_fisheye_cam3", "l2": 5.491742, "mean_abs": 0.115292, "relative": 0.919526 }, { "name": "video_stereo_left", "l2": 5.240882, "mean_abs": 0.114989, "relative": 0.851654 }, { "name": "video_stereo_right", "l2": 6.3723, "mean_abs": 0.141727, "relative": 0.898361 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.024728, "mean_abs": 0.079735, "relative": 0.72462 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 329, "start_frame": 1645, "end_frame": 1664, "center_frame": 1654, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.016138, "mean_abs": 0.2291, "relative": 0.844904 }, { "name": "hand_right_joints", "l2": 10.496188, "mean_abs": 0.2761, "relative": 0.82738 }, { "name": "body_joints", "l2": 6.465413, "mean_abs": 0.102645, "relative": 0.873278 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.075963, "mean_abs": 0.011183, "relative": 0.179263 }, { "name": "camera_rotation_matrix", "l2": 3.001742, "mean_abs": 0.193103, "relative": 0.921786 }, { "name": "imu_accel_gyro", "l2": 17.148989, "mean_abs": 1.05043, "relative": 0.925247 }, { "name": "depth_confidence", "l2": 9.969829, "mean_abs": 0.188022, "relative": 0.000331 }, { "name": "video_fisheye_cam0", "l2": 5.172781, "mean_abs": 0.107175, "relative": 0.93668 }, { "name": "video_fisheye_cam1", "l2": 5.301836, "mean_abs": 0.112834, "relative": 0.876084 }, { "name": "video_fisheye_cam2", "l2": 5.576889, "mean_abs": 0.112999, "relative": 0.890267 }, { "name": "video_fisheye_cam3", "l2": 5.506737, "mean_abs": 0.115964, "relative": 0.922037 }, { "name": "video_stereo_left", "l2": 5.250811, "mean_abs": 0.115543, "relative": 0.853267 }, { "name": "video_stereo_right", "l2": 6.350329, "mean_abs": 0.141965, "relative": 0.895263 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.035164, "mean_abs": 0.079286, "relative": 0.728355 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 330, "start_frame": 1650, "end_frame": 1669, "center_frame": 1659, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.035883, "mean_abs": 0.231104, "relative": 0.84657 }, { "name": "hand_right_joints", "l2": 10.640776, "mean_abs": 0.280887, "relative": 0.838778 }, { "name": "body_joints", "l2": 6.464356, "mean_abs": 0.103752, "relative": 0.873135 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.05933, "mean_abs": 0.008978, "relative": 0.14001 }, { "name": "camera_rotation_matrix", "l2": 3.001591, "mean_abs": 0.192388, "relative": 0.92174 }, { "name": "imu_accel_gyro", "l2": 17.054657, "mean_abs": 1.054782, "relative": 0.920157 }, { "name": "depth_confidence", "l2": 9.905505, "mean_abs": 0.186951, "relative": 0.000329 }, { "name": "video_fisheye_cam0", "l2": 5.172852, "mean_abs": 0.107043, "relative": 0.936693 }, { "name": "video_fisheye_cam1", "l2": 5.283834, "mean_abs": 0.112294, "relative": 0.873109 }, { "name": "video_fisheye_cam2", "l2": 5.57468, "mean_abs": 0.112789, "relative": 0.889915 }, { "name": "video_fisheye_cam3", "l2": 5.512646, "mean_abs": 0.1153, "relative": 0.923026 }, { "name": "video_stereo_left", "l2": 5.259409, "mean_abs": 0.1155, "relative": 0.854665 }, { "name": "video_stereo_right", "l2": 6.341077, "mean_abs": 0.141399, "relative": 0.893959 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.057243, "mean_abs": 0.080864, "relative": 0.736257 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 331, "start_frame": 1655, "end_frame": 1674, "center_frame": 1664, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.055925, "mean_abs": 0.232191, "relative": 0.84826 }, { "name": "hand_right_joints", "l2": 10.858197, "mean_abs": 0.282514, "relative": 0.855916 }, { "name": "body_joints", "l2": 6.412377, "mean_abs": 0.102816, "relative": 0.866115 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.035382, "mean_abs": 0.005561, "relative": 0.083496 }, { "name": "camera_rotation_matrix", "l2": 3.001286, "mean_abs": 0.190684, "relative": 0.921646 }, { "name": "imu_accel_gyro", "l2": 17.065611, "mean_abs": 1.04177, "relative": 0.920748 }, { "name": "depth_confidence", "l2": 9.774107, "mean_abs": 0.183021, "relative": 0.000324 }, { "name": "video_fisheye_cam0", "l2": 5.167293, "mean_abs": 0.106761, "relative": 0.935686 }, { "name": "video_fisheye_cam1", "l2": 5.25737, "mean_abs": 0.111329, "relative": 0.868736 }, { "name": "video_fisheye_cam2", "l2": 5.56742, "mean_abs": 0.112427, "relative": 0.888756 }, { "name": "video_fisheye_cam3", "l2": 5.50854, "mean_abs": 0.114522, "relative": 0.922339 }, { "name": "video_stereo_left", "l2": 5.264787, "mean_abs": 0.114908, "relative": 0.855538 }, { "name": "video_stereo_right", "l2": 6.31768, "mean_abs": 0.139994, "relative": 0.89066 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.049081, "mean_abs": 0.080296, "relative": 0.733335 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 332, "start_frame": 1660, "end_frame": 1679, "center_frame": 1669, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.072169, "mean_abs": 0.231295, "relative": 0.84963 }, { "name": "hand_right_joints", "l2": 11.02661, "mean_abs": 0.281657, "relative": 0.869192 }, { "name": "body_joints", "l2": 6.353518, "mean_abs": 0.101331, "relative": 0.858164 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016745, "mean_abs": 0.002631, "relative": 0.039515 }, { "name": "camera_rotation_matrix", "l2": 3.000009, "mean_abs": 0.188546, "relative": 0.921254 }, { "name": "imu_accel_gyro", "l2": 17.139107, "mean_abs": 1.030487, "relative": 0.924714 }, { "name": "depth_confidence", "l2": 9.717874, "mean_abs": 0.180117, "relative": 0.000323 }, { "name": "video_fisheye_cam0", "l2": 5.158103, "mean_abs": 0.106237, "relative": 0.934022 }, { "name": "video_fisheye_cam1", "l2": 5.235639, "mean_abs": 0.110267, "relative": 0.865145 }, { "name": "video_fisheye_cam2", "l2": 5.564909, "mean_abs": 0.112016, "relative": 0.888355 }, { "name": "video_fisheye_cam3", "l2": 5.493762, "mean_abs": 0.113098, "relative": 0.919864 }, { "name": "video_stereo_left", "l2": 5.263448, "mean_abs": 0.113527, "relative": 0.855321 }, { "name": "video_stereo_right", "l2": 6.303975, "mean_abs": 0.138109, "relative": 0.888728 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.110009, "mean_abs": 0.081289, "relative": 0.75514 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 333, "start_frame": 1665, "end_frame": 1684, "center_frame": 1674, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.104138, "mean_abs": 0.229226, "relative": 0.852327 }, { "name": "hand_right_joints", "l2": 11.026785, "mean_abs": 0.277494, "relative": 0.869206 }, { "name": "body_joints", "l2": 6.243777, "mean_abs": 0.097695, "relative": 0.843342 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01384, "mean_abs": 0.001985, "relative": 0.03266 }, { "name": "camera_rotation_matrix", "l2": 3.000239, "mean_abs": 0.187524, "relative": 0.921325 }, { "name": "imu_accel_gyro", "l2": 17.062479, "mean_abs": 1.027919, "relative": 0.920579 }, { "name": "depth_confidence", "l2": 9.71221, "mean_abs": 0.179776, "relative": 0.000322 }, { "name": "video_fisheye_cam0", "l2": 5.157877, "mean_abs": 0.106035, "relative": 0.933981 }, { "name": "video_fisheye_cam1", "l2": 5.232468, "mean_abs": 0.109988, "relative": 0.864621 }, { "name": "video_fisheye_cam2", "l2": 5.563609, "mean_abs": 0.111668, "relative": 0.888147 }, { "name": "video_fisheye_cam3", "l2": 5.490766, "mean_abs": 0.113, "relative": 0.919362 }, { "name": "video_stereo_left", "l2": 5.271077, "mean_abs": 0.112924, "relative": 0.856561 }, { "name": "video_stereo_right", "l2": 6.30067, "mean_abs": 0.137383, "relative": 0.888262 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.077039, "mean_abs": 0.080604, "relative": 0.743341 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 334, "start_frame": 1670, "end_frame": 1689, "center_frame": 1679, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.144407, "mean_abs": 0.229244, "relative": 0.855724 }, { "name": "hand_right_joints", "l2": 11.083079, "mean_abs": 0.276353, "relative": 0.873643 }, { "name": "body_joints", "l2": 6.143201, "mean_abs": 0.095617, "relative": 0.829757 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020531, "mean_abs": 0.00316, "relative": 0.048451 }, { "name": "camera_rotation_matrix", "l2": 3.000093, "mean_abs": 0.188374, "relative": 0.92128 }, { "name": "imu_accel_gyro", "l2": 17.083704, "mean_abs": 1.027987, "relative": 0.921725 }, { "name": "depth_confidence", "l2": 9.714969, "mean_abs": 0.180634, "relative": 0.000322 }, { "name": "video_fisheye_cam0", "l2": 5.162936, "mean_abs": 0.106156, "relative": 0.934897 }, { "name": "video_fisheye_cam1", "l2": 5.242817, "mean_abs": 0.110456, "relative": 0.866331 }, { "name": "video_fisheye_cam2", "l2": 5.563778, "mean_abs": 0.111855, "relative": 0.888174 }, { "name": "video_fisheye_cam3", "l2": 5.492891, "mean_abs": 0.113049, "relative": 0.919718 }, { "name": "video_stereo_left", "l2": 5.26084, "mean_abs": 0.113431, "relative": 0.854897 }, { "name": "video_stereo_right", "l2": 6.291358, "mean_abs": 0.137956, "relative": 0.886949 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.120237, "mean_abs": 0.082368, "relative": 0.758801 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 335, "start_frame": 1675, "end_frame": 1694, "center_frame": 1684, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.16942, "mean_abs": 0.228738, "relative": 0.857834 }, { "name": "hand_right_joints", "l2": 9.947792, "mean_abs": 0.29205, "relative": 0.784152 }, { "name": "body_joints", "l2": 6.019132, "mean_abs": 0.094847, "relative": 0.812999 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030637, "mean_abs": 0.004731, "relative": 0.072299 }, { "name": "camera_rotation_matrix", "l2": 3.000618, "mean_abs": 0.18983, "relative": 0.921441 }, { "name": "imu_accel_gyro", "l2": 17.091455, "mean_abs": 1.03961, "relative": 0.922143 }, { "name": "depth_confidence", "l2": 9.753366, "mean_abs": 0.182336, "relative": 0.000324 }, { "name": "video_fisheye_cam0", "l2": 5.167932, "mean_abs": 0.106272, "relative": 0.935802 }, { "name": "video_fisheye_cam1", "l2": 5.24791, "mean_abs": 0.110809, "relative": 0.867173 }, { "name": "video_fisheye_cam2", "l2": 5.564795, "mean_abs": 0.112254, "relative": 0.888337 }, { "name": "video_fisheye_cam3", "l2": 5.476603, "mean_abs": 0.112763, "relative": 0.916991 }, { "name": "video_stereo_left", "l2": 5.254829, "mean_abs": 0.11438, "relative": 0.85392 }, { "name": "video_stereo_right", "l2": 6.281631, "mean_abs": 0.138862, "relative": 0.885578 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.185887, "mean_abs": 0.082755, "relative": 0.782296 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 336, "start_frame": 1680, "end_frame": 1699, "center_frame": 1689, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.198277, "mean_abs": 0.229127, "relative": 0.860268 }, { "name": "hand_right_joints", "l2": 9.31191, "mean_abs": 0.259618, "relative": 0.734028 }, { "name": "body_joints", "l2": 6.040073, "mean_abs": 0.098896, "relative": 0.815828 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.036907, "mean_abs": 0.005468, "relative": 0.087095 }, { "name": "camera_rotation_matrix", "l2": 2.999872, "mean_abs": 0.191896, "relative": 0.921212 }, { "name": "imu_accel_gyro", "l2": 17.073292, "mean_abs": 1.035578, "relative": 0.921163 }, { "name": "depth_confidence", "l2": 9.823405, "mean_abs": 0.184601, "relative": 0.000326 }, { "name": "video_fisheye_cam0", "l2": 5.173217, "mean_abs": 0.106702, "relative": 0.936759 }, { "name": "video_fisheye_cam1", "l2": 5.260523, "mean_abs": 0.111439, "relative": 0.869257 }, { "name": "video_fisheye_cam2", "l2": 5.589857, "mean_abs": 0.113079, "relative": 0.892338 }, { "name": "video_fisheye_cam3", "l2": 5.465346, "mean_abs": 0.112712, "relative": 0.915106 }, { "name": "video_stereo_left", "l2": 5.272887, "mean_abs": 0.115404, "relative": 0.856855 }, { "name": "video_stereo_right", "l2": 6.30702, "mean_abs": 0.140178, "relative": 0.889157 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.278, "mean_abs": 0.095775, "relative": 0.815262 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 337, "start_frame": 1685, "end_frame": 1704, "center_frame": 1694, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.22108, "mean_abs": 0.230407, "relative": 0.862192 }, { "name": "hand_right_joints", "l2": 9.191598, "mean_abs": 0.25514, "relative": 0.724544 }, { "name": "body_joints", "l2": 5.989427, "mean_abs": 0.098853, "relative": 0.808987 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.048337, "mean_abs": 0.006737, "relative": 0.114068 }, { "name": "camera_rotation_matrix", "l2": 3.001225, "mean_abs": 0.197121, "relative": 0.921628 }, { "name": "imu_accel_gyro", "l2": 17.058825, "mean_abs": 1.045882, "relative": 0.920382 }, { "name": "depth_confidence", "l2": 9.94976, "mean_abs": 0.188715, "relative": 0.00033 }, { "name": "video_fisheye_cam0", "l2": 5.174505, "mean_abs": 0.107423, "relative": 0.936992 }, { "name": "video_fisheye_cam1", "l2": 5.271119, "mean_abs": 0.112949, "relative": 0.871008 }, { "name": "video_fisheye_cam2", "l2": 5.620271, "mean_abs": 0.114738, "relative": 0.897193 }, { "name": "video_fisheye_cam3", "l2": 5.434053, "mean_abs": 0.112765, "relative": 0.909867 }, { "name": "video_stereo_left", "l2": 5.307775, "mean_abs": 0.117624, "relative": 0.862524 }, { "name": "video_stereo_right", "l2": 6.35939, "mean_abs": 0.143238, "relative": 0.896541 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.33212, "mean_abs": 0.094491, "relative": 0.834631 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 338, "start_frame": 1690, "end_frame": 1709, "center_frame": 1699, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.24272, "mean_abs": 0.233026, "relative": 0.864017 }, { "name": "hand_right_joints", "l2": 9.016383, "mean_abs": 0.24546, "relative": 0.710732 }, { "name": "body_joints", "l2": 5.932832, "mean_abs": 0.098313, "relative": 0.801343 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.049684, "mean_abs": 0.00663, "relative": 0.117248 }, { "name": "camera_rotation_matrix", "l2": 3.002555, "mean_abs": 0.199804, "relative": 0.922036 }, { "name": "imu_accel_gyro", "l2": 17.066616, "mean_abs": 1.056376, "relative": 0.920803 }, { "name": "depth_confidence", "l2": 10.037729, "mean_abs": 0.190907, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.174463, "mean_abs": 0.107742, "relative": 0.936985 }, { "name": "video_fisheye_cam1", "l2": 5.275879, "mean_abs": 0.113827, "relative": 0.871795 }, { "name": "video_fisheye_cam2", "l2": 5.642022, "mean_abs": 0.115798, "relative": 0.900665 }, { "name": "video_fisheye_cam3", "l2": 5.375977, "mean_abs": 0.111723, "relative": 0.900143 }, { "name": "video_stereo_left", "l2": 5.32668, "mean_abs": 0.118657, "relative": 0.865596 }, { "name": "video_stereo_right", "l2": 6.392567, "mean_abs": 0.144929, "relative": 0.901218 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.324329, "mean_abs": 0.09341, "relative": 0.831842 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 339, "start_frame": 1695, "end_frame": 1714, "center_frame": 1704, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.254094, "mean_abs": 0.2337, "relative": 0.864977 }, { "name": "hand_right_joints", "l2": 8.993509, "mean_abs": 0.249566, "relative": 0.708929 }, { "name": "body_joints", "l2": 5.843029, "mean_abs": 0.095465, "relative": 0.789213 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.043574, "mean_abs": 0.006, "relative": 0.102828 }, { "name": "camera_rotation_matrix", "l2": 3.002039, "mean_abs": 0.199679, "relative": 0.921877 }, { "name": "imu_accel_gyro", "l2": 17.049623, "mean_abs": 1.04517, "relative": 0.919886 }, { "name": "depth_confidence", "l2": 10.126323, "mean_abs": 0.191812, "relative": 0.000336 }, { "name": "video_fisheye_cam0", "l2": 5.1792, "mean_abs": 0.107743, "relative": 0.937842 }, { "name": "video_fisheye_cam1", "l2": 5.282334, "mean_abs": 0.113661, "relative": 0.872861 }, { "name": "video_fisheye_cam2", "l2": 5.659122, "mean_abs": 0.115766, "relative": 0.903395 }, { "name": "video_fisheye_cam3", "l2": 5.360367, "mean_abs": 0.110882, "relative": 0.897529 }, { "name": "video_stereo_left", "l2": 5.343867, "mean_abs": 0.118302, "relative": 0.868389 }, { "name": "video_stereo_right", "l2": 6.415547, "mean_abs": 0.144684, "relative": 0.904458 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.287638, "mean_abs": 0.09233, "relative": 0.818711 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 340, "start_frame": 1700, "end_frame": 1719, "center_frame": 1709, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.281255, "mean_abs": 0.233944, "relative": 0.867268 }, { "name": "hand_right_joints", "l2": 9.918709, "mean_abs": 0.238759, "relative": 0.78186 }, { "name": "body_joints", "l2": 5.847174, "mean_abs": 0.094818, "relative": 0.789773 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.02448, "mean_abs": 0.003329, "relative": 0.057769 }, { "name": "camera_rotation_matrix", "l2": 3.000683, "mean_abs": 0.197466, "relative": 0.921461 }, { "name": "imu_accel_gyro", "l2": 17.070679, "mean_abs": 1.052847, "relative": 0.921022 }, { "name": "depth_confidence", "l2": 10.200148, "mean_abs": 0.189849, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.180233, "mean_abs": 0.107004, "relative": 0.938029 }, { "name": "video_fisheye_cam1", "l2": 5.287439, "mean_abs": 0.112376, "relative": 0.873705 }, { "name": "video_fisheye_cam2", "l2": 5.680888, "mean_abs": 0.114588, "relative": 0.906869 }, { "name": "video_fisheye_cam3", "l2": 5.330605, "mean_abs": 0.109499, "relative": 0.892546 }, { "name": "video_stereo_left", "l2": 5.361157, "mean_abs": 0.115314, "relative": 0.871199 }, { "name": "video_stereo_right", "l2": 6.439579, "mean_abs": 0.141289, "relative": 0.907845 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.417314, "mean_abs": 0.094423, "relative": 0.865121 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 341, "start_frame": 1705, "end_frame": 1724, "center_frame": 1714, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.291662, "mean_abs": 0.233658, "relative": 0.868146 }, { "name": "hand_right_joints", "l2": 10.000814, "mean_abs": 0.239471, "relative": 0.788332 }, { "name": "body_joints", "l2": 5.853828, "mean_abs": 0.095151, "relative": 0.790672 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007402, "mean_abs": 0.001196, "relative": 0.017468 }, { "name": "camera_rotation_matrix", "l2": 3.000483, "mean_abs": 0.198102, "relative": 0.921399 }, { "name": "imu_accel_gyro", "l2": 17.069256, "mean_abs": 1.034354, "relative": 0.920945 }, { "name": "depth_confidence", "l2": 10.231107, "mean_abs": 0.189966, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.180882, "mean_abs": 0.107217, "relative": 0.938147 }, { "name": "video_fisheye_cam1", "l2": 5.300754, "mean_abs": 0.112854, "relative": 0.875905 }, { "name": "video_fisheye_cam2", "l2": 5.682665, "mean_abs": 0.114815, "relative": 0.907153 }, { "name": "video_fisheye_cam3", "l2": 5.340642, "mean_abs": 0.109672, "relative": 0.894226 }, { "name": "video_stereo_left", "l2": 5.357542, "mean_abs": 0.11541, "relative": 0.870611 }, { "name": "video_stereo_right", "l2": 6.458891, "mean_abs": 0.141698, "relative": 0.910568 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.028164, "mean_abs": 0.078881, "relative": 0.72585 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 342, "start_frame": 1710, "end_frame": 1729, "center_frame": 1719, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.297805, "mean_abs": 0.234288, "relative": 0.868664 }, { "name": "hand_right_joints", "l2": 10.064452, "mean_abs": 0.240364, "relative": 0.793348 }, { "name": "body_joints", "l2": 5.842209, "mean_abs": 0.095111, "relative": 0.789102 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018177, "mean_abs": 0.002481, "relative": 0.042896 }, { "name": "camera_rotation_matrix", "l2": 3.001355, "mean_abs": 0.199064, "relative": 0.921668 }, { "name": "imu_accel_gyro", "l2": 17.056751, "mean_abs": 1.034226, "relative": 0.92027 }, { "name": "depth_confidence", "l2": 10.202791, "mean_abs": 0.191449, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.189795, "mean_abs": 0.107746, "relative": 0.939761 }, { "name": "video_fisheye_cam1", "l2": 5.310585, "mean_abs": 0.113586, "relative": 0.877529 }, { "name": "video_fisheye_cam2", "l2": 5.67247, "mean_abs": 0.115371, "relative": 0.905525 }, { "name": "video_fisheye_cam3", "l2": 5.388577, "mean_abs": 0.111601, "relative": 0.902252 }, { "name": "video_stereo_left", "l2": 5.35039, "mean_abs": 0.116997, "relative": 0.869449 }, { "name": "video_stereo_right", "l2": 6.457656, "mean_abs": 0.144167, "relative": 0.910394 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.056334, "mean_abs": 0.079023, "relative": 0.735931 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 343, "start_frame": 1715, "end_frame": 1734, "center_frame": 1724, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.277847, "mean_abs": 0.23265, "relative": 0.86698 }, { "name": "hand_right_joints", "l2": 9.060199, "mean_abs": 0.251382, "relative": 0.714186 }, { "name": "body_joints", "l2": 5.803178, "mean_abs": 0.09513, "relative": 0.783831 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.027746, "mean_abs": 0.003827, "relative": 0.065476 }, { "name": "camera_rotation_matrix", "l2": 3.002003, "mean_abs": 0.199105, "relative": 0.921867 }, { "name": "imu_accel_gyro", "l2": 17.046795, "mean_abs": 1.029954, "relative": 0.919733 }, { "name": "depth_confidence", "l2": 10.148218, "mean_abs": 0.191606, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.191415, "mean_abs": 0.108101, "relative": 0.940054 }, { "name": "video_fisheye_cam1", "l2": 5.321963, "mean_abs": 0.114077, "relative": 0.879409 }, { "name": "video_fisheye_cam2", "l2": 5.649057, "mean_abs": 0.115515, "relative": 0.901788 }, { "name": "video_fisheye_cam3", "l2": 5.430439, "mean_abs": 0.113202, "relative": 0.909261 }, { "name": "video_stereo_left", "l2": 5.32409, "mean_abs": 0.118086, "relative": 0.865175 }, { "name": "video_stereo_right", "l2": 6.441338, "mean_abs": 0.145175, "relative": 0.908094 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.094386, "mean_abs": 0.081015, "relative": 0.749549 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 344, "start_frame": 1720, "end_frame": 1739, "center_frame": 1729, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.256534, "mean_abs": 0.230159, "relative": 0.865182 }, { "name": "hand_right_joints", "l2": 8.781454, "mean_abs": 0.245573, "relative": 0.692213 }, { "name": "body_joints", "l2": 5.794445, "mean_abs": 0.094417, "relative": 0.782651 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030477, "mean_abs": 0.004139, "relative": 0.071922 }, { "name": "camera_rotation_matrix", "l2": 3.0014, "mean_abs": 0.197041, "relative": 0.921681 }, { "name": "imu_accel_gyro", "l2": 17.041271, "mean_abs": 1.026756, "relative": 0.919435 }, { "name": "depth_confidence", "l2": 10.098229, "mean_abs": 0.190002, "relative": 0.000335 }, { "name": "video_fisheye_cam0", "l2": 5.191342, "mean_abs": 0.107674, "relative": 0.940041 }, { "name": "video_fisheye_cam1", "l2": 5.329582, "mean_abs": 0.113701, "relative": 0.880668 }, { "name": "video_fisheye_cam2", "l2": 5.622621, "mean_abs": 0.114663, "relative": 0.897568 }, { "name": "video_fisheye_cam3", "l2": 5.462679, "mean_abs": 0.112651, "relative": 0.91466 }, { "name": "video_stereo_left", "l2": 5.290364, "mean_abs": 0.117472, "relative": 0.859695 }, { "name": "video_stereo_right", "l2": 6.410657, "mean_abs": 0.144148, "relative": 0.903768 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.120102, "mean_abs": 0.080684, "relative": 0.758753 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 345, "start_frame": 1725, "end_frame": 1744, "center_frame": 1734, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.231865, "mean_abs": 0.228488, "relative": 0.863101 }, { "name": "hand_right_joints", "l2": 8.604541, "mean_abs": 0.228415, "relative": 0.678268 }, { "name": "body_joints", "l2": 5.864316, "mean_abs": 0.094377, "relative": 0.792088 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.02736, "mean_abs": 0.003826, "relative": 0.064566 }, { "name": "camera_rotation_matrix", "l2": 3.000592, "mean_abs": 0.194283, "relative": 0.921433 }, { "name": "imu_accel_gyro", "l2": 17.067329, "mean_abs": 1.021853, "relative": 0.920841 }, { "name": "depth_confidence", "l2": 10.063553, "mean_abs": 0.188918, "relative": 0.000334 }, { "name": "video_fisheye_cam0", "l2": 5.192381, "mean_abs": 0.107312, "relative": 0.940229 }, { "name": "video_fisheye_cam1", "l2": 5.329033, "mean_abs": 0.113016, "relative": 0.880578 }, { "name": "video_fisheye_cam2", "l2": 5.607389, "mean_abs": 0.113715, "relative": 0.895136 }, { "name": "video_fisheye_cam3", "l2": 5.484812, "mean_abs": 0.112524, "relative": 0.918366 }, { "name": "video_stereo_left", "l2": 5.269549, "mean_abs": 0.115989, "relative": 0.856312 }, { "name": "video_stereo_right", "l2": 6.376228, "mean_abs": 0.141908, "relative": 0.898914 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.079212, "mean_abs": 0.077127, "relative": 0.744119 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 346, "start_frame": 1730, "end_frame": 1749, "center_frame": 1739, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.215459, "mean_abs": 0.226667, "relative": 0.861718 }, { "name": "hand_right_joints", "l2": 9.189097, "mean_abs": 0.266266, "relative": 0.724347 }, { "name": "body_joints", "l2": 5.868767, "mean_abs": 0.093756, "relative": 0.79269 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020884, "mean_abs": 0.003213, "relative": 0.049282 }, { "name": "camera_rotation_matrix", "l2": 3.000232, "mean_abs": 0.191892, "relative": 0.921323 }, { "name": "imu_accel_gyro", "l2": 17.050755, "mean_abs": 1.008011, "relative": 0.919947 }, { "name": "depth_confidence", "l2": 10.017854, "mean_abs": 0.18609, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.193913, "mean_abs": 0.10672, "relative": 0.940507 }, { "name": "video_fisheye_cam1", "l2": 5.327393, "mean_abs": 0.112538, "relative": 0.880307 }, { "name": "video_fisheye_cam2", "l2": 5.591296, "mean_abs": 0.112752, "relative": 0.892567 }, { "name": "video_fisheye_cam3", "l2": 5.508774, "mean_abs": 0.112764, "relative": 0.922378 }, { "name": "video_stereo_left", "l2": 5.248467, "mean_abs": 0.113904, "relative": 0.852886 }, { "name": "video_stereo_right", "l2": 6.341159, "mean_abs": 0.139478, "relative": 0.89397 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.997126, "mean_abs": 0.077097, "relative": 0.714741 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 347, "start_frame": 1735, "end_frame": 1754, "center_frame": 1744, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.18155, "mean_abs": 0.223747, "relative": 0.858857 }, { "name": "hand_right_joints", "l2": 9.613523, "mean_abs": 0.277859, "relative": 0.757803 }, { "name": "body_joints", "l2": 5.884612, "mean_abs": 0.093987, "relative": 0.79483 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019266, "mean_abs": 0.002966, "relative": 0.045464 }, { "name": "camera_rotation_matrix", "l2": 3.000566, "mean_abs": 0.193826, "relative": 0.921425 }, { "name": "imu_accel_gyro", "l2": 17.079784, "mean_abs": 1.027587, "relative": 0.921513 }, { "name": "depth_confidence", "l2": 9.997225, "mean_abs": 0.187083, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.193876, "mean_abs": 0.107093, "relative": 0.9405 }, { "name": "video_fisheye_cam1", "l2": 5.320565, "mean_abs": 0.113168, "relative": 0.879178 }, { "name": "video_fisheye_cam2", "l2": 5.594769, "mean_abs": 0.113523, "relative": 0.893122 }, { "name": "video_fisheye_cam3", "l2": 5.509918, "mean_abs": 0.113524, "relative": 0.922569 }, { "name": "video_stereo_left", "l2": 5.272024, "mean_abs": 0.115727, "relative": 0.856714 }, { "name": "video_stereo_right", "l2": 6.356089, "mean_abs": 0.141367, "relative": 0.896075 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.968805, "mean_abs": 0.075477, "relative": 0.704606 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 348, "start_frame": 1740, "end_frame": 1759, "center_frame": 1749, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.163262, "mean_abs": 0.221349, "relative": 0.857315 }, { "name": "hand_right_joints", "l2": 9.969091, "mean_abs": 0.283152, "relative": 0.785831 }, { "name": "body_joints", "l2": 5.933433, "mean_abs": 0.093155, "relative": 0.801424 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015876, "mean_abs": 0.002248, "relative": 0.037466 }, { "name": "camera_rotation_matrix", "l2": 3.000226, "mean_abs": 0.193166, "relative": 0.921321 }, { "name": "imu_accel_gyro", "l2": 17.08853, "mean_abs": 1.018125, "relative": 0.921985 }, { "name": "depth_confidence", "l2": 9.973804, "mean_abs": 0.186648, "relative": 0.000331 }, { "name": "video_fisheye_cam0", "l2": 5.19431, "mean_abs": 0.106979, "relative": 0.940578 }, { "name": "video_fisheye_cam1", "l2": 5.316827, "mean_abs": 0.112666, "relative": 0.878561 }, { "name": "video_fisheye_cam2", "l2": 5.601106, "mean_abs": 0.113486, "relative": 0.894133 }, { "name": "video_fisheye_cam3", "l2": 5.511847, "mean_abs": 0.113429, "relative": 0.922892 }, { "name": "video_stereo_left", "l2": 5.282464, "mean_abs": 0.115494, "relative": 0.858411 }, { "name": "video_stereo_right", "l2": 6.370848, "mean_abs": 0.140923, "relative": 0.898156 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.0149, "mean_abs": 0.07928, "relative": 0.721102 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00768, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 349, "start_frame": 1745, "end_frame": 1764, "center_frame": 1754, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.15496, "mean_abs": 0.220894, "relative": 0.856614 }, { "name": "hand_right_joints", "l2": 11.029758, "mean_abs": 0.267254, "relative": 0.86944 }, { "name": "body_joints", "l2": 5.996549, "mean_abs": 0.091636, "relative": 0.809949 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012204, "mean_abs": 0.001705, "relative": 0.028799 }, { "name": "camera_rotation_matrix", "l2": 2.999849, "mean_abs": 0.191786, "relative": 0.921205 }, { "name": "imu_accel_gyro", "l2": 17.094852, "mean_abs": 1.026464, "relative": 0.922326 }, { "name": "depth_confidence", "l2": 9.950951, "mean_abs": 0.185903, "relative": 0.00033 }, { "name": "video_fisheye_cam0", "l2": 5.197034, "mean_abs": 0.106814, "relative": 0.941072 }, { "name": "video_fisheye_cam1", "l2": 5.312744, "mean_abs": 0.111878, "relative": 0.877886 }, { "name": "video_fisheye_cam2", "l2": 5.605785, "mean_abs": 0.113089, "relative": 0.89488 }, { "name": "video_fisheye_cam3", "l2": 5.514862, "mean_abs": 0.112766, "relative": 0.923397 }, { "name": "video_stereo_left", "l2": 5.295238, "mean_abs": 0.114787, "relative": 0.860487 }, { "name": "video_stereo_right", "l2": 6.381742, "mean_abs": 0.139825, "relative": 0.899692 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.029638, "mean_abs": 0.080483, "relative": 0.726377 }, { "name": "caption_objects_interaction_text", "l2": 1.845294, "mean_abs": 0.010771, "relative": 0.847314 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 350, "start_frame": 1750, "end_frame": 1769, "center_frame": 1759, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.151055, "mean_abs": 0.221196, "relative": 0.856285 }, { "name": "hand_right_joints", "l2": 11.041099, "mean_abs": 0.266885, "relative": 0.870334 }, { "name": "body_joints", "l2": 5.999486, "mean_abs": 0.091695, "relative": 0.810346 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008641, "mean_abs": 0.001319, "relative": 0.020391 }, { "name": "camera_rotation_matrix", "l2": 2.999949, "mean_abs": 0.191288, "relative": 0.921236 }, { "name": "imu_accel_gyro", "l2": 17.06624, "mean_abs": 1.020939, "relative": 0.920782 }, { "name": "depth_confidence", "l2": 9.915881, "mean_abs": 0.183802, "relative": 0.000329 }, { "name": "video_fisheye_cam0", "l2": 5.198224, "mean_abs": 0.10681, "relative": 0.941287 }, { "name": "video_fisheye_cam1", "l2": 5.310197, "mean_abs": 0.111631, "relative": 0.877465 }, { "name": "video_fisheye_cam2", "l2": 5.60673, "mean_abs": 0.11292, "relative": 0.895031 }, { "name": "video_fisheye_cam3", "l2": 5.522496, "mean_abs": 0.112774, "relative": 0.924675 }, { "name": "video_stereo_left", "l2": 5.299057, "mean_abs": 0.114001, "relative": 0.861107 }, { "name": "video_stereo_right", "l2": 6.385046, "mean_abs": 0.139064, "relative": 0.900157 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.018815, "mean_abs": 0.077457, "relative": 0.722504 }, { "name": "caption_objects_interaction_text", "l2": 1.845294, "mean_abs": 0.010899, "relative": 0.847314 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 351, "start_frame": 1755, "end_frame": 1774, "center_frame": 1764, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.146642, "mean_abs": 0.221336, "relative": 0.855913 }, { "name": "hand_right_joints", "l2": 11.055829, "mean_abs": 0.270245, "relative": 0.871495 }, { "name": "body_joints", "l2": 6.03319, "mean_abs": 0.093401, "relative": 0.814898 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014763, "mean_abs": 0.002336, "relative": 0.034838 }, { "name": "camera_rotation_matrix", "l2": 3.000284, "mean_abs": 0.191089, "relative": 0.921338 }, { "name": "imu_accel_gyro", "l2": 17.096727, "mean_abs": 1.022892, "relative": 0.922427 }, { "name": "depth_confidence", "l2": 9.921248, "mean_abs": 0.183849, "relative": 0.000329 }, { "name": "video_fisheye_cam0", "l2": 5.200449, "mean_abs": 0.106763, "relative": 0.94169 }, { "name": "video_fisheye_cam1", "l2": 5.322947, "mean_abs": 0.111848, "relative": 0.879572 }, { "name": "video_fisheye_cam2", "l2": 5.607262, "mean_abs": 0.112936, "relative": 0.895116 }, { "name": "video_fisheye_cam3", "l2": 5.525161, "mean_abs": 0.112975, "relative": 0.925122 }, { "name": "video_stereo_left", "l2": 5.296412, "mean_abs": 0.114104, "relative": 0.860678 }, { "name": "video_stereo_right", "l2": 6.389208, "mean_abs": 0.139159, "relative": 0.900744 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.066028, "mean_abs": 0.078645, "relative": 0.739401 }, { "name": "caption_objects_interaction_text", "l2": 1.845294, "mean_abs": 0.010788, "relative": 0.847314 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 352, "start_frame": 1760, "end_frame": 1779, "center_frame": 1769, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.14449, "mean_abs": 0.221412, "relative": 0.855731 }, { "name": "hand_right_joints", "l2": 11.034338, "mean_abs": 0.271544, "relative": 0.869801 }, { "name": "body_joints", "l2": 6.049215, "mean_abs": 0.094427, "relative": 0.817063 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028761, "mean_abs": 0.004258, "relative": 0.067872 }, { "name": "camera_rotation_matrix", "l2": 3.0, "mean_abs": 0.192736, "relative": 0.921251 }, { "name": "imu_accel_gyro", "l2": 17.069641, "mean_abs": 1.031158, "relative": 0.920966 }, { "name": "depth_confidence", "l2": 9.970843, "mean_abs": 0.18584, "relative": 0.000331 }, { "name": "video_fisheye_cam0", "l2": 5.203319, "mean_abs": 0.107141, "relative": 0.94221 }, { "name": "video_fisheye_cam1", "l2": 5.334149, "mean_abs": 0.112386, "relative": 0.881423 }, { "name": "video_fisheye_cam2", "l2": 5.621444, "mean_abs": 0.113578, "relative": 0.89738 }, { "name": "video_fisheye_cam3", "l2": 5.521723, "mean_abs": 0.113018, "relative": 0.924546 }, { "name": "video_stereo_left", "l2": 5.312657, "mean_abs": 0.115307, "relative": 0.863317 }, { "name": "video_stereo_right", "l2": 6.407248, "mean_abs": 0.140602, "relative": 0.903288 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.125988, "mean_abs": 0.08203, "relative": 0.760859 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007781, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 353, "start_frame": 1765, "end_frame": 1784, "center_frame": 1774, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.139791, "mean_abs": 0.222035, "relative": 0.855335 }, { "name": "hand_right_joints", "l2": 10.946304, "mean_abs": 0.270779, "relative": 0.862862 }, { "name": "body_joints", "l2": 6.069739, "mean_abs": 0.095278, "relative": 0.819835 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.034011, "mean_abs": 0.005311, "relative": 0.080262 }, { "name": "camera_rotation_matrix", "l2": 2.999912, "mean_abs": 0.192994, "relative": 0.921224 }, { "name": "imu_accel_gyro", "l2": 17.084106, "mean_abs": 1.035893, "relative": 0.921746 }, { "name": "depth_confidence", "l2": 10.005773, "mean_abs": 0.18687, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.207629, "mean_abs": 0.107315, "relative": 0.94299 }, { "name": "video_fisheye_cam1", "l2": 5.345224, "mean_abs": 0.112941, "relative": 0.883253 }, { "name": "video_fisheye_cam2", "l2": 5.623281, "mean_abs": 0.113632, "relative": 0.897673 }, { "name": "video_fisheye_cam3", "l2": 5.52577, "mean_abs": 0.113373, "relative": 0.925224 }, { "name": "video_stereo_left", "l2": 5.31403, "mean_abs": 0.115536, "relative": 0.86354 }, { "name": "video_stereo_right", "l2": 6.411243, "mean_abs": 0.141362, "relative": 0.903851 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.115006, "mean_abs": 0.078917, "relative": 0.756929 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007781, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 354, "start_frame": 1770, "end_frame": 1789, "center_frame": 1779, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.119873, "mean_abs": 0.223509, "relative": 0.853654 }, { "name": "hand_right_joints", "l2": 10.72234, "mean_abs": 0.271026, "relative": 0.845207 }, { "name": "body_joints", "l2": 6.125235, "mean_abs": 0.098558, "relative": 0.827331 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.048965, "mean_abs": 0.00716, "relative": 0.115551 }, { "name": "camera_rotation_matrix", "l2": 3.001135, "mean_abs": 0.195894, "relative": 0.9216 }, { "name": "imu_accel_gyro", "l2": 17.097998, "mean_abs": 1.052103, "relative": 0.922496 }, { "name": "depth_confidence", "l2": 10.044782, "mean_abs": 0.188779, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.214081, "mean_abs": 0.107895, "relative": 0.944158 }, { "name": "video_fisheye_cam1", "l2": 5.373662, "mean_abs": 0.114947, "relative": 0.887952 }, { "name": "video_fisheye_cam2", "l2": 5.609507, "mean_abs": 0.114727, "relative": 0.895474 }, { "name": "video_fisheye_cam3", "l2": 5.532161, "mean_abs": 0.114345, "relative": 0.926294 }, { "name": "video_stereo_left", "l2": 5.288207, "mean_abs": 0.117187, "relative": 0.859344 }, { "name": "video_stereo_right", "l2": 6.391285, "mean_abs": 0.14338, "relative": 0.901037 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.104578, "mean_abs": 0.080829, "relative": 0.753197 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007781, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 355, "start_frame": 1775, "end_frame": 1794, "center_frame": 1784, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.083068, "mean_abs": 0.22725, "relative": 0.85055 }, { "name": "hand_right_joints", "l2": 10.598371, "mean_abs": 0.266552, "relative": 0.835435 }, { "name": "body_joints", "l2": 6.173504, "mean_abs": 0.101195, "relative": 0.83385 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.086445, "mean_abs": 0.009756, "relative": 0.203999 }, { "name": "camera_rotation_matrix", "l2": 3.008329, "mean_abs": 0.19983, "relative": 0.923809 }, { "name": "imu_accel_gyro", "l2": 17.074781, "mean_abs": 1.063595, "relative": 0.921243 }, { "name": "depth_confidence", "l2": 10.025955, "mean_abs": 0.191096, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.212847, "mean_abs": 0.109312, "relative": 0.943935 }, { "name": "video_fisheye_cam1", "l2": 5.402626, "mean_abs": 0.117784, "relative": 0.892738 }, { "name": "video_fisheye_cam2", "l2": 5.566432, "mean_abs": 0.116474, "relative": 0.888598 }, { "name": "video_fisheye_cam3", "l2": 5.542959, "mean_abs": 0.115919, "relative": 0.928102 }, { "name": "video_stereo_left", "l2": 5.263132, "mean_abs": 0.120518, "relative": 0.855269 }, { "name": "video_stereo_right", "l2": 6.35685, "mean_abs": 0.147343, "relative": 0.896182 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.076131, "mean_abs": 0.078103, "relative": 0.743016 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007781, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 356, "start_frame": 1780, "end_frame": 1799, "center_frame": 1789, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.060606, "mean_abs": 0.226898, "relative": 0.848655 }, { "name": "hand_right_joints", "l2": 10.5097, "mean_abs": 0.260475, "relative": 0.828445 }, { "name": "body_joints", "l2": 6.18883, "mean_abs": 0.101451, "relative": 0.83592 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.141237, "mean_abs": 0.01465, "relative": 0.3333 }, { "name": "camera_rotation_matrix", "l2": 3.020207, "mean_abs": 0.201924, "relative": 0.927457 }, { "name": "imu_accel_gyro", "l2": 17.04917, "mean_abs": 1.059392, "relative": 0.919861 }, { "name": "depth_confidence", "l2": 9.993045, "mean_abs": 0.192228, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.203722, "mean_abs": 0.110386, "relative": 0.942283 }, { "name": "video_fisheye_cam1", "l2": 5.431684, "mean_abs": 0.120349, "relative": 0.89754 }, { "name": "video_fisheye_cam2", "l2": 5.515615, "mean_abs": 0.117111, "relative": 0.880486 }, { "name": "video_fisheye_cam3", "l2": 5.520213, "mean_abs": 0.116759, "relative": 0.924293 }, { "name": "video_stereo_left", "l2": 5.195697, "mean_abs": 0.12089, "relative": 0.844311 }, { "name": "video_stereo_right", "l2": 6.344961, "mean_abs": 0.149317, "relative": 0.894506 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.998526, "mean_abs": 0.075616, "relative": 0.715243 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007781, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 357, "start_frame": 1785, "end_frame": 1804, "center_frame": 1794, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper", "coffee container", "lid", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.067886, "mean_abs": 0.225388, "relative": 0.849269 }, { "name": "hand_right_joints", "l2": 10.706036, "mean_abs": 0.272542, "relative": 0.843922 }, { "name": "body_joints", "l2": 6.195277, "mean_abs": 0.102436, "relative": 0.836791 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.175368, "mean_abs": 0.018665, "relative": 0.413843 }, { "name": "camera_rotation_matrix", "l2": 3.018781, "mean_abs": 0.198067, "relative": 0.927019 }, { "name": "imu_accel_gyro", "l2": 17.022514, "mean_abs": 1.053491, "relative": 0.918423 }, { "name": "depth_confidence", "l2": 9.931289, "mean_abs": 0.1904, "relative": 0.00033 }, { "name": "video_fisheye_cam0", "l2": 5.188213, "mean_abs": 0.109944, "relative": 0.939474 }, { "name": "video_fisheye_cam1", "l2": 5.423136, "mean_abs": 0.120749, "relative": 0.896127 }, { "name": "video_fisheye_cam2", "l2": 5.464213, "mean_abs": 0.116035, "relative": 0.87228 }, { "name": "video_fisheye_cam3", "l2": 5.481695, "mean_abs": 0.115852, "relative": 0.917844 }, { "name": "video_stereo_left", "l2": 5.118337, "mean_abs": 0.118852, "relative": 0.83174 }, { "name": "video_stereo_right", "l2": 6.263648, "mean_abs": 0.148512, "relative": 0.883043 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.067364, "mean_abs": 0.079518, "relative": 0.739879 }, { "name": "caption_objects_interaction_text", "l2": 2.0214, "mean_abs": 0.016335, "relative": 0.928178 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 358, "start_frame": 1790, "end_frame": 1809, "center_frame": 1799, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper", "coffee container", "lid", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.080385, "mean_abs": 0.226196, "relative": 0.850323 }, { "name": "hand_right_joints", "l2": 11.132354, "mean_abs": 0.292372, "relative": 0.877527 }, { "name": "body_joints", "l2": 6.268281, "mean_abs": 0.103832, "relative": 0.846652 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.174767, "mean_abs": 0.020425, "relative": 0.412425 }, { "name": "camera_rotation_matrix", "l2": 3.00777, "mean_abs": 0.192564, "relative": 0.923637 }, { "name": "imu_accel_gyro", "l2": 17.065262, "mean_abs": 1.087085, "relative": 0.92073 }, { "name": "depth_confidence", "l2": 9.834743, "mean_abs": 0.187385, "relative": 0.000326 }, { "name": "video_fisheye_cam0", "l2": 5.163604, "mean_abs": 0.108315, "relative": 0.935018 }, { "name": "video_fisheye_cam1", "l2": 5.387622, "mean_abs": 0.118979, "relative": 0.890259 }, { "name": "video_fisheye_cam2", "l2": 5.404994, "mean_abs": 0.113764, "relative": 0.862827 }, { "name": "video_fisheye_cam3", "l2": 5.403146, "mean_abs": 0.113263, "relative": 0.904692 }, { "name": "video_stereo_left", "l2": 5.043606, "mean_abs": 0.115916, "relative": 0.819596 }, { "name": "video_stereo_right", "l2": 6.11974, "mean_abs": 0.143624, "relative": 0.862755 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.957008, "mean_abs": 0.076693, "relative": 0.700384 }, { "name": "caption_objects_interaction_text", "l2": 2.0214, "mean_abs": 0.016878, "relative": 0.928178 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 359, "start_frame": 1795, "end_frame": 1814, "center_frame": 1804, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee scoop", "dripper", "coffee container", "lid", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.129659, "mean_abs": 0.229957, "relative": 0.85448 }, { "name": "hand_right_joints", "l2": 11.583902, "mean_abs": 0.306448, "relative": 0.913121 }, { "name": "body_joints", "l2": 6.257068, "mean_abs": 0.100747, "relative": 0.845137 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.14491, "mean_abs": 0.020009, "relative": 0.341968 }, { "name": "camera_rotation_matrix", "l2": 3.002547, "mean_abs": 0.186379, "relative": 0.922033 }, { "name": "imu_accel_gyro", "l2": 17.20339, "mean_abs": 1.14003, "relative": 0.928182 }, { "name": "depth_confidence", "l2": 9.708462, "mean_abs": 0.183992, "relative": 0.000322 }, { "name": "video_fisheye_cam0", "l2": 5.129808, "mean_abs": 0.107221, "relative": 0.928899 }, { "name": "video_fisheye_cam1", "l2": 5.31675, "mean_abs": 0.114906, "relative": 0.878548 }, { "name": "video_fisheye_cam2", "l2": 5.351352, "mean_abs": 0.110838, "relative": 0.854264 }, { "name": "video_fisheye_cam3", "l2": 5.337992, "mean_abs": 0.110912, "relative": 0.893782 }, { "name": "video_stereo_left", "l2": 4.964126, "mean_abs": 0.112306, "relative": 0.80668 }, { "name": "video_stereo_right", "l2": 5.980199, "mean_abs": 0.136881, "relative": 0.843083 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.039446, "mean_abs": 0.081405, "relative": 0.729887 }, { "name": "caption_objects_interaction_text", "l2": 2.0214, "mean_abs": 0.016851, "relative": 0.928178 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 360, "start_frame": 1800, "end_frame": 1819, "center_frame": 1809, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.115113, "mean_abs": 0.235892, "relative": 0.853253 }, { "name": "hand_right_joints", "l2": 11.942782, "mean_abs": 0.312599, "relative": 0.941411 }, { "name": "body_joints", "l2": 6.252306, "mean_abs": 0.100473, "relative": 0.844494 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.102331, "mean_abs": 0.015984, "relative": 0.241487 }, { "name": "camera_rotation_matrix", "l2": 3.005335, "mean_abs": 0.18713, "relative": 0.92289 }, { "name": "imu_accel_gyro", "l2": 17.130066, "mean_abs": 1.131284, "relative": 0.924226 }, { "name": "depth_confidence", "l2": 9.605367, "mean_abs": 0.182534, "relative": 0.000319 }, { "name": "video_fisheye_cam0", "l2": 5.123602, "mean_abs": 0.107541, "relative": 0.927775 }, { "name": "video_fisheye_cam1", "l2": 5.252257, "mean_abs": 0.113561, "relative": 0.867891 }, { "name": "video_fisheye_cam2", "l2": 5.368126, "mean_abs": 0.110977, "relative": 0.856942 }, { "name": "video_fisheye_cam3", "l2": 5.339853, "mean_abs": 0.111262, "relative": 0.894094 }, { "name": "video_stereo_left", "l2": 4.990511, "mean_abs": 0.112903, "relative": 0.810968 }, { "name": "video_stereo_right", "l2": 5.961661, "mean_abs": 0.137227, "relative": 0.840469 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.062804, "mean_abs": 0.082818, "relative": 0.738247 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 361, "start_frame": 1805, "end_frame": 1824, "center_frame": 1814, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.089702, "mean_abs": 0.238528, "relative": 0.851109 }, { "name": "hand_right_joints", "l2": 12.285629, "mean_abs": 0.316381, "relative": 0.968436 }, { "name": "body_joints", "l2": 6.300606, "mean_abs": 0.100764, "relative": 0.851018 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.07004, "mean_abs": 0.011008, "relative": 0.165284 }, { "name": "camera_rotation_matrix", "l2": 3.005756, "mean_abs": 0.187154, "relative": 0.923019 }, { "name": "imu_accel_gyro", "l2": 17.131145, "mean_abs": 1.108555, "relative": 0.924284 }, { "name": "depth_confidence", "l2": 9.524655, "mean_abs": 0.181175, "relative": 0.000316 }, { "name": "video_fisheye_cam0", "l2": 5.134096, "mean_abs": 0.107998, "relative": 0.929675 }, { "name": "video_fisheye_cam1", "l2": 5.208222, "mean_abs": 0.113001, "relative": 0.860615 }, { "name": "video_fisheye_cam2", "l2": 5.404216, "mean_abs": 0.111896, "relative": 0.862703 }, { "name": "video_fisheye_cam3", "l2": 5.356944, "mean_abs": 0.112071, "relative": 0.896956 }, { "name": "video_stereo_left", "l2": 5.030241, "mean_abs": 0.113314, "relative": 0.817424 }, { "name": "video_stereo_right", "l2": 5.944026, "mean_abs": 0.136427, "relative": 0.837983 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.073962, "mean_abs": 0.086217, "relative": 0.74224 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 362, "start_frame": 1810, "end_frame": 1829, "center_frame": 1819, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.019928, "mean_abs": 0.235491, "relative": 0.845224 }, { "name": "hand_right_joints", "l2": 12.51477, "mean_abs": 0.314953, "relative": 0.986499 }, { "name": "body_joints", "l2": 6.336098, "mean_abs": 0.100152, "relative": 0.855812 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.04968, "mean_abs": 0.006635, "relative": 0.117237 }, { "name": "camera_rotation_matrix", "l2": 3.003074, "mean_abs": 0.184774, "relative": 0.922195 }, { "name": "imu_accel_gyro", "l2": 17.191071, "mean_abs": 1.116156, "relative": 0.927517 }, { "name": "depth_confidence", "l2": 9.37468, "mean_abs": 0.178404, "relative": 0.000311 }, { "name": "video_fisheye_cam0", "l2": 5.128699, "mean_abs": 0.107064, "relative": 0.928698 }, { "name": "video_fisheye_cam1", "l2": 5.16563, "mean_abs": 0.110786, "relative": 0.853577 }, { "name": "video_fisheye_cam2", "l2": 5.426229, "mean_abs": 0.111545, "relative": 0.866217 }, { "name": "video_fisheye_cam3", "l2": 5.397207, "mean_abs": 0.113596, "relative": 0.903697 }, { "name": "video_stereo_left", "l2": 5.05388, "mean_abs": 0.113683, "relative": 0.821266 }, { "name": "video_stereo_right", "l2": 5.855723, "mean_abs": 0.133563, "relative": 0.825534 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.081732, "mean_abs": 0.086748, "relative": 0.745021 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 363, "start_frame": 1815, "end_frame": 1834, "center_frame": 1824, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.952847, "mean_abs": 0.233858, "relative": 0.839565 }, { "name": "hand_right_joints", "l2": 12.536514, "mean_abs": 0.320993, "relative": 0.988213 }, { "name": "body_joints", "l2": 6.443069, "mean_abs": 0.101847, "relative": 0.87026 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.049118, "mean_abs": 0.007077, "relative": 0.11591 }, { "name": "camera_rotation_matrix", "l2": 3.003609, "mean_abs": 0.183747, "relative": 0.92236 }, { "name": "imu_accel_gyro", "l2": 17.124617, "mean_abs": 1.104652, "relative": 0.923932 }, { "name": "depth_confidence", "l2": 9.29724, "mean_abs": 0.177853, "relative": 0.000309 }, { "name": "video_fisheye_cam0", "l2": 5.108531, "mean_abs": 0.107117, "relative": 0.925046 }, { "name": "video_fisheye_cam1", "l2": 5.161461, "mean_abs": 0.110644, "relative": 0.852888 }, { "name": "video_fisheye_cam2", "l2": 5.42601, "mean_abs": 0.111188, "relative": 0.866182 }, { "name": "video_fisheye_cam3", "l2": 5.480828, "mean_abs": 0.114904, "relative": 0.917699 }, { "name": "video_stereo_left", "l2": 5.049735, "mean_abs": 0.114129, "relative": 0.820592 }, { "name": "video_stereo_right", "l2": 5.805633, "mean_abs": 0.133829, "relative": 0.818472 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.005573, "mean_abs": 0.077472, "relative": 0.717764 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 364, "start_frame": 1820, "end_frame": 1839, "center_frame": 1829, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.859664, "mean_abs": 0.231444, "relative": 0.831705 }, { "name": "hand_right_joints", "l2": 12.519385, "mean_abs": 0.326479, "relative": 0.986862 }, { "name": "body_joints", "l2": 6.536668, "mean_abs": 0.103213, "relative": 0.882902 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.064164, "mean_abs": 0.009502, "relative": 0.151417 }, { "name": "camera_rotation_matrix", "l2": 3.003407, "mean_abs": 0.184367, "relative": 0.922297 }, { "name": "imu_accel_gyro", "l2": 17.05447, "mean_abs": 1.092303, "relative": 0.920147 }, { "name": "depth_confidence", "l2": 9.262338, "mean_abs": 0.176684, "relative": 0.000307 }, { "name": "video_fisheye_cam0", "l2": 5.109172, "mean_abs": 0.106953, "relative": 0.925162 }, { "name": "video_fisheye_cam1", "l2": 5.185987, "mean_abs": 0.111576, "relative": 0.856941 }, { "name": "video_fisheye_cam2", "l2": 5.451785, "mean_abs": 0.112028, "relative": 0.870296 }, { "name": "video_fisheye_cam3", "l2": 5.541183, "mean_abs": 0.114984, "relative": 0.927804 }, { "name": "video_stereo_left", "l2": 5.044275, "mean_abs": 0.113535, "relative": 0.819705 }, { "name": "video_stereo_right", "l2": 5.793118, "mean_abs": 0.133398, "relative": 0.816708 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.044275, "mean_abs": 0.080205, "relative": 0.731615 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 365, "start_frame": 1825, "end_frame": 1844, "center_frame": 1834, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.821734, "mean_abs": 0.230026, "relative": 0.828505 }, { "name": "hand_right_joints", "l2": 12.493633, "mean_abs": 0.325663, "relative": 0.984833 }, { "name": "body_joints", "l2": 6.584039, "mean_abs": 0.103816, "relative": 0.889301 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.080291, "mean_abs": 0.011102, "relative": 0.189475 }, { "name": "camera_rotation_matrix", "l2": 3.006114, "mean_abs": 0.187873, "relative": 0.923129 }, { "name": "imu_accel_gyro", "l2": 17.195181, "mean_abs": 1.106204, "relative": 0.927739 }, { "name": "depth_confidence", "l2": 9.191238, "mean_abs": 0.174946, "relative": 0.000305 }, { "name": "video_fisheye_cam0", "l2": 5.103882, "mean_abs": 0.107095, "relative": 0.924204 }, { "name": "video_fisheye_cam1", "l2": 5.237633, "mean_abs": 0.114312, "relative": 0.865475 }, { "name": "video_fisheye_cam2", "l2": 5.526253, "mean_abs": 0.115324, "relative": 0.882184 }, { "name": "video_fisheye_cam3", "l2": 5.589598, "mean_abs": 0.116647, "relative": 0.935911 }, { "name": "video_stereo_left", "l2": 5.139014, "mean_abs": 0.116799, "relative": 0.8351 }, { "name": "video_stereo_right", "l2": 5.893755, "mean_abs": 0.137588, "relative": 0.830896 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.064568, "mean_abs": 0.080786, "relative": 0.738878 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 366, "start_frame": 1830, "end_frame": 1849, "center_frame": 1839, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.784762, "mean_abs": 0.23041, "relative": 0.825386 }, { "name": "hand_right_joints", "l2": 12.129473, "mean_abs": 0.325473, "relative": 0.956127 }, { "name": "body_joints", "l2": 6.665562, "mean_abs": 0.104493, "relative": 0.900312 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.104706, "mean_abs": 0.014527, "relative": 0.247092 }, { "name": "camera_rotation_matrix", "l2": 3.012095, "mean_abs": 0.193134, "relative": 0.924966 }, { "name": "imu_accel_gyro", "l2": 17.303274, "mean_abs": 1.122601, "relative": 0.933571 }, { "name": "depth_confidence", "l2": 9.287547, "mean_abs": 0.179517, "relative": 0.000308 }, { "name": "video_fisheye_cam0", "l2": 5.097416, "mean_abs": 0.10821, "relative": 0.923033 }, { "name": "video_fisheye_cam1", "l2": 5.26807, "mean_abs": 0.116232, "relative": 0.870504 }, { "name": "video_fisheye_cam2", "l2": 5.580437, "mean_abs": 0.117797, "relative": 0.890834 }, { "name": "video_fisheye_cam3", "l2": 5.581743, "mean_abs": 0.116948, "relative": 0.934596 }, { "name": "video_stereo_left", "l2": 5.255231, "mean_abs": 0.120428, "relative": 0.853986 }, { "name": "video_stereo_right", "l2": 6.027314, "mean_abs": 0.143103, "relative": 0.849725 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.048365, "mean_abs": 0.07886, "relative": 0.733079 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 367, "start_frame": 1835, "end_frame": 1854, "center_frame": 1844, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.77393, "mean_abs": 0.230949, "relative": 0.824473 }, { "name": "hand_right_joints", "l2": 11.580742, "mean_abs": 0.324198, "relative": 0.912872 }, { "name": "body_joints", "l2": 6.80501, "mean_abs": 0.106446, "relative": 0.919147 }, { "name": "body_contacts", "l2": 3.464478, "mean_abs": 0.081737, "relative": 0.999733 }, { "name": "camera_translation", "l2": 0.163298, "mean_abs": 0.024636, "relative": 0.385359 }, { "name": "camera_rotation_matrix", "l2": 3.012657, "mean_abs": 0.198417, "relative": 0.925138 }, { "name": "imu_accel_gyro", "l2": 17.297834, "mean_abs": 1.137259, "relative": 0.933278 }, { "name": "depth_confidence", "l2": 9.464845, "mean_abs": 0.1844, "relative": 0.000314 }, { "name": "video_fisheye_cam0", "l2": 5.103305, "mean_abs": 0.108281, "relative": 0.924099 }, { "name": "video_fisheye_cam1", "l2": 5.357857, "mean_abs": 0.118979, "relative": 0.885341 }, { "name": "video_fisheye_cam2", "l2": 5.645588, "mean_abs": 0.119721, "relative": 0.901234 }, { "name": "video_fisheye_cam3", "l2": 5.609457, "mean_abs": 0.11744, "relative": 0.939236 }, { "name": "video_stereo_left", "l2": 5.377988, "mean_abs": 0.124836, "relative": 0.873934 }, { "name": "video_stereo_right", "l2": 6.281435, "mean_abs": 0.150635, "relative": 0.88555 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.017703, "mean_abs": 0.07704, "relative": 0.722105 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 368, "start_frame": 1840, "end_frame": 1859, "center_frame": 1849, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.807355, "mean_abs": 0.232348, "relative": 0.827292 }, { "name": "hand_right_joints", "l2": 11.242356, "mean_abs": 0.323449, "relative": 0.886198 }, { "name": "body_joints", "l2": 6.81564, "mean_abs": 0.111506, "relative": 0.920583 }, { "name": "body_contacts", "l2": 3.464455, "mean_abs": 0.081716, "relative": 0.999727 }, { "name": "camera_translation", "l2": 0.238879, "mean_abs": 0.034955, "relative": 0.563722 }, { "name": "camera_rotation_matrix", "l2": 3.015093, "mean_abs": 0.20487, "relative": 0.925886 }, { "name": "imu_accel_gyro", "l2": 17.262852, "mean_abs": 1.170772, "relative": 0.93139 }, { "name": "depth_confidence", "l2": 9.738885, "mean_abs": 0.189844, "relative": 0.000323 }, { "name": "video_fisheye_cam0", "l2": 5.104307, "mean_abs": 0.107936, "relative": 0.924281 }, { "name": "video_fisheye_cam1", "l2": 5.439823, "mean_abs": 0.120101, "relative": 0.898885 }, { "name": "video_fisheye_cam2", "l2": 5.743453, "mean_abs": 0.121395, "relative": 0.916857 }, { "name": "video_fisheye_cam3", "l2": 5.624017, "mean_abs": 0.118968, "relative": 0.941674 }, { "name": "video_stereo_left", "l2": 5.488386, "mean_abs": 0.126869, "relative": 0.891874 }, { "name": "video_stereo_right", "l2": 6.472217, "mean_abs": 0.15294, "relative": 0.912447 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.015047, "mean_abs": 0.080448, "relative": 0.721155 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 369, "start_frame": 1845, "end_frame": 1864, "center_frame": 1854, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.826078, "mean_abs": 0.233887, "relative": 0.828872 }, { "name": "hand_right_joints", "l2": 10.875857, "mean_abs": 0.314129, "relative": 0.857309 }, { "name": "body_joints", "l2": 6.74549, "mean_abs": 0.113066, "relative": 0.911108 }, { "name": "body_contacts", "l2": 3.464455, "mean_abs": 0.081715, "relative": 0.999727 }, { "name": "camera_translation", "l2": 0.2758, "mean_abs": 0.040054, "relative": 0.65085 }, { "name": "camera_rotation_matrix", "l2": 3.01779, "mean_abs": 0.211717, "relative": 0.926714 }, { "name": "imu_accel_gyro", "l2": 17.452444, "mean_abs": 1.214655, "relative": 0.941619 }, { "name": "depth_confidence", "l2": 10.059453, "mean_abs": 0.195035, "relative": 0.000334 }, { "name": "video_fisheye_cam0", "l2": 5.090113, "mean_abs": 0.108013, "relative": 0.921711 }, { "name": "video_fisheye_cam1", "l2": 5.52508, "mean_abs": 0.122946, "relative": 0.912973 }, { "name": "video_fisheye_cam2", "l2": 5.801311, "mean_abs": 0.122644, "relative": 0.926093 }, { "name": "video_fisheye_cam3", "l2": 5.62668, "mean_abs": 0.119627, "relative": 0.94212 }, { "name": "video_stereo_left", "l2": 5.612292, "mean_abs": 0.12984, "relative": 0.912009 }, { "name": "video_stereo_right", "l2": 6.597552, "mean_abs": 0.156276, "relative": 0.930116 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.134766, "mean_abs": 0.084135, "relative": 0.764001 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 370, "start_frame": 1850, "end_frame": 1869, "center_frame": 1859, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.856841, "mean_abs": 0.234794, "relative": 0.831467 }, { "name": "hand_right_joints", "l2": 10.098669, "mean_abs": 0.284553, "relative": 0.796045 }, { "name": "body_joints", "l2": 6.476601, "mean_abs": 0.106261, "relative": 0.874789 }, { "name": "body_contacts", "l2": 3.464455, "mean_abs": 0.081712, "relative": 0.999727 }, { "name": "camera_translation", "l2": 0.253524, "mean_abs": 0.037115, "relative": 0.598282 }, { "name": "camera_rotation_matrix", "l2": 3.01667, "mean_abs": 0.214333, "relative": 0.92637 }, { "name": "imu_accel_gyro", "l2": 17.306866, "mean_abs": 1.190927, "relative": 0.933765 }, { "name": "depth_confidence", "l2": 10.378851, "mean_abs": 0.198962, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.084055, "mean_abs": 0.108106, "relative": 0.920614 }, { "name": "video_fisheye_cam1", "l2": 5.588005, "mean_abs": 0.123551, "relative": 0.923371 }, { "name": "video_fisheye_cam2", "l2": 5.818797, "mean_abs": 0.122824, "relative": 0.928884 }, { "name": "video_fisheye_cam3", "l2": 5.653819, "mean_abs": 0.120162, "relative": 0.946664 }, { "name": "video_stereo_left", "l2": 5.632265, "mean_abs": 0.130637, "relative": 0.915254 }, { "name": "video_stereo_right", "l2": 6.649487, "mean_abs": 0.157253, "relative": 0.937438 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.196934, "mean_abs": 0.087793, "relative": 0.78625 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 371, "start_frame": 1855, "end_frame": 1874, "center_frame": 1864, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.856291, "mean_abs": 0.233222, "relative": 0.83142 }, { "name": "hand_right_joints", "l2": 9.486117, "mean_abs": 0.252051, "relative": 0.74776 }, { "name": "body_joints", "l2": 6.21739, "mean_abs": 0.098294, "relative": 0.839778 }, { "name": "body_contacts", "l2": 3.464389, "mean_abs": 0.081732, "relative": 0.999708 }, { "name": "camera_translation", "l2": 0.18667, "mean_abs": 0.027625, "relative": 0.440516 }, { "name": "camera_rotation_matrix", "l2": 3.010698, "mean_abs": 0.213846, "relative": 0.924537 }, { "name": "imu_accel_gyro", "l2": 16.996742, "mean_abs": 1.115518, "relative": 0.917033 }, { "name": "depth_confidence", "l2": 10.626122, "mean_abs": 0.200312, "relative": 0.000353 }, { "name": "video_fisheye_cam0", "l2": 5.040927, "mean_abs": 0.106017, "relative": 0.912804 }, { "name": "video_fisheye_cam1", "l2": 5.609415, "mean_abs": 0.122132, "relative": 0.926908 }, { "name": "video_fisheye_cam2", "l2": 5.802143, "mean_abs": 0.121209, "relative": 0.926226 }, { "name": "video_fisheye_cam3", "l2": 5.568371, "mean_abs": 0.116137, "relative": 0.932357 }, { "name": "video_stereo_left", "l2": 5.579394, "mean_abs": 0.128167, "relative": 0.906663 }, { "name": "video_stereo_right", "l2": 6.605237, "mean_abs": 0.154481, "relative": 0.9312 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.184077, "mean_abs": 0.087091, "relative": 0.781648 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 372, "start_frame": 1860, "end_frame": 1879, "center_frame": 1869, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.858518, "mean_abs": 0.231908, "relative": 0.831608 }, { "name": "hand_right_joints", "l2": 9.453374, "mean_abs": 0.241899, "relative": 0.745179 }, { "name": "body_joints", "l2": 6.08634, "mean_abs": 0.095381, "relative": 0.822077 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.099569, "mean_abs": 0.014875, "relative": 0.234969 }, { "name": "camera_rotation_matrix", "l2": 3.00323, "mean_abs": 0.212311, "relative": 0.922243 }, { "name": "imu_accel_gyro", "l2": 16.923166, "mean_abs": 1.070926, "relative": 0.913063 }, { "name": "depth_confidence", "l2": 10.813644, "mean_abs": 0.200498, "relative": 0.000359 }, { "name": "video_fisheye_cam0", "l2": 5.030437, "mean_abs": 0.105192, "relative": 0.910905 }, { "name": "video_fisheye_cam1", "l2": 5.62152, "mean_abs": 0.120296, "relative": 0.928909 }, { "name": "video_fisheye_cam2", "l2": 5.793601, "mean_abs": 0.120094, "relative": 0.924862 }, { "name": "video_fisheye_cam3", "l2": 5.533865, "mean_abs": 0.114729, "relative": 0.926579 }, { "name": "video_stereo_left", "l2": 5.555049, "mean_abs": 0.125331, "relative": 0.902707 }, { "name": "video_stereo_right", "l2": 6.543861, "mean_abs": 0.14892, "relative": 0.922547 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.168358, "mean_abs": 0.087326, "relative": 0.776023 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 373, "start_frame": 1865, "end_frame": 1884, "center_frame": 1874, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.84145, "mean_abs": 0.229332, "relative": 0.830168 }, { "name": "hand_right_joints", "l2": 9.458223, "mean_abs": 0.237325, "relative": 0.745561 }, { "name": "body_joints", "l2": 6.043831, "mean_abs": 0.094769, "relative": 0.816335 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.051205, "mean_abs": 0.007767, "relative": 0.120836 }, { "name": "camera_rotation_matrix", "l2": 3.001867, "mean_abs": 0.213007, "relative": 0.921825 }, { "name": "imu_accel_gyro", "l2": 17.101532, "mean_abs": 1.057853, "relative": 0.922687 }, { "name": "depth_confidence", "l2": 10.968949, "mean_abs": 0.200126, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.036184, "mean_abs": 0.105387, "relative": 0.911945 }, { "name": "video_fisheye_cam1", "l2": 5.62869, "mean_abs": 0.120649, "relative": 0.930094 }, { "name": "video_fisheye_cam2", "l2": 5.776247, "mean_abs": 0.119526, "relative": 0.922092 }, { "name": "video_fisheye_cam3", "l2": 5.516412, "mean_abs": 0.114225, "relative": 0.923657 }, { "name": "video_stereo_left", "l2": 5.536456, "mean_abs": 0.12454, "relative": 0.899685 }, { "name": "video_stereo_right", "l2": 6.589636, "mean_abs": 0.150034, "relative": 0.929 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.146259, "mean_abs": 0.082804, "relative": 0.768114 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 374, "start_frame": 1870, "end_frame": 1889, "center_frame": 1879, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.819613, "mean_abs": 0.225674, "relative": 0.828326 }, { "name": "hand_right_joints", "l2": 9.453807, "mean_abs": 0.235173, "relative": 0.745213 }, { "name": "body_joints", "l2": 5.993269, "mean_abs": 0.09404, "relative": 0.809506 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.046846, "mean_abs": 0.007142, "relative": 0.11055 }, { "name": "camera_rotation_matrix", "l2": 3.000388, "mean_abs": 0.210715, "relative": 0.921371 }, { "name": "imu_accel_gyro", "l2": 17.081787, "mean_abs": 1.040331, "relative": 0.921621 }, { "name": "depth_confidence", "l2": 11.052956, "mean_abs": 0.200575, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.042643, "mean_abs": 0.105192, "relative": 0.913115 }, { "name": "video_fisheye_cam1", "l2": 5.617744, "mean_abs": 0.119823, "relative": 0.928285 }, { "name": "video_fisheye_cam2", "l2": 5.748018, "mean_abs": 0.117907, "relative": 0.917586 }, { "name": "video_fisheye_cam3", "l2": 5.487475, "mean_abs": 0.112496, "relative": 0.918812 }, { "name": "video_stereo_left", "l2": 5.492287, "mean_abs": 0.121867, "relative": 0.892508 }, { "name": "video_stereo_right", "l2": 6.566731, "mean_abs": 0.147881, "relative": 0.925771 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.994363, "mean_abs": 0.076162, "relative": 0.713753 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 375, "start_frame": 1875, "end_frame": 1894, "center_frame": 1884, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.814085, "mean_abs": 0.223144, "relative": 0.82786 }, { "name": "hand_right_joints", "l2": 9.436211, "mean_abs": 0.232471, "relative": 0.743826 }, { "name": "body_joints", "l2": 5.940283, "mean_abs": 0.093115, "relative": 0.802349 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.043575, "mean_abs": 0.006278, "relative": 0.102832 }, { "name": "camera_rotation_matrix", "l2": 2.999311, "mean_abs": 0.212042, "relative": 0.92104 }, { "name": "imu_accel_gyro", "l2": 17.074839, "mean_abs": 1.029038, "relative": 0.921246 }, { "name": "depth_confidence", "l2": 11.08884, "mean_abs": 0.200556, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.062356, "mean_abs": 0.106124, "relative": 0.916685 }, { "name": "video_fisheye_cam1", "l2": 5.586674, "mean_abs": 0.119174, "relative": 0.923151 }, { "name": "video_fisheye_cam2", "l2": 5.738211, "mean_abs": 0.117955, "relative": 0.91602 }, { "name": "video_fisheye_cam3", "l2": 5.471222, "mean_abs": 0.111701, "relative": 0.91609 }, { "name": "video_stereo_left", "l2": 5.47691, "mean_abs": 0.121527, "relative": 0.890009 }, { "name": "video_stereo_right", "l2": 6.560889, "mean_abs": 0.147702, "relative": 0.924948 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.133479, "mean_abs": 0.084568, "relative": 0.76354 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 376, "start_frame": 1880, "end_frame": 1899, "center_frame": 1889, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.806965, "mean_abs": 0.220394, "relative": 0.827259 }, { "name": "hand_right_joints", "l2": 9.331573, "mean_abs": 0.227327, "relative": 0.735578 }, { "name": "body_joints", "l2": 5.915676, "mean_abs": 0.092069, "relative": 0.799025 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.03447, "mean_abs": 0.004814, "relative": 0.081345 }, { "name": "camera_rotation_matrix", "l2": 3.000438, "mean_abs": 0.212318, "relative": 0.921386 }, { "name": "imu_accel_gyro", "l2": 17.101595, "mean_abs": 1.021971, "relative": 0.92269 }, { "name": "depth_confidence", "l2": 11.095958, "mean_abs": 0.20024, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.086246, "mean_abs": 0.106584, "relative": 0.92101 }, { "name": "video_fisheye_cam1", "l2": 5.594495, "mean_abs": 0.119093, "relative": 0.924443 }, { "name": "video_fisheye_cam2", "l2": 5.742205, "mean_abs": 0.118187, "relative": 0.916658 }, { "name": "video_fisheye_cam3", "l2": 5.486045, "mean_abs": 0.111993, "relative": 0.918572 }, { "name": "video_stereo_left", "l2": 5.496927, "mean_abs": 0.121935, "relative": 0.893262 }, { "name": "video_stereo_right", "l2": 6.621238, "mean_abs": 0.148896, "relative": 0.933456 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.193192, "mean_abs": 0.087738, "relative": 0.78491 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 377, "start_frame": 1885, "end_frame": 1904, "center_frame": 1894, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.803063, "mean_abs": 0.217932, "relative": 0.82693 }, { "name": "hand_right_joints", "l2": 9.274956, "mean_abs": 0.219718, "relative": 0.731115 }, { "name": "body_joints", "l2": 5.841639, "mean_abs": 0.089179, "relative": 0.789025 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021829, "mean_abs": 0.003282, "relative": 0.051513 }, { "name": "camera_rotation_matrix", "l2": 3.001713, "mean_abs": 0.212839, "relative": 0.921777 }, { "name": "imu_accel_gyro", "l2": 17.042738, "mean_abs": 1.034951, "relative": 0.919514 }, { "name": "depth_confidence", "l2": 11.099134, "mean_abs": 0.199536, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.108537, "mean_abs": 0.106979, "relative": 0.925047 }, { "name": "video_fisheye_cam1", "l2": 5.573927, "mean_abs": 0.117986, "relative": 0.921044 }, { "name": "video_fisheye_cam2", "l2": 5.741057, "mean_abs": 0.117729, "relative": 0.916474 }, { "name": "video_fisheye_cam3", "l2": 5.503939, "mean_abs": 0.112266, "relative": 0.921568 }, { "name": "video_stereo_left", "l2": 5.496046, "mean_abs": 0.120554, "relative": 0.893118 }, { "name": "video_stereo_right", "l2": 6.628598, "mean_abs": 0.146468, "relative": 0.934493 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.140247, "mean_abs": 0.083978, "relative": 0.765962 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 378, "start_frame": 1890, "end_frame": 1909, "center_frame": 1899, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.769926, "mean_abs": 0.216793, "relative": 0.824135 }, { "name": "hand_right_joints", "l2": 9.261652, "mean_abs": 0.217349, "relative": 0.730066 }, { "name": "body_joints", "l2": 5.804899, "mean_abs": 0.086632, "relative": 0.784063 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018707, "mean_abs": 0.002943, "relative": 0.044145 }, { "name": "camera_rotation_matrix", "l2": 3.000376, "mean_abs": 0.213053, "relative": 0.921367 }, { "name": "imu_accel_gyro", "l2": 17.039694, "mean_abs": 1.032961, "relative": 0.91935 }, { "name": "depth_confidence", "l2": 11.074677, "mean_abs": 0.199992, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.126327, "mean_abs": 0.106802, "relative": 0.928268 }, { "name": "video_fisheye_cam1", "l2": 5.513188, "mean_abs": 0.116789, "relative": 0.911008 }, { "name": "video_fisheye_cam2", "l2": 5.717513, "mean_abs": 0.117529, "relative": 0.912716 }, { "name": "video_fisheye_cam3", "l2": 5.522756, "mean_abs": 0.112224, "relative": 0.924719 }, { "name": "video_stereo_left", "l2": 5.488827, "mean_abs": 0.121359, "relative": 0.891945 }, { "name": "video_stereo_right", "l2": 6.55889, "mean_abs": 0.146365, "relative": 0.924666 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.150052, "mean_abs": 0.086527, "relative": 0.769471 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 379, "start_frame": 1895, "end_frame": 1914, "center_frame": 1904, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.732209, "mean_abs": 0.216099, "relative": 0.820953 }, { "name": "hand_right_joints", "l2": 9.22236, "mean_abs": 0.214623, "relative": 0.726969 }, { "name": "body_joints", "l2": 5.829525, "mean_abs": 0.08802, "relative": 0.787389 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.031131, "mean_abs": 0.003989, "relative": 0.073464 }, { "name": "camera_rotation_matrix", "l2": 3.001442, "mean_abs": 0.214796, "relative": 0.921694 }, { "name": "imu_accel_gyro", "l2": 17.051466, "mean_abs": 1.029216, "relative": 0.919985 }, { "name": "depth_confidence", "l2": 11.161093, "mean_abs": 0.201957, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.133442, "mean_abs": 0.107175, "relative": 0.929557 }, { "name": "video_fisheye_cam1", "l2": 5.49638, "mean_abs": 0.11767, "relative": 0.90823 }, { "name": "video_fisheye_cam2", "l2": 5.682431, "mean_abs": 0.117959, "relative": 0.907116 }, { "name": "video_fisheye_cam3", "l2": 5.540194, "mean_abs": 0.112864, "relative": 0.927639 }, { "name": "video_stereo_left", "l2": 5.467261, "mean_abs": 0.123365, "relative": 0.888441 }, { "name": "video_stereo_right", "l2": 6.531274, "mean_abs": 0.149062, "relative": 0.920773 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.048796, "mean_abs": 0.076711, "relative": 0.733233 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 380, "start_frame": 1900, "end_frame": 1919, "center_frame": 1909, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.714021, "mean_abs": 0.216649, "relative": 0.819419 }, { "name": "hand_right_joints", "l2": 9.243092, "mean_abs": 0.216332, "relative": 0.728603 }, { "name": "body_joints", "l2": 5.882611, "mean_abs": 0.090555, "relative": 0.794559 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.04968, "mean_abs": 0.005918, "relative": 0.117237 }, { "name": "camera_rotation_matrix", "l2": 3.003277, "mean_abs": 0.216321, "relative": 0.922258 }, { "name": "imu_accel_gyro", "l2": 17.027054, "mean_abs": 1.044786, "relative": 0.918668 }, { "name": "depth_confidence", "l2": 11.173311, "mean_abs": 0.204485, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.144331, "mean_abs": 0.107751, "relative": 0.931528 }, { "name": "video_fisheye_cam1", "l2": 5.459914, "mean_abs": 0.117623, "relative": 0.902205 }, { "name": "video_fisheye_cam2", "l2": 5.643554, "mean_abs": 0.118232, "relative": 0.900909 }, { "name": "video_fisheye_cam3", "l2": 5.546136, "mean_abs": 0.113591, "relative": 0.928634 }, { "name": "video_stereo_left", "l2": 5.44437, "mean_abs": 0.125397, "relative": 0.884721 }, { "name": "video_stereo_right", "l2": 6.458526, "mean_abs": 0.149096, "relative": 0.910517 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.05973, "mean_abs": 0.077731, "relative": 0.737146 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 381, "start_frame": 1905, "end_frame": 1924, "center_frame": 1914, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.711769, "mean_abs": 0.213256, "relative": 0.819229 }, { "name": "hand_right_joints", "l2": 9.397765, "mean_abs": 0.220908, "relative": 0.740795 }, { "name": "body_joints", "l2": 5.980564, "mean_abs": 0.094503, "relative": 0.80779 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.082627, "mean_abs": 0.010399, "relative": 0.194989 }, { "name": "camera_rotation_matrix", "l2": 3.007584, "mean_abs": 0.217733, "relative": 0.92358 }, { "name": "imu_accel_gyro", "l2": 17.034822, "mean_abs": 1.050008, "relative": 0.919087 }, { "name": "depth_confidence", "l2": 11.155043, "mean_abs": 0.205602, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.145336, "mean_abs": 0.108321, "relative": 0.93171 }, { "name": "video_fisheye_cam1", "l2": 5.420325, "mean_abs": 0.117546, "relative": 0.895663 }, { "name": "video_fisheye_cam2", "l2": 5.589823, "mean_abs": 0.118202, "relative": 0.892332 }, { "name": "video_fisheye_cam3", "l2": 5.547019, "mean_abs": 0.114155, "relative": 0.928782 }, { "name": "video_stereo_left", "l2": 5.378826, "mean_abs": 0.124814, "relative": 0.87407 }, { "name": "video_stereo_right", "l2": 6.312993, "mean_abs": 0.14664, "relative": 0.89 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.038281, "mean_abs": 0.077226, "relative": 0.72947 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 382, "start_frame": 1910, "end_frame": 1929, "center_frame": 1919, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.839031, "mean_abs": 0.221541, "relative": 0.829964 }, { "name": "hand_right_joints", "l2": 9.543279, "mean_abs": 0.232555, "relative": 0.752266 }, { "name": "body_joints", "l2": 6.185809, "mean_abs": 0.099745, "relative": 0.835512 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.128527, "mean_abs": 0.016802, "relative": 0.303305 }, { "name": "camera_rotation_matrix", "l2": 3.009983, "mean_abs": 0.218324, "relative": 0.924317 }, { "name": "imu_accel_gyro", "l2": 17.052347, "mean_abs": 1.050262, "relative": 0.920033 }, { "name": "depth_confidence", "l2": 11.073103, "mean_abs": 0.205502, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.139086, "mean_abs": 0.108333, "relative": 0.930579 }, { "name": "video_fisheye_cam1", "l2": 5.364787, "mean_abs": 0.116945, "relative": 0.886486 }, { "name": "video_fisheye_cam2", "l2": 5.537304, "mean_abs": 0.117402, "relative": 0.883948 }, { "name": "video_fisheye_cam3", "l2": 5.569718, "mean_abs": 0.114942, "relative": 0.932582 }, { "name": "video_stereo_left", "l2": 5.301558, "mean_abs": 0.123398, "relative": 0.861514 }, { "name": "video_stereo_right", "l2": 6.152771, "mean_abs": 0.143416, "relative": 0.867412 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.094582, "mean_abs": 0.08029, "relative": 0.749619 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 383, "start_frame": 1915, "end_frame": 1934, "center_frame": 1924, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.945955, "mean_abs": 0.223143, "relative": 0.838984 }, { "name": "hand_right_joints", "l2": 9.888312, "mean_abs": 0.260786, "relative": 0.779464 }, { "name": "body_joints", "l2": 6.49495, "mean_abs": 0.103849, "relative": 0.877268 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.181936, "mean_abs": 0.02386, "relative": 0.429344 }, { "name": "camera_rotation_matrix", "l2": 3.011775, "mean_abs": 0.216919, "relative": 0.924867 }, { "name": "imu_accel_gyro", "l2": 17.105429, "mean_abs": 1.103291, "relative": 0.922897 }, { "name": "depth_confidence", "l2": 10.904178, "mean_abs": 0.204948, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.125039, "mean_abs": 0.107785, "relative": 0.928035 }, { "name": "video_fisheye_cam1", "l2": 5.286983, "mean_abs": 0.116007, "relative": 0.873629 }, { "name": "video_fisheye_cam2", "l2": 5.462188, "mean_abs": 0.115878, "relative": 0.871957 }, { "name": "video_fisheye_cam3", "l2": 5.597654, "mean_abs": 0.116279, "relative": 0.93726 }, { "name": "video_stereo_left", "l2": 5.206398, "mean_abs": 0.121658, "relative": 0.84605 }, { "name": "video_stereo_right", "l2": 6.030411, "mean_abs": 0.142685, "relative": 0.850161 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.038852, "mean_abs": 0.078193, "relative": 0.729675 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 384, "start_frame": 1920, "end_frame": 1939, "center_frame": 1929, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.009666, "mean_abs": 0.226075, "relative": 0.844358 }, { "name": "hand_right_joints", "l2": 10.547745, "mean_abs": 0.292003, "relative": 0.831444 }, { "name": "body_joints", "l2": 6.713742, "mean_abs": 0.109541, "relative": 0.90682 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.209682, "mean_abs": 0.027647, "relative": 0.49482 }, { "name": "camera_rotation_matrix", "l2": 3.00797, "mean_abs": 0.211472, "relative": 0.923699 }, { "name": "imu_accel_gyro", "l2": 17.121347, "mean_abs": 1.140146, "relative": 0.923756 }, { "name": "depth_confidence", "l2": 10.725698, "mean_abs": 0.203429, "relative": 0.000356 }, { "name": "video_fisheye_cam0", "l2": 5.121396, "mean_abs": 0.10742, "relative": 0.927375 }, { "name": "video_fisheye_cam1", "l2": 5.232763, "mean_abs": 0.114273, "relative": 0.86467 }, { "name": "video_fisheye_cam2", "l2": 5.406163, "mean_abs": 0.113658, "relative": 0.863013 }, { "name": "video_fisheye_cam3", "l2": 5.664528, "mean_abs": 0.118722, "relative": 0.948457 }, { "name": "video_stereo_left", "l2": 5.112134, "mean_abs": 0.118548, "relative": 0.830732 }, { "name": "video_stereo_right", "l2": 5.936655, "mean_abs": 0.140549, "relative": 0.836944 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.091119, "mean_abs": 0.082107, "relative": 0.74838 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 385, "start_frame": 1925, "end_frame": 1944, "center_frame": 1934, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.119594, "mean_abs": 0.229535, "relative": 0.853631 }, { "name": "hand_right_joints", "l2": 11.097045, "mean_abs": 0.312079, "relative": 0.874744 }, { "name": "body_joints", "l2": 6.834147, "mean_abs": 0.112761, "relative": 0.923083 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.208346, "mean_abs": 0.02674, "relative": 0.491668 }, { "name": "camera_rotation_matrix", "l2": 3.009799, "mean_abs": 0.206631, "relative": 0.92426 }, { "name": "imu_accel_gyro", "l2": 17.150249, "mean_abs": 1.161182, "relative": 0.925315 }, { "name": "depth_confidence", "l2": 10.418454, "mean_abs": 0.198947, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.119375, "mean_abs": 0.107289, "relative": 0.927009 }, { "name": "video_fisheye_cam1", "l2": 5.147928, "mean_abs": 0.112933, "relative": 0.850652 }, { "name": "video_fisheye_cam2", "l2": 5.364637, "mean_abs": 0.111723, "relative": 0.856385 }, { "name": "video_fisheye_cam3", "l2": 5.681076, "mean_abs": 0.119765, "relative": 0.951228 }, { "name": "video_stereo_left", "l2": 5.083014, "mean_abs": 0.116881, "relative": 0.826 }, { "name": "video_stereo_right", "l2": 5.829729, "mean_abs": 0.137209, "relative": 0.821869 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.028609, "mean_abs": 0.077645, "relative": 0.726009 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 386, "start_frame": 1930, "end_frame": 1949, "center_frame": 1939, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.143889, "mean_abs": 0.225719, "relative": 0.85568 }, { "name": "hand_right_joints", "l2": 11.388841, "mean_abs": 0.318821, "relative": 0.897745 }, { "name": "body_joints", "l2": 6.900315, "mean_abs": 0.110692, "relative": 0.93202 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.178074, "mean_abs": 0.022997, "relative": 0.420229 }, { "name": "camera_rotation_matrix", "l2": 3.01028, "mean_abs": 0.20413, "relative": 0.924408 }, { "name": "imu_accel_gyro", "l2": 17.207623, "mean_abs": 1.150216, "relative": 0.92841 }, { "name": "depth_confidence", "l2": 10.149562, "mean_abs": 0.19452, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.115256, "mean_abs": 0.107041, "relative": 0.926264 }, { "name": "video_fisheye_cam1", "l2": 5.082023, "mean_abs": 0.110795, "relative": 0.839761 }, { "name": "video_fisheye_cam2", "l2": 5.368219, "mean_abs": 0.111526, "relative": 0.856956 }, { "name": "video_fisheye_cam3", "l2": 5.649022, "mean_abs": 0.119557, "relative": 0.945861 }, { "name": "video_stereo_left", "l2": 5.081226, "mean_abs": 0.116331, "relative": 0.825709 }, { "name": "video_stereo_right", "l2": 5.814761, "mean_abs": 0.135978, "relative": 0.819759 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.107452, "mean_abs": 0.0836, "relative": 0.754226 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 387, "start_frame": 1935, "end_frame": 1954, "center_frame": 1944, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.125953, "mean_abs": 0.219441, "relative": 0.854167 }, { "name": "hand_right_joints", "l2": 11.729875, "mean_abs": 0.324101, "relative": 0.924628 }, { "name": "body_joints", "l2": 6.89498, "mean_abs": 0.102927, "relative": 0.931299 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.115682, "mean_abs": 0.016487, "relative": 0.272992 }, { "name": "camera_rotation_matrix", "l2": 3.008827, "mean_abs": 0.203826, "relative": 0.923962 }, { "name": "imu_accel_gyro", "l2": 17.285698, "mean_abs": 1.110543, "relative": 0.932623 }, { "name": "depth_confidence", "l2": 9.93673, "mean_abs": 0.190593, "relative": 0.00033 }, { "name": "video_fisheye_cam0", "l2": 5.128193, "mean_abs": 0.107219, "relative": 0.928606 }, { "name": "video_fisheye_cam1", "l2": 5.056812, "mean_abs": 0.109929, "relative": 0.835595 }, { "name": "video_fisheye_cam2", "l2": 5.415648, "mean_abs": 0.112806, "relative": 0.864528 }, { "name": "video_fisheye_cam3", "l2": 5.617246, "mean_abs": 0.118804, "relative": 0.94054 }, { "name": "video_stereo_left", "l2": 5.131391, "mean_abs": 0.117412, "relative": 0.833861 }, { "name": "video_stereo_right", "l2": 5.879198, "mean_abs": 0.136638, "relative": 0.828844 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.516621, "mean_abs": 0.103699, "relative": 0.900661 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 388, "start_frame": 1940, "end_frame": 1959, "center_frame": 1949, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.183931, "mean_abs": 0.212168, "relative": 0.859058 }, { "name": "hand_right_joints", "l2": 12.152624, "mean_abs": 0.323947, "relative": 0.957952 }, { "name": "body_joints", "l2": 6.826167, "mean_abs": 0.096363, "relative": 0.922005 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.071174, "mean_abs": 0.011115, "relative": 0.16796 }, { "name": "camera_rotation_matrix", "l2": 3.006972, "mean_abs": 0.202331, "relative": 0.923392 }, { "name": "imu_accel_gyro", "l2": 17.266411, "mean_abs": 1.090742, "relative": 0.931582 }, { "name": "depth_confidence", "l2": 9.82094, "mean_abs": 0.187091, "relative": 0.000326 }, { "name": "video_fisheye_cam0", "l2": 5.132329, "mean_abs": 0.107209, "relative": 0.929355 }, { "name": "video_fisheye_cam1", "l2": 5.032366, "mean_abs": 0.109038, "relative": 0.831556 }, { "name": "video_fisheye_cam2", "l2": 5.470804, "mean_abs": 0.113499, "relative": 0.873332 }, { "name": "video_fisheye_cam3", "l2": 5.553787, "mean_abs": 0.116782, "relative": 0.929915 }, { "name": "video_stereo_left", "l2": 5.214371, "mean_abs": 0.118131, "relative": 0.847346 }, { "name": "video_stereo_right", "l2": 5.934664, "mean_abs": 0.136356, "relative": 0.836663 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.341348, "mean_abs": 0.09396, "relative": 0.837933 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010882, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 389, "start_frame": 1945, "end_frame": 1964, "center_frame": 1954, "action": "Transfer coffee to dripper", "subtask": "Transfer coffee grounds to dripper", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale", "closed coffee container" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.202674, "mean_abs": 0.216067, "relative": 0.860639 }, { "name": "hand_right_joints", "l2": 12.219134, "mean_abs": 0.32692, "relative": 0.963195 }, { "name": "body_joints", "l2": 6.837286, "mean_abs": 0.095442, "relative": 0.923507 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.046382, "mean_abs": 0.006702, "relative": 0.109454 }, { "name": "camera_rotation_matrix", "l2": 3.004108, "mean_abs": 0.19859, "relative": 0.922513 }, { "name": "imu_accel_gyro", "l2": 17.378456, "mean_abs": 1.115635, "relative": 0.937628 }, { "name": "depth_confidence", "l2": 9.678542, "mean_abs": 0.184427, "relative": 0.000321 }, { "name": "video_fisheye_cam0", "l2": 5.131931, "mean_abs": 0.106909, "relative": 0.929283 }, { "name": "video_fisheye_cam1", "l2": 5.026733, "mean_abs": 0.10869, "relative": 0.830625 }, { "name": "video_fisheye_cam2", "l2": 5.514827, "mean_abs": 0.113679, "relative": 0.88036 }, { "name": "video_fisheye_cam3", "l2": 5.486164, "mean_abs": 0.11393, "relative": 0.918592 }, { "name": "video_stereo_left", "l2": 5.306429, "mean_abs": 0.119499, "relative": 0.862305 }, { "name": "video_stereo_right", "l2": 6.038513, "mean_abs": 0.138503, "relative": 0.851304 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.263792, "mean_abs": 0.091371, "relative": 0.810177 }, { "name": "caption_objects_interaction_text", "l2": 1.885409, "mean_abs": 0.016567, "relative": 0.865734 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 390, "start_frame": 1950, "end_frame": 1969, "center_frame": 1959, "action": "Transfer coffee to dripper", "subtask": "", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale", "closed coffee container" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.224007, "mean_abs": 0.22052, "relative": 0.862439 }, { "name": "hand_right_joints", "l2": 11.903557, "mean_abs": 0.326706, "relative": 0.938319 }, { "name": "body_joints", "l2": 6.723825, "mean_abs": 0.098748, "relative": 0.908182 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.065598, "mean_abs": 0.009011, "relative": 0.154802 }, { "name": "camera_rotation_matrix", "l2": 3.001818, "mean_abs": 0.194572, "relative": 0.921809 }, { "name": "imu_accel_gyro", "l2": 17.175774, "mean_abs": 1.122058, "relative": 0.926692 }, { "name": "depth_confidence", "l2": 9.632365, "mean_abs": 0.184308, "relative": 0.00032 }, { "name": "video_fisheye_cam0", "l2": 5.130331, "mean_abs": 0.106601, "relative": 0.928993 }, { "name": "video_fisheye_cam1", "l2": 5.046335, "mean_abs": 0.109288, "relative": 0.833864 }, { "name": "video_fisheye_cam2", "l2": 5.543857, "mean_abs": 0.113676, "relative": 0.884994 }, { "name": "video_fisheye_cam3", "l2": 5.477067, "mean_abs": 0.113118, "relative": 0.917069 }, { "name": "video_stereo_left", "l2": 5.343927, "mean_abs": 0.119592, "relative": 0.868399 }, { "name": "video_stereo_right", "l2": 6.080405, "mean_abs": 0.139987, "relative": 0.85721 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.252897, "mean_abs": 0.089277, "relative": 0.806278 }, { "name": "caption_objects_interaction_text", "l2": 1.885409, "mean_abs": 0.016795, "relative": 0.865734 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 391, "start_frame": 1955, "end_frame": 1974, "center_frame": 1964, "action": "Transfer coffee to dripper", "subtask": "Secure coffee container", "objects": [ "coffee container", "lid", "coffee scoop", "dripper", "scale", "closed coffee container" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.22821, "mean_abs": 0.221536, "relative": 0.862793 }, { "name": "hand_right_joints", "l2": 11.704214, "mean_abs": 0.322587, "relative": 0.922605 }, { "name": "body_joints", "l2": 6.531216, "mean_abs": 0.098353, "relative": 0.882166 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.093347, "mean_abs": 0.013213, "relative": 0.220287 }, { "name": "camera_rotation_matrix", "l2": 3.00062, "mean_abs": 0.19176, "relative": 0.921442 }, { "name": "imu_accel_gyro", "l2": 17.111574, "mean_abs": 1.096123, "relative": 0.923228 }, { "name": "depth_confidence", "l2": 9.622285, "mean_abs": 0.182455, "relative": 0.000319 }, { "name": "video_fisheye_cam0", "l2": 5.132648, "mean_abs": 0.106477, "relative": 0.929413 }, { "name": "video_fisheye_cam1", "l2": 5.111849, "mean_abs": 0.110212, "relative": 0.84469 }, { "name": "video_fisheye_cam2", "l2": 5.566793, "mean_abs": 0.113113, "relative": 0.888656 }, { "name": "video_fisheye_cam3", "l2": 5.496027, "mean_abs": 0.113577, "relative": 0.920243 }, { "name": "video_stereo_left", "l2": 5.372325, "mean_abs": 0.119224, "relative": 0.873014 }, { "name": "video_stereo_right", "l2": 6.14659, "mean_abs": 0.140791, "relative": 0.86654 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.477004, "mean_abs": 0.100495, "relative": 0.886482 }, { "name": "caption_objects_interaction_text", "l2": 1.885409, "mean_abs": 0.01658, "relative": 0.865734 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 392, "start_frame": 1960, "end_frame": 1979, "center_frame": 1969, "action": "Transfer coffee to dripper", "subtask": "Secure coffee container", "objects": [ "closed coffee container", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.240627, "mean_abs": 0.220876, "relative": 0.863841 }, { "name": "hand_right_joints", "l2": 11.610445, "mean_abs": 0.319236, "relative": 0.915214 }, { "name": "body_joints", "l2": 6.446451, "mean_abs": 0.097623, "relative": 0.870717 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.116875, "mean_abs": 0.016398, "relative": 0.275809 }, { "name": "camera_rotation_matrix", "l2": 2.999081, "mean_abs": 0.191637, "relative": 0.920969 }, { "name": "imu_accel_gyro", "l2": 17.104816, "mean_abs": 1.091484, "relative": 0.922864 }, { "name": "depth_confidence", "l2": 9.69163, "mean_abs": 0.183846, "relative": 0.000322 }, { "name": "video_fisheye_cam0", "l2": 5.142078, "mean_abs": 0.106745, "relative": 0.93112 }, { "name": "video_fisheye_cam1", "l2": 5.171624, "mean_abs": 0.111663, "relative": 0.854567 }, { "name": "video_fisheye_cam2", "l2": 5.583195, "mean_abs": 0.113516, "relative": 0.891274 }, { "name": "video_fisheye_cam3", "l2": 5.500902, "mean_abs": 0.114808, "relative": 0.92106 }, { "name": "video_stereo_left", "l2": 5.408278, "mean_abs": 0.119545, "relative": 0.878856 }, { "name": "video_stereo_right", "l2": 6.224364, "mean_abs": 0.141968, "relative": 0.877505 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.077768, "mean_abs": 0.082206, "relative": 0.743602 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01096, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 393, "start_frame": 1965, "end_frame": 1984, "center_frame": 1974, "action": "Transfer coffee to dripper", "subtask": "Secure coffee container", "objects": [ "closed coffee container", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.319805, "mean_abs": 0.220929, "relative": 0.87052 }, { "name": "hand_right_joints", "l2": 11.635808, "mean_abs": 0.321281, "relative": 0.917213 }, { "name": "body_joints", "l2": 6.344884, "mean_abs": 0.096498, "relative": 0.856998 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.130145, "mean_abs": 0.017464, "relative": 0.307124 }, { "name": "camera_rotation_matrix", "l2": 3.001884, "mean_abs": 0.196728, "relative": 0.92183 }, { "name": "imu_accel_gyro", "l2": 17.116583, "mean_abs": 1.122877, "relative": 0.923499 }, { "name": "depth_confidence", "l2": 9.89005, "mean_abs": 0.190435, "relative": 0.000328 }, { "name": "video_fisheye_cam0", "l2": 5.141737, "mean_abs": 0.107505, "relative": 0.931059 }, { "name": "video_fisheye_cam1", "l2": 5.220408, "mean_abs": 0.112844, "relative": 0.862628 }, { "name": "video_fisheye_cam2", "l2": 5.607142, "mean_abs": 0.114933, "relative": 0.895097 }, { "name": "video_fisheye_cam3", "l2": 5.507168, "mean_abs": 0.115391, "relative": 0.922109 }, { "name": "video_stereo_left", "l2": 5.44076, "mean_abs": 0.121932, "relative": 0.884134 }, { "name": "video_stereo_right", "l2": 6.277957, "mean_abs": 0.14433, "relative": 0.88506 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.10858, "mean_abs": 0.081778, "relative": 0.754629 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01096, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 394, "start_frame": 1970, "end_frame": 1989, "center_frame": 1979, "action": "Transfer coffee to dripper", "subtask": "Secure coffee container", "objects": [ "closed coffee container", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.371699, "mean_abs": 0.225202, "relative": 0.874897 }, { "name": "hand_right_joints", "l2": 11.515804, "mean_abs": 0.295081, "relative": 0.907753 }, { "name": "body_joints", "l2": 5.897405, "mean_abs": 0.090837, "relative": 0.796558 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.107323, "mean_abs": 0.014195, "relative": 0.253266 }, { "name": "camera_rotation_matrix", "l2": 3.004852, "mean_abs": 0.198866, "relative": 0.922741 }, { "name": "imu_accel_gyro", "l2": 17.003563, "mean_abs": 1.108284, "relative": 0.917401 }, { "name": "depth_confidence", "l2": 10.038894, "mean_abs": 0.192846, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.139276, "mean_abs": 0.107635, "relative": 0.930613 }, { "name": "video_fisheye_cam1", "l2": 5.274603, "mean_abs": 0.113551, "relative": 0.871584 }, { "name": "video_fisheye_cam2", "l2": 5.633739, "mean_abs": 0.115605, "relative": 0.899343 }, { "name": "video_fisheye_cam3", "l2": 5.494148, "mean_abs": 0.11467, "relative": 0.919929 }, { "name": "video_stereo_left", "l2": 5.453016, "mean_abs": 0.122262, "relative": 0.886126 }, { "name": "video_stereo_right", "l2": 6.312741, "mean_abs": 0.145308, "relative": 0.889964 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.091255, "mean_abs": 0.081627, "relative": 0.748429 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01096, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 395, "start_frame": 1975, "end_frame": 1994, "center_frame": 1984, "action": "Transfer coffee to dripper", "subtask": "Secure coffee container", "objects": [ "closed coffee container", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.435087, "mean_abs": 0.227994, "relative": 0.880244 }, { "name": "hand_right_joints", "l2": 11.717416, "mean_abs": 0.279044, "relative": 0.923646 }, { "name": "body_joints", "l2": 5.61223, "mean_abs": 0.085101, "relative": 0.758039 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.071862, "mean_abs": 0.009699, "relative": 0.169583 }, { "name": "camera_rotation_matrix", "l2": 3.001173, "mean_abs": 0.199748, "relative": 0.921611 }, { "name": "imu_accel_gyro", "l2": 17.06237, "mean_abs": 1.099174, "relative": 0.920574 }, { "name": "depth_confidence", "l2": 10.19173, "mean_abs": 0.193232, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.135642, "mean_abs": 0.107322, "relative": 0.929955 }, { "name": "video_fisheye_cam1", "l2": 5.324171, "mean_abs": 0.113436, "relative": 0.879774 }, { "name": "video_fisheye_cam2", "l2": 5.656483, "mean_abs": 0.115451, "relative": 0.902973 }, { "name": "video_fisheye_cam3", "l2": 5.451123, "mean_abs": 0.112534, "relative": 0.912725 }, { "name": "video_stereo_left", "l2": 5.431753, "mean_abs": 0.121077, "relative": 0.882671 }, { "name": "video_stereo_right", "l2": 6.328186, "mean_abs": 0.144552, "relative": 0.892141 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.035723, "mean_abs": 0.07986, "relative": 0.728555 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01096, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 396, "start_frame": 1980, "end_frame": 1999, "center_frame": 1989, "action": "Transfer coffee to dripper", "subtask": "Secure coffee container", "objects": [ "closed coffee container", "dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.479756, "mean_abs": 0.226463, "relative": 0.884012 }, { "name": "hand_right_joints", "l2": 11.839748, "mean_abs": 0.275942, "relative": 0.933289 }, { "name": "body_joints", "l2": 5.514603, "mean_abs": 0.083126, "relative": 0.744853 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.042709, "mean_abs": 0.005693, "relative": 0.100788 }, { "name": "camera_rotation_matrix", "l2": 3.000393, "mean_abs": 0.20137, "relative": 0.921372 }, { "name": "imu_accel_gyro", "l2": 17.018507, "mean_abs": 1.08928, "relative": 0.918207 }, { "name": "depth_confidence", "l2": 10.365791, "mean_abs": 0.194898, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.126712, "mean_abs": 0.106679, "relative": 0.928338 }, { "name": "video_fisheye_cam1", "l2": 5.341458, "mean_abs": 0.113098, "relative": 0.882631 }, { "name": "video_fisheye_cam2", "l2": 5.679869, "mean_abs": 0.115179, "relative": 0.906707 }, { "name": "video_fisheye_cam3", "l2": 5.39282, "mean_abs": 0.109879, "relative": 0.902963 }, { "name": "video_stereo_left", "l2": 5.414521, "mean_abs": 0.118691, "relative": 0.87987 }, { "name": "video_stereo_right", "l2": 6.329265, "mean_abs": 0.142081, "relative": 0.892294 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.029225, "mean_abs": 0.078371, "relative": 0.726229 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01096, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 397, "start_frame": 1985, "end_frame": 2004, "center_frame": 1994, "action": "Transfer coffee to dripper", "subtask": "Secure coffee container", "objects": [ "closed coffee container", "dripper", "scale", "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.542141, "mean_abs": 0.224716, "relative": 0.889275 }, { "name": "hand_right_joints", "l2": 11.721061, "mean_abs": 0.274092, "relative": 0.923933 }, { "name": "body_joints", "l2": 5.46136, "mean_abs": 0.081309, "relative": 0.737662 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019716, "mean_abs": 0.002887, "relative": 0.046527 }, { "name": "camera_rotation_matrix", "l2": 3.000799, "mean_abs": 0.201387, "relative": 0.921497 }, { "name": "imu_accel_gyro", "l2": 17.109118, "mean_abs": 1.07641, "relative": 0.923096 }, { "name": "depth_confidence", "l2": 10.460066, "mean_abs": 0.194168, "relative": 0.000347 }, { "name": "video_fisheye_cam0", "l2": 5.118441, "mean_abs": 0.106191, "relative": 0.92684 }, { "name": "video_fisheye_cam1", "l2": 5.347411, "mean_abs": 0.11277, "relative": 0.883614 }, { "name": "video_fisheye_cam2", "l2": 5.686891, "mean_abs": 0.11478, "relative": 0.907827 }, { "name": "video_fisheye_cam3", "l2": 5.387069, "mean_abs": 0.109033, "relative": 0.902 }, { "name": "video_stereo_left", "l2": 5.389146, "mean_abs": 0.11638, "relative": 0.875747 }, { "name": "video_stereo_right", "l2": 6.315195, "mean_abs": 0.139765, "relative": 0.89031 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.042018, "mean_abs": 0.080346, "relative": 0.730808 }, { "name": "caption_objects_interaction_text", "l2": 2.041447, "mean_abs": 0.02019, "relative": 0.937383 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 398, "start_frame": 1990, "end_frame": 2009, "center_frame": 1999, "action": "", "subtask": "", "objects": [ "closed coffee container", "dripper", "scale", "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.521925, "mean_abs": 0.224258, "relative": 0.887569 }, { "name": "hand_right_joints", "l2": 11.522476, "mean_abs": 0.278019, "relative": 0.908279 }, { "name": "body_joints", "l2": 5.494554, "mean_abs": 0.081805, "relative": 0.742145 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017612, "mean_abs": 0.002602, "relative": 0.041561 }, { "name": "camera_rotation_matrix", "l2": 3.001607, "mean_abs": 0.202282, "relative": 0.921745 }, { "name": "imu_accel_gyro", "l2": 17.114357, "mean_abs": 1.082812, "relative": 0.923378 }, { "name": "depth_confidence", "l2": 10.492121, "mean_abs": 0.196729, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.113659, "mean_abs": 0.106226, "relative": 0.925974 }, { "name": "video_fisheye_cam1", "l2": 5.347951, "mean_abs": 0.11334, "relative": 0.883704 }, { "name": "video_fisheye_cam2", "l2": 5.680848, "mean_abs": 0.114923, "relative": 0.906863 }, { "name": "video_fisheye_cam3", "l2": 5.415653, "mean_abs": 0.11032, "relative": 0.906786 }, { "name": "video_stereo_left", "l2": 5.373117, "mean_abs": 0.117658, "relative": 0.873142 }, { "name": "video_stereo_right", "l2": 6.299114, "mean_abs": 0.141086, "relative": 0.888043 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.04622, "mean_abs": 0.08176, "relative": 0.732311 }, { "name": "caption_objects_interaction_text", "l2": 2.041447, "mean_abs": 0.020538, "relative": 0.937383 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 399, "start_frame": 1995, "end_frame": 2014, "center_frame": 2004, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "closed coffee container", "dripper", "scale", "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.487745, "mean_abs": 0.22476, "relative": 0.884686 }, { "name": "hand_right_joints", "l2": 11.38523, "mean_abs": 0.285311, "relative": 0.897461 }, { "name": "body_joints", "l2": 5.611731, "mean_abs": 0.084527, "relative": 0.757972 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.034848, "mean_abs": 0.004846, "relative": 0.082237 }, { "name": "camera_rotation_matrix", "l2": 3.002326, "mean_abs": 0.203904, "relative": 0.921965 }, { "name": "imu_accel_gyro", "l2": 17.140862, "mean_abs": 1.081391, "relative": 0.924809 }, { "name": "depth_confidence", "l2": 10.534232, "mean_abs": 0.198564, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.109997, "mean_abs": 0.106553, "relative": 0.925311 }, { "name": "video_fisheye_cam1", "l2": 5.362171, "mean_abs": 0.114361, "relative": 0.886053 }, { "name": "video_fisheye_cam2", "l2": 5.667742, "mean_abs": 0.115297, "relative": 0.904771 }, { "name": "video_fisheye_cam3", "l2": 5.475029, "mean_abs": 0.112243, "relative": 0.916728 }, { "name": "video_stereo_left", "l2": 5.370681, "mean_abs": 0.119233, "relative": 0.872746 }, { "name": "video_stereo_right", "l2": 6.315842, "mean_abs": 0.143353, "relative": 0.890401 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.989314, "mean_abs": 0.078109, "relative": 0.711946 }, { "name": "caption_objects_interaction_text", "l2": 2.041447, "mean_abs": 0.020186, "relative": 0.937383 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 400, "start_frame": 2000, "end_frame": 2019, "center_frame": 2009, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.425134, "mean_abs": 0.225148, "relative": 0.879404 }, { "name": "hand_right_joints", "l2": 11.170373, "mean_abs": 0.285633, "relative": 0.880524 }, { "name": "body_joints", "l2": 5.717457, "mean_abs": 0.086193, "relative": 0.772252 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.048766, "mean_abs": 0.006627, "relative": 0.115082 }, { "name": "camera_rotation_matrix", "l2": 3.004293, "mean_abs": 0.205061, "relative": 0.92257 }, { "name": "imu_accel_gyro", "l2": 17.134623, "mean_abs": 1.080011, "relative": 0.924472 }, { "name": "depth_confidence", "l2": 10.558311, "mean_abs": 0.19991, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.103755, "mean_abs": 0.106539, "relative": 0.924181 }, { "name": "video_fisheye_cam1", "l2": 5.370174, "mean_abs": 0.114715, "relative": 0.887376 }, { "name": "video_fisheye_cam2", "l2": 5.658919, "mean_abs": 0.115426, "relative": 0.903362 }, { "name": "video_fisheye_cam3", "l2": 5.517288, "mean_abs": 0.113665, "relative": 0.923803 }, { "name": "video_stereo_left", "l2": 5.37018, "mean_abs": 0.119795, "relative": 0.872665 }, { "name": "video_stereo_right", "l2": 6.331092, "mean_abs": 0.144256, "relative": 0.892551 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.999484, "mean_abs": 0.078355, "relative": 0.715585 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 401, "start_frame": 2005, "end_frame": 2024, "center_frame": 2014, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.375519, "mean_abs": 0.224219, "relative": 0.875219 }, { "name": "hand_right_joints", "l2": 10.999288, "mean_abs": 0.282892, "relative": 0.867038 }, { "name": "body_joints", "l2": 5.81411, "mean_abs": 0.085657, "relative": 0.785307 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.061345, "mean_abs": 0.007817, "relative": 0.144765 }, { "name": "camera_rotation_matrix", "l2": 3.002683, "mean_abs": 0.204626, "relative": 0.922075 }, { "name": "imu_accel_gyro", "l2": 17.059898, "mean_abs": 1.050128, "relative": 0.92044 }, { "name": "depth_confidence", "l2": 10.510556, "mean_abs": 0.199103, "relative": 0.000349 }, { "name": "video_fisheye_cam0", "l2": 5.10879, "mean_abs": 0.106618, "relative": 0.925093 }, { "name": "video_fisheye_cam1", "l2": 5.377578, "mean_abs": 0.114916, "relative": 0.888599 }, { "name": "video_fisheye_cam2", "l2": 5.647484, "mean_abs": 0.115232, "relative": 0.901537 }, { "name": "video_fisheye_cam3", "l2": 5.560372, "mean_abs": 0.114358, "relative": 0.931017 }, { "name": "video_stereo_left", "l2": 5.35659, "mean_abs": 0.118505, "relative": 0.870457 }, { "name": "video_stereo_right", "l2": 6.327069, "mean_abs": 0.14275, "relative": 0.891984 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.112501, "mean_abs": 0.081513, "relative": 0.756033 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 402, "start_frame": 2010, "end_frame": 2029, "center_frame": 2019, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.329762, "mean_abs": 0.222473, "relative": 0.87136 }, { "name": "hand_right_joints", "l2": 10.874642, "mean_abs": 0.277699, "relative": 0.857213 }, { "name": "body_joints", "l2": 5.901022, "mean_abs": 0.084745, "relative": 0.797046 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.055803, "mean_abs": 0.006543, "relative": 0.131687 }, { "name": "camera_rotation_matrix", "l2": 3.002375, "mean_abs": 0.202688, "relative": 0.921981 }, { "name": "imu_accel_gyro", "l2": 17.083832, "mean_abs": 1.015887, "relative": 0.921732 }, { "name": "depth_confidence", "l2": 10.523607, "mean_abs": 0.19747, "relative": 0.000349 }, { "name": "video_fisheye_cam0", "l2": 5.111617, "mean_abs": 0.106494, "relative": 0.925605 }, { "name": "video_fisheye_cam1", "l2": 5.386781, "mean_abs": 0.115168, "relative": 0.89012 }, { "name": "video_fisheye_cam2", "l2": 5.637401, "mean_abs": 0.114967, "relative": 0.899927 }, { "name": "video_fisheye_cam3", "l2": 5.60897, "mean_abs": 0.116, "relative": 0.939154 }, { "name": "video_stereo_left", "l2": 5.3603, "mean_abs": 0.118402, "relative": 0.871059 }, { "name": "video_stereo_right", "l2": 6.355877, "mean_abs": 0.143236, "relative": 0.896045 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.064568, "mean_abs": 0.079534, "relative": 0.738878 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 403, "start_frame": 2015, "end_frame": 2034, "center_frame": 2024, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.303697, "mean_abs": 0.221086, "relative": 0.869161 }, { "name": "hand_right_joints", "l2": 10.894135, "mean_abs": 0.278158, "relative": 0.858749 }, { "name": "body_joints", "l2": 5.99122, "mean_abs": 0.084437, "relative": 0.809229 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.047407, "mean_abs": 0.005918, "relative": 0.111874 }, { "name": "camera_rotation_matrix", "l2": 3.001455, "mean_abs": 0.201514, "relative": 0.921698 }, { "name": "imu_accel_gyro", "l2": 17.080481, "mean_abs": 1.005092, "relative": 0.921551 }, { "name": "depth_confidence", "l2": 10.53275, "mean_abs": 0.197305, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.109306, "mean_abs": 0.106521, "relative": 0.925186 }, { "name": "video_fisheye_cam1", "l2": 5.392096, "mean_abs": 0.11511, "relative": 0.890998 }, { "name": "video_fisheye_cam2", "l2": 5.631254, "mean_abs": 0.114754, "relative": 0.898946 }, { "name": "video_fisheye_cam3", "l2": 5.635349, "mean_abs": 0.116632, "relative": 0.943571 }, { "name": "video_stereo_left", "l2": 5.372306, "mean_abs": 0.118387, "relative": 0.87301 }, { "name": "video_stereo_right", "l2": 6.377003, "mean_abs": 0.143174, "relative": 0.899024 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.080181, "mean_abs": 0.077898, "relative": 0.744466 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 404, "start_frame": 2020, "end_frame": 2039, "center_frame": 2029, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.306195, "mean_abs": 0.221102, "relative": 0.869372 }, { "name": "hand_right_joints", "l2": 10.944276, "mean_abs": 0.280644, "relative": 0.862702 }, { "name": "body_joints", "l2": 6.04694, "mean_abs": 0.084043, "relative": 0.816755 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.03908, "mean_abs": 0.005698, "relative": 0.092223 }, { "name": "camera_rotation_matrix", "l2": 3.000665, "mean_abs": 0.20076, "relative": 0.921456 }, { "name": "imu_accel_gyro", "l2": 17.024399, "mean_abs": 1.005143, "relative": 0.918525 }, { "name": "depth_confidence", "l2": 10.529711, "mean_abs": 0.196595, "relative": 0.000349 }, { "name": "video_fisheye_cam0", "l2": 5.112371, "mean_abs": 0.10651, "relative": 0.925741 }, { "name": "video_fisheye_cam1", "l2": 5.386616, "mean_abs": 0.114684, "relative": 0.890093 }, { "name": "video_fisheye_cam2", "l2": 5.63056, "mean_abs": 0.114541, "relative": 0.898835 }, { "name": "video_fisheye_cam3", "l2": 5.668371, "mean_abs": 0.117401, "relative": 0.9491 }, { "name": "video_stereo_left", "l2": 5.385264, "mean_abs": 0.118058, "relative": 0.875116 }, { "name": "video_stereo_right", "l2": 6.391343, "mean_abs": 0.142649, "relative": 0.901045 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.096535, "mean_abs": 0.078253, "relative": 0.750318 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 405, "start_frame": 2025, "end_frame": 2044, "center_frame": 2034, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.326573, "mean_abs": 0.222757, "relative": 0.871091 }, { "name": "hand_right_joints", "l2": 10.959313, "mean_abs": 0.280103, "relative": 0.863887 }, { "name": "body_joints", "l2": 6.079676, "mean_abs": 0.083635, "relative": 0.821177 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.027808, "mean_abs": 0.004139, "relative": 0.065624 }, { "name": "camera_rotation_matrix", "l2": 3.000482, "mean_abs": 0.199854, "relative": 0.921399 }, { "name": "imu_accel_gyro", "l2": 17.054731, "mean_abs": 0.997416, "relative": 0.920161 }, { "name": "depth_confidence", "l2": 10.535695, "mean_abs": 0.195907, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.114826, "mean_abs": 0.106517, "relative": 0.926186 }, { "name": "video_fisheye_cam1", "l2": 5.37292, "mean_abs": 0.114491, "relative": 0.88783 }, { "name": "video_fisheye_cam2", "l2": 5.628157, "mean_abs": 0.114345, "relative": 0.898452 }, { "name": "video_fisheye_cam3", "l2": 5.662174, "mean_abs": 0.117087, "relative": 0.948063 }, { "name": "video_stereo_left", "l2": 5.397445, "mean_abs": 0.117566, "relative": 0.877096 }, { "name": "video_stereo_right", "l2": 6.401242, "mean_abs": 0.142045, "relative": 0.902441 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.066548, "mean_abs": 0.079941, "relative": 0.739586 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 406, "start_frame": 2030, "end_frame": 2049, "center_frame": 2039, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.371952, "mean_abs": 0.222426, "relative": 0.874918 }, { "name": "hand_right_joints", "l2": 10.994669, "mean_abs": 0.280234, "relative": 0.866674 }, { "name": "body_joints", "l2": 6.053609, "mean_abs": 0.082877, "relative": 0.817656 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019844, "mean_abs": 0.002783, "relative": 0.04683 }, { "name": "camera_rotation_matrix", "l2": 3.000086, "mean_abs": 0.19873, "relative": 0.921278 }, { "name": "imu_accel_gyro", "l2": 17.076527, "mean_abs": 1.00181, "relative": 0.921337 }, { "name": "depth_confidence", "l2": 10.501875, "mean_abs": 0.195058, "relative": 0.000349 }, { "name": "video_fisheye_cam0", "l2": 5.114265, "mean_abs": 0.106133, "relative": 0.926084 }, { "name": "video_fisheye_cam1", "l2": 5.361025, "mean_abs": 0.113949, "relative": 0.885864 }, { "name": "video_fisheye_cam2", "l2": 5.61933, "mean_abs": 0.113921, "relative": 0.897043 }, { "name": "video_fisheye_cam3", "l2": 5.669919, "mean_abs": 0.11669, "relative": 0.94936 }, { "name": "video_stereo_left", "l2": 5.402693, "mean_abs": 0.11706, "relative": 0.877948 }, { "name": "video_stereo_right", "l2": 6.403727, "mean_abs": 0.141391, "relative": 0.902791 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.192134, "mean_abs": 0.085554, "relative": 0.784532 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 407, "start_frame": 2035, "end_frame": 2054, "center_frame": 2044, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.408597, "mean_abs": 0.222539, "relative": 0.87801 }, { "name": "hand_right_joints", "l2": 11.004894, "mean_abs": 0.279437, "relative": 0.86748 }, { "name": "body_joints", "l2": 6.012625, "mean_abs": 0.08212, "relative": 0.81212 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015366, "mean_abs": 0.00217, "relative": 0.036261 }, { "name": "camera_rotation_matrix", "l2": 2.999863, "mean_abs": 0.197375, "relative": 0.921209 }, { "name": "imu_accel_gyro", "l2": 17.066154, "mean_abs": 1.002385, "relative": 0.920778 }, { "name": "depth_confidence", "l2": 10.466149, "mean_abs": 0.19383, "relative": 0.000347 }, { "name": "video_fisheye_cam0", "l2": 5.116707, "mean_abs": 0.106123, "relative": 0.926526 }, { "name": "video_fisheye_cam1", "l2": 5.355545, "mean_abs": 0.11361, "relative": 0.884959 }, { "name": "video_fisheye_cam2", "l2": 5.617661, "mean_abs": 0.113709, "relative": 0.896776 }, { "name": "video_fisheye_cam3", "l2": 5.683424, "mean_abs": 0.116983, "relative": 0.951621 }, { "name": "video_stereo_left", "l2": 5.417908, "mean_abs": 0.116712, "relative": 0.880421 }, { "name": "video_stereo_right", "l2": 6.410931, "mean_abs": 0.141229, "relative": 0.903807 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.016429, "mean_abs": 0.078, "relative": 0.72165 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 408, "start_frame": 2040, "end_frame": 2059, "center_frame": 2049, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.421867, "mean_abs": 0.22159, "relative": 0.879129 }, { "name": "hand_right_joints", "l2": 10.997714, "mean_abs": 0.277827, "relative": 0.866914 }, { "name": "body_joints", "l2": 5.971871, "mean_abs": 0.081799, "relative": 0.806616 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009398, "mean_abs": 0.00143, "relative": 0.022178 }, { "name": "camera_rotation_matrix", "l2": 3.000313, "mean_abs": 0.196902, "relative": 0.921347 }, { "name": "imu_accel_gyro", "l2": 17.053596, "mean_abs": 0.997501, "relative": 0.9201 }, { "name": "depth_confidence", "l2": 10.418299, "mean_abs": 0.192794, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.114665, "mean_abs": 0.105682, "relative": 0.926156 }, { "name": "video_fisheye_cam1", "l2": 5.340544, "mean_abs": 0.112793, "relative": 0.88248 }, { "name": "video_fisheye_cam2", "l2": 5.609137, "mean_abs": 0.113363, "relative": 0.895415 }, { "name": "video_fisheye_cam3", "l2": 5.637115, "mean_abs": 0.114786, "relative": 0.943867 }, { "name": "video_stereo_left", "l2": 5.410417, "mean_abs": 0.116039, "relative": 0.879203 }, { "name": "video_stereo_right", "l2": 6.393126, "mean_abs": 0.139975, "relative": 0.901297 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.033357, "mean_abs": 0.079727, "relative": 0.727708 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 409, "start_frame": 2045, "end_frame": 2064, "center_frame": 2054, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.425198, "mean_abs": 0.221283, "relative": 0.87941 }, { "name": "hand_right_joints", "l2": 10.995091, "mean_abs": 0.275231, "relative": 0.866707 }, { "name": "body_joints", "l2": 5.934928, "mean_abs": 0.08087, "relative": 0.801626 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008197, "mean_abs": 0.001356, "relative": 0.019344 }, { "name": "camera_rotation_matrix", "l2": 3.000067, "mean_abs": 0.196807, "relative": 0.921272 }, { "name": "imu_accel_gyro", "l2": 17.034739, "mean_abs": 0.997274, "relative": 0.919083 }, { "name": "depth_confidence", "l2": 10.393578, "mean_abs": 0.192088, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.114518, "mean_abs": 0.105566, "relative": 0.92613 }, { "name": "video_fisheye_cam1", "l2": 5.334642, "mean_abs": 0.112698, "relative": 0.881505 }, { "name": "video_fisheye_cam2", "l2": 5.605099, "mean_abs": 0.11335, "relative": 0.894771 }, { "name": "video_fisheye_cam3", "l2": 5.631701, "mean_abs": 0.114426, "relative": 0.942961 }, { "name": "video_stereo_left", "l2": 5.409744, "mean_abs": 0.115836, "relative": 0.879094 }, { "name": "video_stereo_right", "l2": 6.387211, "mean_abs": 0.139564, "relative": 0.900463 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.049857, "mean_abs": 0.081404, "relative": 0.733613 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 410, "start_frame": 2050, "end_frame": 2069, "center_frame": 2059, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.437686, "mean_abs": 0.220523, "relative": 0.880463 }, { "name": "hand_right_joints", "l2": 10.980423, "mean_abs": 0.274871, "relative": 0.865551 }, { "name": "body_joints", "l2": 5.905377, "mean_abs": 0.08014, "relative": 0.797635 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010909, "mean_abs": 0.001635, "relative": 0.025744 }, { "name": "camera_rotation_matrix", "l2": 3.000033, "mean_abs": 0.196111, "relative": 0.921261 }, { "name": "imu_accel_gyro", "l2": 17.03936, "mean_abs": 0.994037, "relative": 0.919332 }, { "name": "depth_confidence", "l2": 10.379226, "mean_abs": 0.191832, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.108167, "mean_abs": 0.105268, "relative": 0.92498 }, { "name": "video_fisheye_cam1", "l2": 5.329293, "mean_abs": 0.112368, "relative": 0.880621 }, { "name": "video_fisheye_cam2", "l2": 5.602607, "mean_abs": 0.113146, "relative": 0.894373 }, { "name": "video_fisheye_cam3", "l2": 5.591085, "mean_abs": 0.113359, "relative": 0.93616 }, { "name": "video_stereo_left", "l2": 5.409219, "mean_abs": 0.115604, "relative": 0.879009 }, { "name": "video_stereo_right", "l2": 6.378366, "mean_abs": 0.139011, "relative": 0.899216 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.056278, "mean_abs": 0.080619, "relative": 0.735911 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 411, "start_frame": 2055, "end_frame": 2074, "center_frame": 2064, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.424678, "mean_abs": 0.220474, "relative": 0.879366 }, { "name": "hand_right_joints", "l2": 10.981993, "mean_abs": 0.273166, "relative": 0.865675 }, { "name": "body_joints", "l2": 5.873747, "mean_abs": 0.079241, "relative": 0.793362 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012159, "mean_abs": 0.00159, "relative": 0.028693 }, { "name": "camera_rotation_matrix", "l2": 2.99995, "mean_abs": 0.195482, "relative": 0.921236 }, { "name": "imu_accel_gyro", "l2": 17.064878, "mean_abs": 0.996634, "relative": 0.920709 }, { "name": "depth_confidence", "l2": 10.34594, "mean_abs": 0.191438, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.108083, "mean_abs": 0.105162, "relative": 0.924965 }, { "name": "video_fisheye_cam1", "l2": 5.328076, "mean_abs": 0.112195, "relative": 0.88042 }, { "name": "video_fisheye_cam2", "l2": 5.600244, "mean_abs": 0.113017, "relative": 0.893996 }, { "name": "video_fisheye_cam3", "l2": 5.55344, "mean_abs": 0.112363, "relative": 0.929857 }, { "name": "video_stereo_left", "l2": 5.405731, "mean_abs": 0.115303, "relative": 0.878442 }, { "name": "video_stereo_right", "l2": 6.374979, "mean_abs": 0.13874, "relative": 0.898738 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.106468, "mean_abs": 0.081751, "relative": 0.753873 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 412, "start_frame": 2060, "end_frame": 2079, "center_frame": 2069, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.430908, "mean_abs": 0.221104, "relative": 0.879892 }, { "name": "hand_right_joints", "l2": 10.987245, "mean_abs": 0.273477, "relative": 0.866089 }, { "name": "body_joints", "l2": 5.854965, "mean_abs": 0.078753, "relative": 0.790825 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010723, "mean_abs": 0.001502, "relative": 0.025304 }, { "name": "camera_rotation_matrix", "l2": 3.000023, "mean_abs": 0.194548, "relative": 0.921259 }, { "name": "imu_accel_gyro", "l2": 17.061825, "mean_abs": 1.000247, "relative": 0.920544 }, { "name": "depth_confidence", "l2": 10.330457, "mean_abs": 0.190623, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.107935, "mean_abs": 0.105114, "relative": 0.924938 }, { "name": "video_fisheye_cam1", "l2": 5.332149, "mean_abs": 0.112046, "relative": 0.881093 }, { "name": "video_fisheye_cam2", "l2": 5.595802, "mean_abs": 0.112698, "relative": 0.893287 }, { "name": "video_fisheye_cam3", "l2": 5.550721, "mean_abs": 0.111884, "relative": 0.929401 }, { "name": "video_stereo_left", "l2": 5.400745, "mean_abs": 0.115214, "relative": 0.877632 }, { "name": "video_stereo_right", "l2": 6.37631, "mean_abs": 0.138431, "relative": 0.898926 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.111345, "mean_abs": 0.080048, "relative": 0.755619 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 413, "start_frame": 2065, "end_frame": 2084, "center_frame": 2074, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.431915, "mean_abs": 0.22067, "relative": 0.879977 }, { "name": "hand_right_joints", "l2": 11.000841, "mean_abs": 0.272419, "relative": 0.867161 }, { "name": "body_joints", "l2": 5.847573, "mean_abs": 0.078757, "relative": 0.789827 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011797, "mean_abs": 0.001704, "relative": 0.02784 }, { "name": "camera_rotation_matrix", "l2": 2.999974, "mean_abs": 0.194305, "relative": 0.921243 }, { "name": "imu_accel_gyro", "l2": 17.06391, "mean_abs": 0.996605, "relative": 0.920657 }, { "name": "depth_confidence", "l2": 10.319864, "mean_abs": 0.190334, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.108685, "mean_abs": 0.105142, "relative": 0.925074 }, { "name": "video_fisheye_cam1", "l2": 5.338312, "mean_abs": 0.112078, "relative": 0.882111 }, { "name": "video_fisheye_cam2", "l2": 5.596525, "mean_abs": 0.112643, "relative": 0.893402 }, { "name": "video_fisheye_cam3", "l2": 5.552978, "mean_abs": 0.112068, "relative": 0.929779 }, { "name": "video_stereo_left", "l2": 5.406237, "mean_abs": 0.115129, "relative": 0.878524 }, { "name": "video_stereo_right", "l2": 6.384159, "mean_abs": 0.138412, "relative": 0.900032 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.091702, "mean_abs": 0.079635, "relative": 0.748589 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 414, "start_frame": 2070, "end_frame": 2089, "center_frame": 2079, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.434385, "mean_abs": 0.22072, "relative": 0.880185 }, { "name": "hand_right_joints", "l2": 10.999738, "mean_abs": 0.272995, "relative": 0.867074 }, { "name": "body_joints", "l2": 5.839369, "mean_abs": 0.078527, "relative": 0.788719 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008712, "mean_abs": 0.001243, "relative": 0.02056 }, { "name": "camera_rotation_matrix", "l2": 2.999973, "mean_abs": 0.19443, "relative": 0.921243 }, { "name": "imu_accel_gyro", "l2": 17.063154, "mean_abs": 0.999526, "relative": 0.920616 }, { "name": "depth_confidence", "l2": 10.307699, "mean_abs": 0.189916, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.106819, "mean_abs": 0.105089, "relative": 0.924736 }, { "name": "video_fisheye_cam1", "l2": 5.344366, "mean_abs": 0.112255, "relative": 0.883111 }, { "name": "video_fisheye_cam2", "l2": 5.595481, "mean_abs": 0.112684, "relative": 0.893235 }, { "name": "video_fisheye_cam3", "l2": 5.543338, "mean_abs": 0.111832, "relative": 0.928165 }, { "name": "video_stereo_left", "l2": 5.408162, "mean_abs": 0.11511, "relative": 0.878837 }, { "name": "video_stereo_right", "l2": 6.391135, "mean_abs": 0.138496, "relative": 0.901016 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.105954, "mean_abs": 0.078229, "relative": 0.753689 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 415, "start_frame": 2075, "end_frame": 2094, "center_frame": 2084, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.464971, "mean_abs": 0.220592, "relative": 0.882765 }, { "name": "hand_right_joints", "l2": 10.974872, "mean_abs": 0.273413, "relative": 0.865113 }, { "name": "body_joints", "l2": 5.816472, "mean_abs": 0.078976, "relative": 0.785626 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008678, "mean_abs": 0.001264, "relative": 0.020479 }, { "name": "camera_rotation_matrix", "l2": 3.000001, "mean_abs": 0.194503, "relative": 0.921252 }, { "name": "imu_accel_gyro", "l2": 17.072941, "mean_abs": 1.000653, "relative": 0.921144 }, { "name": "depth_confidence", "l2": 10.322747, "mean_abs": 0.190225, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.110126, "mean_abs": 0.105219, "relative": 0.925335 }, { "name": "video_fisheye_cam1", "l2": 5.35111, "mean_abs": 0.112312, "relative": 0.884226 }, { "name": "video_fisheye_cam2", "l2": 5.595315, "mean_abs": 0.112669, "relative": 0.893209 }, { "name": "video_fisheye_cam3", "l2": 5.552793, "mean_abs": 0.112113, "relative": 0.929748 }, { "name": "video_stereo_left", "l2": 5.407919, "mean_abs": 0.115146, "relative": 0.878798 }, { "name": "video_stereo_right", "l2": 6.397799, "mean_abs": 0.13858, "relative": 0.901955 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.102246, "mean_abs": 0.080558, "relative": 0.752362 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 416, "start_frame": 2080, "end_frame": 2099, "center_frame": 2089, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.492774, "mean_abs": 0.221206, "relative": 0.88511 }, { "name": "hand_right_joints", "l2": 10.977424, "mean_abs": 0.272531, "relative": 0.865315 }, { "name": "body_joints", "l2": 5.804064, "mean_abs": 0.078291, "relative": 0.78395 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005479, "mean_abs": 0.000768, "relative": 0.01293 }, { "name": "camera_rotation_matrix", "l2": 3.000005, "mean_abs": 0.194511, "relative": 0.921253 }, { "name": "imu_accel_gyro", "l2": 17.062107, "mean_abs": 0.99633, "relative": 0.920559 }, { "name": "depth_confidence", "l2": 10.322303, "mean_abs": 0.190278, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.110317, "mean_abs": 0.105163, "relative": 0.925369 }, { "name": "video_fisheye_cam1", "l2": 5.353119, "mean_abs": 0.112231, "relative": 0.884558 }, { "name": "video_fisheye_cam2", "l2": 5.594338, "mean_abs": 0.112641, "relative": 0.893053 }, { "name": "video_fisheye_cam3", "l2": 5.549108, "mean_abs": 0.111936, "relative": 0.929131 }, { "name": "video_stereo_left", "l2": 5.407563, "mean_abs": 0.115171, "relative": 0.87874 }, { "name": "video_stereo_right", "l2": 6.400412, "mean_abs": 0.138523, "relative": 0.902324 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.042055, "mean_abs": 0.078466, "relative": 0.730821 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 417, "start_frame": 2085, "end_frame": 2104, "center_frame": 2094, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.496531, "mean_abs": 0.221074, "relative": 0.885427 }, { "name": "hand_right_joints", "l2": 10.991945, "mean_abs": 0.273322, "relative": 0.866459 }, { "name": "body_joints", "l2": 5.804674, "mean_abs": 0.078421, "relative": 0.784033 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003057, "mean_abs": 0.000521, "relative": 0.007214 }, { "name": "camera_rotation_matrix", "l2": 3.000024, "mean_abs": 0.19446, "relative": 0.921259 }, { "name": "imu_accel_gyro", "l2": 17.06362, "mean_abs": 0.998456, "relative": 0.920641 }, { "name": "depth_confidence", "l2": 10.321741, "mean_abs": 0.190107, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.112588, "mean_abs": 0.105262, "relative": 0.92578 }, { "name": "video_fisheye_cam1", "l2": 5.35708, "mean_abs": 0.112366, "relative": 0.885212 }, { "name": "video_fisheye_cam2", "l2": 5.593668, "mean_abs": 0.112674, "relative": 0.892946 }, { "name": "video_fisheye_cam3", "l2": 5.553197, "mean_abs": 0.112107, "relative": 0.929816 }, { "name": "video_stereo_left", "l2": 5.408176, "mean_abs": 0.115306, "relative": 0.878839 }, { "name": "video_stereo_right", "l2": 6.408848, "mean_abs": 0.138798, "relative": 0.903513 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.030753, "mean_abs": 0.076374, "relative": 0.726776 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 418, "start_frame": 2090, "end_frame": 2109, "center_frame": 2099, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.517687, "mean_abs": 0.221674, "relative": 0.887212 }, { "name": "hand_right_joints", "l2": 11.014563, "mean_abs": 0.274087, "relative": 0.868242 }, { "name": "body_joints", "l2": 5.804368, "mean_abs": 0.07824, "relative": 0.783991 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005441, "mean_abs": 0.000864, "relative": 0.01284 }, { "name": "camera_rotation_matrix", "l2": 2.999997, "mean_abs": 0.194624, "relative": 0.92125 }, { "name": "imu_accel_gyro", "l2": 17.078176, "mean_abs": 1.001048, "relative": 0.921426 }, { "name": "depth_confidence", "l2": 10.310832, "mean_abs": 0.190094, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.114907, "mean_abs": 0.105349, "relative": 0.9262 }, { "name": "video_fisheye_cam1", "l2": 5.358568, "mean_abs": 0.112514, "relative": 0.885458 }, { "name": "video_fisheye_cam2", "l2": 5.59408, "mean_abs": 0.112835, "relative": 0.893012 }, { "name": "video_fisheye_cam3", "l2": 5.566751, "mean_abs": 0.112622, "relative": 0.932085 }, { "name": "video_stereo_left", "l2": 5.413825, "mean_abs": 0.115602, "relative": 0.879757 }, { "name": "video_stereo_right", "l2": 6.417074, "mean_abs": 0.139258, "relative": 0.904673 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.009787, "mean_abs": 0.076338, "relative": 0.719273 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 419, "start_frame": 2095, "end_frame": 2114, "center_frame": 2104, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.534948, "mean_abs": 0.220708, "relative": 0.888668 }, { "name": "hand_right_joints", "l2": 11.049918, "mean_abs": 0.276586, "relative": 0.871029 }, { "name": "body_joints", "l2": 5.808629, "mean_abs": 0.078001, "relative": 0.784567 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00622, "mean_abs": 0.000955, "relative": 0.014678 }, { "name": "camera_rotation_matrix", "l2": 3.000006, "mean_abs": 0.194597, "relative": 0.921253 }, { "name": "imu_accel_gyro", "l2": 17.037746, "mean_abs": 0.999746, "relative": 0.919245 }, { "name": "depth_confidence", "l2": 10.301505, "mean_abs": 0.190639, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.116051, "mean_abs": 0.105476, "relative": 0.926407 }, { "name": "video_fisheye_cam1", "l2": 5.358498, "mean_abs": 0.112647, "relative": 0.885446 }, { "name": "video_fisheye_cam2", "l2": 5.594261, "mean_abs": 0.112911, "relative": 0.893041 }, { "name": "video_fisheye_cam3", "l2": 5.577411, "mean_abs": 0.113261, "relative": 0.93387 }, { "name": "video_stereo_left", "l2": 5.420056, "mean_abs": 0.115911, "relative": 0.88077 }, { "name": "video_stereo_right", "l2": 6.422999, "mean_abs": 0.13975, "relative": 0.905508 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.026966, "mean_abs": 0.076946, "relative": 0.725421 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 420, "start_frame": 2100, "end_frame": 2119, "center_frame": 2109, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.535036, "mean_abs": 0.220601, "relative": 0.888675 }, { "name": "hand_right_joints", "l2": 11.125361, "mean_abs": 0.275802, "relative": 0.876976 }, { "name": "body_joints", "l2": 5.828988, "mean_abs": 0.077206, "relative": 0.787317 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005894, "mean_abs": 0.000883, "relative": 0.01391 }, { "name": "camera_rotation_matrix", "l2": 3.000035, "mean_abs": 0.194493, "relative": 0.921262 }, { "name": "imu_accel_gyro", "l2": 17.061255, "mean_abs": 1.0021, "relative": 0.920513 }, { "name": "depth_confidence", "l2": 10.294348, "mean_abs": 0.190119, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.120316, "mean_abs": 0.10553, "relative": 0.92718 }, { "name": "video_fisheye_cam1", "l2": 5.361152, "mean_abs": 0.112593, "relative": 0.885885 }, { "name": "video_fisheye_cam2", "l2": 5.595852, "mean_abs": 0.112947, "relative": 0.893295 }, { "name": "video_fisheye_cam3", "l2": 5.590883, "mean_abs": 0.113283, "relative": 0.936126 }, { "name": "video_stereo_left", "l2": 5.429003, "mean_abs": 0.115991, "relative": 0.882224 }, { "name": "video_stereo_right", "l2": 6.435143, "mean_abs": 0.139812, "relative": 0.90722 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.081944, "mean_abs": 0.080646, "relative": 0.745097 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 421, "start_frame": 2105, "end_frame": 2124, "center_frame": 2114, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.535646, "mean_abs": 0.221097, "relative": 0.888727 }, { "name": "hand_right_joints", "l2": 11.094364, "mean_abs": 0.276367, "relative": 0.874533 }, { "name": "body_joints", "l2": 5.837206, "mean_abs": 0.077786, "relative": 0.788427 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005756, "mean_abs": 0.000871, "relative": 0.013583 }, { "name": "camera_rotation_matrix", "l2": 3.00013, "mean_abs": 0.194563, "relative": 0.921291 }, { "name": "imu_accel_gyro", "l2": 17.060272, "mean_abs": 0.997787, "relative": 0.92046 }, { "name": "depth_confidence", "l2": 10.270262, "mean_abs": 0.189692, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.12312, "mean_abs": 0.105798, "relative": 0.927688 }, { "name": "video_fisheye_cam1", "l2": 5.356025, "mean_abs": 0.112746, "relative": 0.885038 }, { "name": "video_fisheye_cam2", "l2": 5.5925, "mean_abs": 0.113024, "relative": 0.892759 }, { "name": "video_fisheye_cam3", "l2": 5.600291, "mean_abs": 0.114007, "relative": 0.937701 }, { "name": "video_stereo_left", "l2": 5.433, "mean_abs": 0.116378, "relative": 0.882873 }, { "name": "video_stereo_right", "l2": 6.436324, "mean_abs": 0.140266, "relative": 0.907387 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.050677, "mean_abs": 0.080499, "relative": 0.733906 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 422, "start_frame": 2110, "end_frame": 2129, "center_frame": 2119, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.521047, "mean_abs": 0.220862, "relative": 0.887495 }, { "name": "hand_right_joints", "l2": 11.046843, "mean_abs": 0.275367, "relative": 0.870787 }, { "name": "body_joints", "l2": 5.840417, "mean_abs": 0.078421, "relative": 0.78886 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005545, "mean_abs": 0.000825, "relative": 0.013085 }, { "name": "camera_rotation_matrix", "l2": 2.99995, "mean_abs": 0.194185, "relative": 0.921236 }, { "name": "imu_accel_gyro", "l2": 17.074903, "mean_abs": 1.002645, "relative": 0.92125 }, { "name": "depth_confidence", "l2": 10.247671, "mean_abs": 0.189304, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.125819, "mean_abs": 0.106007, "relative": 0.928176 }, { "name": "video_fisheye_cam1", "l2": 5.35835, "mean_abs": 0.113021, "relative": 0.885422 }, { "name": "video_fisheye_cam2", "l2": 5.592124, "mean_abs": 0.113235, "relative": 0.892699 }, { "name": "video_fisheye_cam3", "l2": 5.609985, "mean_abs": 0.114554, "relative": 0.939324 }, { "name": "video_stereo_left", "l2": 5.443335, "mean_abs": 0.116682, "relative": 0.884553 }, { "name": "video_stereo_right", "l2": 6.448041, "mean_abs": 0.140879, "relative": 0.909038 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.107694, "mean_abs": 0.081184, "relative": 0.754312 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 423, "start_frame": 2115, "end_frame": 2134, "center_frame": 2124, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.518246, "mean_abs": 0.220984, "relative": 0.887259 }, { "name": "hand_right_joints", "l2": 11.027316, "mean_abs": 0.274547, "relative": 0.869247 }, { "name": "body_joints", "l2": 5.840046, "mean_abs": 0.078424, "relative": 0.78881 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006585, "mean_abs": 0.000971, "relative": 0.015539 }, { "name": "camera_rotation_matrix", "l2": 2.999952, "mean_abs": 0.193847, "relative": 0.921237 }, { "name": "imu_accel_gyro", "l2": 17.048708, "mean_abs": 1.000726, "relative": 0.919837 }, { "name": "depth_confidence", "l2": 10.200892, "mean_abs": 0.188625, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.132096, "mean_abs": 0.105814, "relative": 0.929313 }, { "name": "video_fisheye_cam1", "l2": 5.358806, "mean_abs": 0.112758, "relative": 0.885497 }, { "name": "video_fisheye_cam2", "l2": 5.596242, "mean_abs": 0.113455, "relative": 0.893357 }, { "name": "video_fisheye_cam3", "l2": 5.619312, "mean_abs": 0.113538, "relative": 0.940886 }, { "name": "video_stereo_left", "l2": 5.444271, "mean_abs": 0.116581, "relative": 0.884705 }, { "name": "video_stereo_right", "l2": 6.451442, "mean_abs": 0.140725, "relative": 0.909518 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.084001, "mean_abs": 0.081024, "relative": 0.745833 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 424, "start_frame": 2120, "end_frame": 2139, "center_frame": 2129, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.516047, "mean_abs": 0.22125, "relative": 0.887073 }, { "name": "hand_right_joints", "l2": 11.026545, "mean_abs": 0.274328, "relative": 0.869187 }, { "name": "body_joints", "l2": 5.837641, "mean_abs": 0.078542, "relative": 0.788485 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005009, "mean_abs": 0.000797, "relative": 0.011821 }, { "name": "camera_rotation_matrix", "l2": 3.000032, "mean_abs": 0.193848, "relative": 0.921261 }, { "name": "imu_accel_gyro", "l2": 17.046902, "mean_abs": 1.000947, "relative": 0.919739 }, { "name": "depth_confidence", "l2": 10.175201, "mean_abs": 0.188488, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.135486, "mean_abs": 0.106211, "relative": 0.929927 }, { "name": "video_fisheye_cam1", "l2": 5.368302, "mean_abs": 0.113623, "relative": 0.887067 }, { "name": "video_fisheye_cam2", "l2": 5.604827, "mean_abs": 0.114152, "relative": 0.894727 }, { "name": "video_fisheye_cam3", "l2": 5.623741, "mean_abs": 0.114853, "relative": 0.941628 }, { "name": "video_stereo_left", "l2": 5.442847, "mean_abs": 0.116813, "relative": 0.884473 }, { "name": "video_stereo_right", "l2": 6.443906, "mean_abs": 0.141069, "relative": 0.908456 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.041296, "mean_abs": 0.078158, "relative": 0.730549 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 425, "start_frame": 2125, "end_frame": 2144, "center_frame": 2134, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.514073, "mean_abs": 0.220871, "relative": 0.886907 }, { "name": "hand_right_joints", "l2": 11.017607, "mean_abs": 0.27445, "relative": 0.868482 }, { "name": "body_joints", "l2": 5.816648, "mean_abs": 0.078401, "relative": 0.78565 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003589, "mean_abs": 0.000591, "relative": 0.008469 }, { "name": "camera_rotation_matrix", "l2": 3.000012, "mean_abs": 0.194134, "relative": 0.921255 }, { "name": "imu_accel_gyro", "l2": 17.055109, "mean_abs": 1.004334, "relative": 0.920182 }, { "name": "depth_confidence", "l2": 10.168546, "mean_abs": 0.188242, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.142245, "mean_abs": 0.106026, "relative": 0.931151 }, { "name": "video_fisheye_cam1", "l2": 5.3921, "mean_abs": 0.114107, "relative": 0.890999 }, { "name": "video_fisheye_cam2", "l2": 5.616281, "mean_abs": 0.114397, "relative": 0.896556 }, { "name": "video_fisheye_cam3", "l2": 5.643368, "mean_abs": 0.11396, "relative": 0.944914 }, { "name": "video_stereo_left", "l2": 5.448352, "mean_abs": 0.116992, "relative": 0.885368 }, { "name": "video_stereo_right", "l2": 6.451794, "mean_abs": 0.140962, "relative": 0.909568 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.108156, "mean_abs": 0.080964, "relative": 0.754477 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 426, "start_frame": 2130, "end_frame": 2149, "center_frame": 2139, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.509036, "mean_abs": 0.221239, "relative": 0.886482 }, { "name": "hand_right_joints", "l2": 11.014963, "mean_abs": 0.274256, "relative": 0.868274 }, { "name": "body_joints", "l2": 5.805216, "mean_abs": 0.078349, "relative": 0.784106 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004521, "mean_abs": 0.000663, "relative": 0.010669 }, { "name": "camera_rotation_matrix", "l2": 3.000019, "mean_abs": 0.193967, "relative": 0.921257 }, { "name": "imu_accel_gyro", "l2": 17.066856, "mean_abs": 1.005931, "relative": 0.920816 }, { "name": "depth_confidence", "l2": 10.138795, "mean_abs": 0.188161, "relative": 0.000336 }, { "name": "video_fisheye_cam0", "l2": 5.150026, "mean_abs": 0.106319, "relative": 0.93256 }, { "name": "video_fisheye_cam1", "l2": 5.406178, "mean_abs": 0.114431, "relative": 0.893325 }, { "name": "video_fisheye_cam2", "l2": 5.622555, "mean_abs": 0.114526, "relative": 0.897557 }, { "name": "video_fisheye_cam3", "l2": 5.617773, "mean_abs": 0.113777, "relative": 0.940628 }, { "name": "video_stereo_left", "l2": 5.432718, "mean_abs": 0.116981, "relative": 0.882828 }, { "name": "video_stereo_right", "l2": 6.43219, "mean_abs": 0.140819, "relative": 0.906804 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.211582, "mean_abs": 0.088097, "relative": 0.791492 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 427, "start_frame": 2135, "end_frame": 2154, "center_frame": 2144, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.519883, "mean_abs": 0.221372, "relative": 0.887397 }, { "name": "hand_right_joints", "l2": 10.991648, "mean_abs": 0.274619, "relative": 0.866436 }, { "name": "body_joints", "l2": 5.789122, "mean_abs": 0.078828, "relative": 0.781932 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006837, "mean_abs": 0.000924, "relative": 0.016135 }, { "name": "camera_rotation_matrix", "l2": 3.000126, "mean_abs": 0.193783, "relative": 0.92129 }, { "name": "imu_accel_gyro", "l2": 17.072857, "mean_abs": 1.008989, "relative": 0.921139 }, { "name": "depth_confidence", "l2": 10.119648, "mean_abs": 0.187514, "relative": 0.000336 }, { "name": "video_fisheye_cam0", "l2": 5.156677, "mean_abs": 0.106294, "relative": 0.933764 }, { "name": "video_fisheye_cam1", "l2": 5.420039, "mean_abs": 0.114409, "relative": 0.895616 }, { "name": "video_fisheye_cam2", "l2": 5.630906, "mean_abs": 0.114525, "relative": 0.89889 }, { "name": "video_fisheye_cam3", "l2": 5.614112, "mean_abs": 0.113726, "relative": 0.940015 }, { "name": "video_stereo_left", "l2": 5.421706, "mean_abs": 0.116532, "relative": 0.881038 }, { "name": "video_stereo_right", "l2": 6.417497, "mean_abs": 0.140108, "relative": 0.904732 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.107849, "mean_abs": 0.08227, "relative": 0.754367 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 428, "start_frame": 2140, "end_frame": 2159, "center_frame": 2149, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.524144, "mean_abs": 0.221996, "relative": 0.887756 }, { "name": "hand_right_joints", "l2": 10.988072, "mean_abs": 0.273182, "relative": 0.866154 }, { "name": "body_joints", "l2": 5.773652, "mean_abs": 0.078252, "relative": 0.779842 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006522, "mean_abs": 0.000844, "relative": 0.015392 }, { "name": "camera_rotation_matrix", "l2": 3.000053, "mean_abs": 0.193441, "relative": 0.921268 }, { "name": "imu_accel_gyro", "l2": 17.077126, "mean_abs": 1.008497, "relative": 0.92137 }, { "name": "depth_confidence", "l2": 10.100741, "mean_abs": 0.186809, "relative": 0.000335 }, { "name": "video_fisheye_cam0", "l2": 5.162677, "mean_abs": 0.106349, "relative": 0.93485 }, { "name": "video_fisheye_cam1", "l2": 5.431513, "mean_abs": 0.114041, "relative": 0.897512 }, { "name": "video_fisheye_cam2", "l2": 5.637711, "mean_abs": 0.114246, "relative": 0.899977 }, { "name": "video_fisheye_cam3", "l2": 5.610967, "mean_abs": 0.113516, "relative": 0.939489 }, { "name": "video_stereo_left", "l2": 5.416752, "mean_abs": 0.116156, "relative": 0.880233 }, { "name": "video_stereo_right", "l2": 6.415754, "mean_abs": 0.139574, "relative": 0.904487 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.234923, "mean_abs": 0.092794, "relative": 0.799846 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 429, "start_frame": 2145, "end_frame": 2164, "center_frame": 2154, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.53469, "mean_abs": 0.221903, "relative": 0.888646 }, { "name": "hand_right_joints", "l2": 10.993453, "mean_abs": 0.273216, "relative": 0.866578 }, { "name": "body_joints", "l2": 5.759816, "mean_abs": 0.077931, "relative": 0.777974 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007279, "mean_abs": 0.001039, "relative": 0.017178 }, { "name": "camera_rotation_matrix", "l2": 3.000017, "mean_abs": 0.193305, "relative": 0.921257 }, { "name": "imu_accel_gyro", "l2": 17.077747, "mean_abs": 1.007612, "relative": 0.921403 }, { "name": "depth_confidence", "l2": 10.110821, "mean_abs": 0.186896, "relative": 0.000336 }, { "name": "video_fisheye_cam0", "l2": 5.164857, "mean_abs": 0.106344, "relative": 0.935245 }, { "name": "video_fisheye_cam1", "l2": 5.426298, "mean_abs": 0.11401, "relative": 0.89665 }, { "name": "video_fisheye_cam2", "l2": 5.636809, "mean_abs": 0.114189, "relative": 0.899833 }, { "name": "video_fisheye_cam3", "l2": 5.602349, "mean_abs": 0.113541, "relative": 0.938046 }, { "name": "video_stereo_left", "l2": 5.413725, "mean_abs": 0.115984, "relative": 0.879741 }, { "name": "video_stereo_right", "l2": 6.418977, "mean_abs": 0.139526, "relative": 0.904941 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.178139, "mean_abs": 0.086659, "relative": 0.779523 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 430, "start_frame": 2150, "end_frame": 2169, "center_frame": 2159, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.538647, "mean_abs": 0.222015, "relative": 0.88898 }, { "name": "hand_right_joints", "l2": 11.006091, "mean_abs": 0.272215, "relative": 0.867574 }, { "name": "body_joints", "l2": 5.753473, "mean_abs": 0.077119, "relative": 0.777117 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008678, "mean_abs": 0.001246, "relative": 0.02048 }, { "name": "camera_rotation_matrix", "l2": 2.999978, "mean_abs": 0.193405, "relative": 0.921245 }, { "name": "imu_accel_gyro", "l2": 17.071749, "mean_abs": 1.005711, "relative": 0.92108 }, { "name": "depth_confidence", "l2": 10.117861, "mean_abs": 0.187322, "relative": 0.000336 }, { "name": "video_fisheye_cam0", "l2": 5.167055, "mean_abs": 0.106404, "relative": 0.935643 }, { "name": "video_fisheye_cam1", "l2": 5.429211, "mean_abs": 0.11459, "relative": 0.897131 }, { "name": "video_fisheye_cam2", "l2": 5.628973, "mean_abs": 0.114658, "relative": 0.898582 }, { "name": "video_fisheye_cam3", "l2": 5.607081, "mean_abs": 0.112916, "relative": 0.938838 }, { "name": "video_stereo_left", "l2": 5.420298, "mean_abs": 0.11687, "relative": 0.880809 }, { "name": "video_stereo_right", "l2": 6.442313, "mean_abs": 0.140835, "relative": 0.908231 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.318788, "mean_abs": 0.095359, "relative": 0.829859 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 431, "start_frame": 2155, "end_frame": 2174, "center_frame": 2164, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.547387, "mean_abs": 0.222723, "relative": 0.889717 }, { "name": "hand_right_joints", "l2": 11.039062, "mean_abs": 0.273054, "relative": 0.870173 }, { "name": "body_joints", "l2": 5.746742, "mean_abs": 0.076765, "relative": 0.776208 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00686, "mean_abs": 0.001067, "relative": 0.016188 }, { "name": "camera_rotation_matrix", "l2": 3.00007, "mean_abs": 0.193104, "relative": 0.921273 }, { "name": "imu_accel_gyro", "l2": 17.066183, "mean_abs": 1.009588, "relative": 0.920779 }, { "name": "depth_confidence", "l2": 10.136486, "mean_abs": 0.188124, "relative": 0.000336 }, { "name": "video_fisheye_cam0", "l2": 5.165028, "mean_abs": 0.106513, "relative": 0.935276 }, { "name": "video_fisheye_cam1", "l2": 5.421324, "mean_abs": 0.114506, "relative": 0.895828 }, { "name": "video_fisheye_cam2", "l2": 5.61908, "mean_abs": 0.114521, "relative": 0.897002 }, { "name": "video_fisheye_cam3", "l2": 5.612417, "mean_abs": 0.113135, "relative": 0.939732 }, { "name": "video_stereo_left", "l2": 5.424571, "mean_abs": 0.117255, "relative": 0.881503 }, { "name": "video_stereo_right", "l2": 6.457105, "mean_abs": 0.141371, "relative": 0.910316 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.123609, "mean_abs": 0.0856, "relative": 0.760008 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 432, "start_frame": 2160, "end_frame": 2179, "center_frame": 2169, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.551558, "mean_abs": 0.222361, "relative": 0.890069 }, { "name": "hand_right_joints", "l2": 11.066266, "mean_abs": 0.272571, "relative": 0.872318 }, { "name": "body_joints", "l2": 5.742393, "mean_abs": 0.076348, "relative": 0.77562 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003768, "mean_abs": 0.000563, "relative": 0.008893 }, { "name": "camera_rotation_matrix", "l2": 3.00008, "mean_abs": 0.192835, "relative": 0.921276 }, { "name": "imu_accel_gyro", "l2": 17.052908, "mean_abs": 1.000606, "relative": 0.920063 }, { "name": "depth_confidence", "l2": 10.158546, "mean_abs": 0.188248, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.164359, "mean_abs": 0.106282, "relative": 0.935155 }, { "name": "video_fisheye_cam1", "l2": 5.407925, "mean_abs": 0.114122, "relative": 0.893614 }, { "name": "video_fisheye_cam2", "l2": 5.612327, "mean_abs": 0.114171, "relative": 0.895925 }, { "name": "video_fisheye_cam3", "l2": 5.619928, "mean_abs": 0.113158, "relative": 0.940989 }, { "name": "video_stereo_left", "l2": 5.431288, "mean_abs": 0.117351, "relative": 0.882595 }, { "name": "video_stereo_right", "l2": 6.466763, "mean_abs": 0.141491, "relative": 0.911678 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.100442, "mean_abs": 0.081105, "relative": 0.751717 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 433, "start_frame": 2165, "end_frame": 2184, "center_frame": 2174, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.545213, "mean_abs": 0.222549, "relative": 0.889534 }, { "name": "hand_right_joints", "l2": 11.023808, "mean_abs": 0.27227, "relative": 0.868971 }, { "name": "body_joints", "l2": 5.737171, "mean_abs": 0.076608, "relative": 0.774915 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003113, "mean_abs": 0.000524, "relative": 0.007346 }, { "name": "camera_rotation_matrix", "l2": 3.000026, "mean_abs": 0.192806, "relative": 0.921259 }, { "name": "imu_accel_gyro", "l2": 17.047169, "mean_abs": 1.001555, "relative": 0.919753 }, { "name": "depth_confidence", "l2": 10.169578, "mean_abs": 0.188426, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.161602, "mean_abs": 0.106196, "relative": 0.934656 }, { "name": "video_fisheye_cam1", "l2": 5.395349, "mean_abs": 0.113627, "relative": 0.891536 }, { "name": "video_fisheye_cam2", "l2": 5.600513, "mean_abs": 0.113766, "relative": 0.894038 }, { "name": "video_fisheye_cam3", "l2": 5.625125, "mean_abs": 0.113263, "relative": 0.941859 }, { "name": "video_stereo_left", "l2": 5.420897, "mean_abs": 0.116781, "relative": 0.880907 }, { "name": "video_stereo_right", "l2": 6.46848, "mean_abs": 0.141228, "relative": 0.91192 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.034628, "mean_abs": 0.077561, "relative": 0.728163 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 434, "start_frame": 2170, "end_frame": 2189, "center_frame": 2179, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.544491, "mean_abs": 0.223131, "relative": 0.889473 }, { "name": "hand_right_joints", "l2": 11.022393, "mean_abs": 0.271541, "relative": 0.868859 }, { "name": "body_joints", "l2": 5.728623, "mean_abs": 0.076882, "relative": 0.77376 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005593, "mean_abs": 0.000855, "relative": 0.013199 }, { "name": "camera_rotation_matrix", "l2": 3.000006, "mean_abs": 0.193203, "relative": 0.921253 }, { "name": "imu_accel_gyro", "l2": 17.048971, "mean_abs": 1.009568, "relative": 0.919851 }, { "name": "depth_confidence", "l2": 10.159406, "mean_abs": 0.188235, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.161393, "mean_abs": 0.10629, "relative": 0.934618 }, { "name": "video_fisheye_cam1", "l2": 5.402211, "mean_abs": 0.113791, "relative": 0.89267 }, { "name": "video_fisheye_cam2", "l2": 5.590402, "mean_abs": 0.113273, "relative": 0.892424 }, { "name": "video_fisheye_cam3", "l2": 5.630823, "mean_abs": 0.113652, "relative": 0.942813 }, { "name": "video_stereo_left", "l2": 5.413851, "mean_abs": 0.116438, "relative": 0.879762 }, { "name": "video_stereo_right", "l2": 6.476939, "mean_abs": 0.141207, "relative": 0.913113 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.98325, "mean_abs": 0.074565, "relative": 0.709775 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 435, "start_frame": 2175, "end_frame": 2194, "center_frame": 2184, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.542861, "mean_abs": 0.224048, "relative": 0.889335 }, { "name": "hand_right_joints", "l2": 11.00549, "mean_abs": 0.272611, "relative": 0.867527 }, { "name": "body_joints", "l2": 5.73523, "mean_abs": 0.077965, "relative": 0.774653 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006111, "mean_abs": 0.000831, "relative": 0.014421 }, { "name": "camera_rotation_matrix", "l2": 3.000011, "mean_abs": 0.193348, "relative": 0.921255 }, { "name": "imu_accel_gyro", "l2": 17.038841, "mean_abs": 1.008792, "relative": 0.919304 }, { "name": "depth_confidence", "l2": 10.153852, "mean_abs": 0.188308, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.164008, "mean_abs": 0.106201, "relative": 0.935091 }, { "name": "video_fisheye_cam1", "l2": 5.412919, "mean_abs": 0.11356, "relative": 0.894439 }, { "name": "video_fisheye_cam2", "l2": 5.598055, "mean_abs": 0.113133, "relative": 0.893646 }, { "name": "video_fisheye_cam3", "l2": 5.64189, "mean_abs": 0.113418, "relative": 0.944666 }, { "name": "video_stereo_left", "l2": 5.421049, "mean_abs": 0.116249, "relative": 0.880931 }, { "name": "video_stereo_right", "l2": 6.489842, "mean_abs": 0.140804, "relative": 0.914932 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.149471, "mean_abs": 0.084445, "relative": 0.769263 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 436, "start_frame": 2180, "end_frame": 2199, "center_frame": 2189, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.541581, "mean_abs": 0.225883, "relative": 0.889227 }, { "name": "hand_right_joints", "l2": 11.012778, "mean_abs": 0.27412, "relative": 0.868102 }, { "name": "body_joints", "l2": 5.747992, "mean_abs": 0.07921, "relative": 0.776377 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004322, "mean_abs": 0.00062, "relative": 0.0102 }, { "name": "camera_rotation_matrix", "l2": 3.000011, "mean_abs": 0.193682, "relative": 0.921255 }, { "name": "imu_accel_gyro", "l2": 17.054712, "mean_abs": 1.01029, "relative": 0.92016 }, { "name": "depth_confidence", "l2": 10.164109, "mean_abs": 0.188512, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.161978, "mean_abs": 0.106234, "relative": 0.934724 }, { "name": "video_fisheye_cam1", "l2": 5.405367, "mean_abs": 0.11379, "relative": 0.893191 }, { "name": "video_fisheye_cam2", "l2": 5.598177, "mean_abs": 0.113225, "relative": 0.893666 }, { "name": "video_fisheye_cam3", "l2": 5.632778, "mean_abs": 0.113374, "relative": 0.943141 }, { "name": "video_stereo_left", "l2": 5.429236, "mean_abs": 0.116794, "relative": 0.882262 }, { "name": "video_stereo_right", "l2": 6.500347, "mean_abs": 0.141699, "relative": 0.916412 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.004895, "mean_abs": 0.079197, "relative": 0.717522 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 437, "start_frame": 2185, "end_frame": 2204, "center_frame": 2194, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.530188, "mean_abs": 0.227607, "relative": 0.888266 }, { "name": "hand_right_joints", "l2": 10.987851, "mean_abs": 0.275123, "relative": 0.866137 }, { "name": "body_joints", "l2": 5.761793, "mean_abs": 0.080708, "relative": 0.778241 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004176, "mean_abs": 0.000632, "relative": 0.009855 }, { "name": "camera_rotation_matrix", "l2": 3.000146, "mean_abs": 0.193973, "relative": 0.921296 }, { "name": "imu_accel_gyro", "l2": 17.053036, "mean_abs": 1.005446, "relative": 0.92007 }, { "name": "depth_confidence", "l2": 10.158553, "mean_abs": 0.188727, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.160525, "mean_abs": 0.10627, "relative": 0.934461 }, { "name": "video_fisheye_cam1", "l2": 5.398141, "mean_abs": 0.113811, "relative": 0.891997 }, { "name": "video_fisheye_cam2", "l2": 5.593934, "mean_abs": 0.113292, "relative": 0.892988 }, { "name": "video_fisheye_cam3", "l2": 5.628516, "mean_abs": 0.113235, "relative": 0.942427 }, { "name": "video_stereo_left", "l2": 5.433445, "mean_abs": 0.116928, "relative": 0.882946 }, { "name": "video_stereo_right", "l2": 6.507809, "mean_abs": 0.142173, "relative": 0.917465 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.043397, "mean_abs": 0.081652, "relative": 0.731301 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 438, "start_frame": 2190, "end_frame": 2209, "center_frame": 2199, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.535432, "mean_abs": 0.22854, "relative": 0.888709 }, { "name": "hand_right_joints", "l2": 10.991756, "mean_abs": 0.276098, "relative": 0.866444 }, { "name": "body_joints", "l2": 5.762522, "mean_abs": 0.081291, "relative": 0.778339 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006853, "mean_abs": 0.000928, "relative": 0.016172 }, { "name": "camera_rotation_matrix", "l2": 3.000075, "mean_abs": 0.193854, "relative": 0.921274 }, { "name": "imu_accel_gyro", "l2": 17.049772, "mean_abs": 1.005593, "relative": 0.919894 }, { "name": "depth_confidence", "l2": 10.168439, "mean_abs": 0.188968, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.156592, "mean_abs": 0.10608, "relative": 0.933749 }, { "name": "video_fisheye_cam1", "l2": 5.384754, "mean_abs": 0.113432, "relative": 0.889785 }, { "name": "video_fisheye_cam2", "l2": 5.592102, "mean_abs": 0.113238, "relative": 0.892696 }, { "name": "video_fisheye_cam3", "l2": 5.622886, "mean_abs": 0.113184, "relative": 0.941485 }, { "name": "video_stereo_left", "l2": 5.441843, "mean_abs": 0.116933, "relative": 0.88431 }, { "name": "video_stereo_right", "l2": 6.51597, "mean_abs": 0.14227, "relative": 0.918615 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.13329, "mean_abs": 0.084069, "relative": 0.763473 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 439, "start_frame": 2195, "end_frame": 2214, "center_frame": 2204, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.51661, "mean_abs": 0.229313, "relative": 0.887121 }, { "name": "hand_right_joints", "l2": 10.97312, "mean_abs": 0.275554, "relative": 0.864975 }, { "name": "body_joints", "l2": 5.753945, "mean_abs": 0.08183, "relative": 0.777181 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008079, "mean_abs": 0.001264, "relative": 0.019066 }, { "name": "camera_rotation_matrix", "l2": 2.999985, "mean_abs": 0.193356, "relative": 0.921247 }, { "name": "imu_accel_gyro", "l2": 17.057152, "mean_abs": 1.013168, "relative": 0.920292 }, { "name": "depth_confidence", "l2": 10.177052, "mean_abs": 0.188549, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.152283, "mean_abs": 0.10602, "relative": 0.932968 }, { "name": "video_fisheye_cam1", "l2": 5.364044, "mean_abs": 0.11283, "relative": 0.886363 }, { "name": "video_fisheye_cam2", "l2": 5.587276, "mean_abs": 0.113067, "relative": 0.891926 }, { "name": "video_fisheye_cam3", "l2": 5.609526, "mean_abs": 0.112996, "relative": 0.939247 }, { "name": "video_stereo_left", "l2": 5.450151, "mean_abs": 0.116771, "relative": 0.88566 }, { "name": "video_stereo_right", "l2": 6.507554, "mean_abs": 0.141556, "relative": 0.917429 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.101385, "mean_abs": 0.084493, "relative": 0.752054 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 440, "start_frame": 2200, "end_frame": 2219, "center_frame": 2209, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.517617, "mean_abs": 0.231427, "relative": 0.887206 }, { "name": "hand_right_joints", "l2": 10.97751, "mean_abs": 0.276634, "relative": 0.865322 }, { "name": "body_joints", "l2": 5.7287, "mean_abs": 0.081675, "relative": 0.773771 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007404, "mean_abs": 0.001075, "relative": 0.017473 }, { "name": "camera_rotation_matrix", "l2": 3.000063, "mean_abs": 0.194241, "relative": 0.921271 }, { "name": "imu_accel_gyro", "l2": 17.069956, "mean_abs": 1.012464, "relative": 0.920983 }, { "name": "depth_confidence", "l2": 10.201067, "mean_abs": 0.189367, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.146761, "mean_abs": 0.105958, "relative": 0.931968 }, { "name": "video_fisheye_cam1", "l2": 5.344991, "mean_abs": 0.112706, "relative": 0.883215 }, { "name": "video_fisheye_cam2", "l2": 5.577905, "mean_abs": 0.112876, "relative": 0.890429 }, { "name": "video_fisheye_cam3", "l2": 5.593483, "mean_abs": 0.112955, "relative": 0.936561 }, { "name": "video_stereo_left", "l2": 5.431938, "mean_abs": 0.117117, "relative": 0.882701 }, { "name": "video_stereo_right", "l2": 6.484143, "mean_abs": 0.141578, "relative": 0.914128 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.13206, "mean_abs": 0.081339, "relative": 0.763032 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 441, "start_frame": 2205, "end_frame": 2224, "center_frame": 2214, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.529255, "mean_abs": 0.23259, "relative": 0.888188 }, { "name": "hand_right_joints", "l2": 10.997723, "mean_abs": 0.276047, "relative": 0.866915 }, { "name": "body_joints", "l2": 5.687776, "mean_abs": 0.081646, "relative": 0.768243 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00884, "mean_abs": 0.001449, "relative": 0.02086 }, { "name": "camera_rotation_matrix", "l2": 3.000431, "mean_abs": 0.194495, "relative": 0.921384 }, { "name": "imu_accel_gyro", "l2": 17.073402, "mean_abs": 1.015248, "relative": 0.921169 }, { "name": "depth_confidence", "l2": 10.201132, "mean_abs": 0.189436, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.145425, "mean_abs": 0.105979, "relative": 0.931726 }, { "name": "video_fisheye_cam1", "l2": 5.338122, "mean_abs": 0.112672, "relative": 0.88208 }, { "name": "video_fisheye_cam2", "l2": 5.570881, "mean_abs": 0.112814, "relative": 0.889308 }, { "name": "video_fisheye_cam3", "l2": 5.588521, "mean_abs": 0.112715, "relative": 0.93573 }, { "name": "video_stereo_left", "l2": 5.418619, "mean_abs": 0.117355, "relative": 0.880536 }, { "name": "video_stereo_right", "l2": 6.468845, "mean_abs": 0.142163, "relative": 0.911971 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.149602, "mean_abs": 0.08262, "relative": 0.76931 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 442, "start_frame": 2210, "end_frame": 2229, "center_frame": 2219, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.5429, "mean_abs": 0.233808, "relative": 0.889339 }, { "name": "hand_right_joints", "l2": 11.007119, "mean_abs": 0.276512, "relative": 0.867655 }, { "name": "body_joints", "l2": 5.612875, "mean_abs": 0.081571, "relative": 0.758126 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015902, "mean_abs": 0.00211, "relative": 0.037526 }, { "name": "camera_rotation_matrix", "l2": 3.000689, "mean_abs": 0.195024, "relative": 0.921463 }, { "name": "imu_accel_gyro", "l2": 17.115274, "mean_abs": 1.029289, "relative": 0.923428 }, { "name": "depth_confidence", "l2": 10.175913, "mean_abs": 0.189637, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.143094, "mean_abs": 0.106384, "relative": 0.931304 }, { "name": "video_fisheye_cam1", "l2": 5.327437, "mean_abs": 0.112999, "relative": 0.880314 }, { "name": "video_fisheye_cam2", "l2": 5.561642, "mean_abs": 0.112906, "relative": 0.887833 }, { "name": "video_fisheye_cam3", "l2": 5.565498, "mean_abs": 0.113763, "relative": 0.931876 }, { "name": "video_stereo_left", "l2": 5.40041, "mean_abs": 0.117853, "relative": 0.877577 }, { "name": "video_stereo_right", "l2": 6.428347, "mean_abs": 0.142647, "relative": 0.906262 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.139837, "mean_abs": 0.079721, "relative": 0.765816 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 443, "start_frame": 2215, "end_frame": 2234, "center_frame": 2224, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.576075, "mean_abs": 0.237245, "relative": 0.892137 }, { "name": "hand_right_joints", "l2": 11.122322, "mean_abs": 0.281141, "relative": 0.876737 }, { "name": "body_joints", "l2": 5.492717, "mean_abs": 0.08151, "relative": 0.741897 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.038623, "mean_abs": 0.004534, "relative": 0.091145 }, { "name": "camera_rotation_matrix", "l2": 3.001188, "mean_abs": 0.197257, "relative": 0.921616 }, { "name": "imu_accel_gyro", "l2": 17.11722, "mean_abs": 1.032842, "relative": 0.923533 }, { "name": "depth_confidence", "l2": 10.198635, "mean_abs": 0.191044, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.143808, "mean_abs": 0.106904, "relative": 0.931434 }, { "name": "video_fisheye_cam1", "l2": 5.325219, "mean_abs": 0.113625, "relative": 0.879947 }, { "name": "video_fisheye_cam2", "l2": 5.568652, "mean_abs": 0.113698, "relative": 0.888952 }, { "name": "video_fisheye_cam3", "l2": 5.517015, "mean_abs": 0.113932, "relative": 0.923758 }, { "name": "video_stereo_left", "l2": 5.382933, "mean_abs": 0.119088, "relative": 0.874737 }, { "name": "video_stereo_right", "l2": 6.388386, "mean_abs": 0.143359, "relative": 0.900628 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.190956, "mean_abs": 0.083561, "relative": 0.78411 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 444, "start_frame": 2220, "end_frame": 2239, "center_frame": 2229, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.657505, "mean_abs": 0.243631, "relative": 0.899006 }, { "name": "hand_right_joints", "l2": 11.300113, "mean_abs": 0.289049, "relative": 0.890751 }, { "name": "body_joints", "l2": 5.366076, "mean_abs": 0.080625, "relative": 0.724792 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.065212, "mean_abs": 0.008318, "relative": 0.153892 }, { "name": "camera_rotation_matrix", "l2": 3.00264, "mean_abs": 0.201515, "relative": 0.922062 }, { "name": "imu_accel_gyro", "l2": 17.121567, "mean_abs": 1.07851, "relative": 0.923768 }, { "name": "depth_confidence", "l2": 10.277238, "mean_abs": 0.194711, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.146088, "mean_abs": 0.107629, "relative": 0.931846 }, { "name": "video_fisheye_cam1", "l2": 5.325349, "mean_abs": 0.113954, "relative": 0.879969 }, { "name": "video_fisheye_cam2", "l2": 5.58879, "mean_abs": 0.114651, "relative": 0.892167 }, { "name": "video_fisheye_cam3", "l2": 5.469656, "mean_abs": 0.113999, "relative": 0.915828 }, { "name": "video_stereo_left", "l2": 5.395149, "mean_abs": 0.120499, "relative": 0.876722 }, { "name": "video_stereo_right", "l2": 6.380767, "mean_abs": 0.144913, "relative": 0.899554 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.237797, "mean_abs": 0.084639, "relative": 0.800874 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 445, "start_frame": 2225, "end_frame": 2244, "center_frame": 2234, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.68281, "mean_abs": 0.246947, "relative": 0.901141 }, { "name": "hand_right_joints", "l2": 11.389741, "mean_abs": 0.291183, "relative": 0.897816 }, { "name": "body_joints", "l2": 5.251408, "mean_abs": 0.078738, "relative": 0.709303 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.081524, "mean_abs": 0.011666, "relative": 0.192385 }, { "name": "camera_rotation_matrix", "l2": 3.005687, "mean_abs": 0.204089, "relative": 0.922998 }, { "name": "imu_accel_gyro", "l2": 17.197968, "mean_abs": 1.099243, "relative": 0.92789 }, { "name": "depth_confidence", "l2": 10.357133, "mean_abs": 0.197548, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.140106, "mean_abs": 0.107983, "relative": 0.930763 }, { "name": "video_fisheye_cam1", "l2": 5.331364, "mean_abs": 0.11428, "relative": 0.880963 }, { "name": "video_fisheye_cam2", "l2": 5.610802, "mean_abs": 0.115475, "relative": 0.895681 }, { "name": "video_fisheye_cam3", "l2": 5.424087, "mean_abs": 0.113793, "relative": 0.908198 }, { "name": "video_stereo_left", "l2": 5.409898, "mean_abs": 0.122024, "relative": 0.879119 }, { "name": "video_stereo_right", "l2": 6.389329, "mean_abs": 0.146822, "relative": 0.900761 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.190759, "mean_abs": 0.081277, "relative": 0.78404 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 446, "start_frame": 2230, "end_frame": 2249, "center_frame": 2239, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.704968, "mean_abs": 0.247926, "relative": 0.90301 }, { "name": "hand_right_joints", "l2": 11.707144, "mean_abs": 0.292339, "relative": 0.922836 }, { "name": "body_joints", "l2": 5.114331, "mean_abs": 0.076085, "relative": 0.690789 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.082244, "mean_abs": 0.012547, "relative": 0.194083 }, { "name": "camera_rotation_matrix", "l2": 3.00646, "mean_abs": 0.204299, "relative": 0.923235 }, { "name": "imu_accel_gyro", "l2": 17.147215, "mean_abs": 1.108243, "relative": 0.925151 }, { "name": "depth_confidence", "l2": 10.438921, "mean_abs": 0.198878, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.133545, "mean_abs": 0.107619, "relative": 0.929575 }, { "name": "video_fisheye_cam1", "l2": 5.331088, "mean_abs": 0.114191, "relative": 0.880917 }, { "name": "video_fisheye_cam2", "l2": 5.630456, "mean_abs": 0.11547, "relative": 0.898819 }, { "name": "video_fisheye_cam3", "l2": 5.362301, "mean_abs": 0.112021, "relative": 0.897853 }, { "name": "video_stereo_left", "l2": 5.402821, "mean_abs": 0.121729, "relative": 0.877969 }, { "name": "video_stereo_right", "l2": 6.373994, "mean_abs": 0.146514, "relative": 0.898599 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.142678, "mean_abs": 0.080754, "relative": 0.766832 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 447, "start_frame": 2235, "end_frame": 2254, "center_frame": 2244, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.75007, "mean_abs": 0.246571, "relative": 0.906814 }, { "name": "hand_right_joints", "l2": 11.953923, "mean_abs": 0.290556, "relative": 0.942289 }, { "name": "body_joints", "l2": 4.969508, "mean_abs": 0.072057, "relative": 0.671227 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.061197, "mean_abs": 0.009693, "relative": 0.144417 }, { "name": "camera_rotation_matrix", "l2": 3.002115, "mean_abs": 0.202526, "relative": 0.921901 }, { "name": "imu_accel_gyro", "l2": 17.095572, "mean_abs": 1.066175, "relative": 0.922365 }, { "name": "depth_confidence", "l2": 10.50734, "mean_abs": 0.197996, "relative": 0.000349 }, { "name": "video_fisheye_cam0", "l2": 5.128599, "mean_abs": 0.106898, "relative": 0.92868 }, { "name": "video_fisheye_cam1", "l2": 5.328645, "mean_abs": 0.113327, "relative": 0.880514 }, { "name": "video_fisheye_cam2", "l2": 5.651418, "mean_abs": 0.115037, "relative": 0.902165 }, { "name": "video_fisheye_cam3", "l2": 5.273419, "mean_abs": 0.108565, "relative": 0.88297 }, { "name": "video_stereo_left", "l2": 5.362159, "mean_abs": 0.119351, "relative": 0.871362 }, { "name": "video_stereo_right", "l2": 6.329485, "mean_abs": 0.143726, "relative": 0.892325 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.100057, "mean_abs": 0.081837, "relative": 0.751579 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 448, "start_frame": 2240, "end_frame": 2259, "center_frame": 2249, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.794416, "mean_abs": 0.244799, "relative": 0.910555 }, { "name": "hand_right_joints", "l2": 12.288875, "mean_abs": 0.298934, "relative": 0.968692 }, { "name": "body_joints", "l2": 4.92874, "mean_abs": 0.070859, "relative": 0.665721 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.038787, "mean_abs": 0.006099, "relative": 0.091531 }, { "name": "camera_rotation_matrix", "l2": 3.000853, "mean_abs": 0.202539, "relative": 0.921513 }, { "name": "imu_accel_gyro", "l2": 17.123919, "mean_abs": 1.048551, "relative": 0.923894 }, { "name": "depth_confidence", "l2": 10.600593, "mean_abs": 0.197689, "relative": 0.000352 }, { "name": "video_fisheye_cam0", "l2": 5.124725, "mean_abs": 0.106523, "relative": 0.927978 }, { "name": "video_fisheye_cam1", "l2": 5.334377, "mean_abs": 0.112768, "relative": 0.881461 }, { "name": "video_fisheye_cam2", "l2": 5.674754, "mean_abs": 0.114727, "relative": 0.90589 }, { "name": "video_fisheye_cam3", "l2": 5.290728, "mean_abs": 0.109388, "relative": 0.885869 }, { "name": "video_stereo_left", "l2": 5.348174, "mean_abs": 0.116865, "relative": 0.869089 }, { "name": "video_stereo_right", "l2": 6.329066, "mean_abs": 0.141345, "relative": 0.892265 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.073465, "mean_abs": 0.081704, "relative": 0.742062 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 449, "start_frame": 2245, "end_frame": 2264, "center_frame": 2254, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.772628, "mean_abs": 0.245729, "relative": 0.908717 }, { "name": "hand_right_joints", "l2": 12.399025, "mean_abs": 0.299339, "relative": 0.977375 }, { "name": "body_joints", "l2": 4.952298, "mean_abs": 0.071606, "relative": 0.668903 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01798, "mean_abs": 0.00284, "relative": 0.042431 }, { "name": "camera_rotation_matrix", "l2": 3.000099, "mean_abs": 0.202418, "relative": 0.921282 }, { "name": "imu_accel_gyro", "l2": 17.061529, "mean_abs": 1.023616, "relative": 0.920528 }, { "name": "depth_confidence", "l2": 10.648616, "mean_abs": 0.196451, "relative": 0.000353 }, { "name": "video_fisheye_cam0", "l2": 5.122136, "mean_abs": 0.106266, "relative": 0.927509 }, { "name": "video_fisheye_cam1", "l2": 5.346063, "mean_abs": 0.112331, "relative": 0.883392 }, { "name": "video_fisheye_cam2", "l2": 5.696739, "mean_abs": 0.114555, "relative": 0.9094 }, { "name": "video_fisheye_cam3", "l2": 5.3108, "mean_abs": 0.108661, "relative": 0.889229 }, { "name": "video_stereo_left", "l2": 5.353113, "mean_abs": 0.115636, "relative": 0.869891 }, { "name": "video_stereo_right", "l2": 6.340129, "mean_abs": 0.13958, "relative": 0.893825 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.990981, "mean_abs": 0.076792, "relative": 0.712542 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 450, "start_frame": 2250, "end_frame": 2269, "center_frame": 2259, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.761303, "mean_abs": 0.245579, "relative": 0.907762 }, { "name": "hand_right_joints", "l2": 12.527474, "mean_abs": 0.300995, "relative": 0.9875 }, { "name": "body_joints", "l2": 4.985557, "mean_abs": 0.072578, "relative": 0.673395 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014127, "mean_abs": 0.001964, "relative": 0.033338 }, { "name": "camera_rotation_matrix", "l2": 3.000318, "mean_abs": 0.203531, "relative": 0.921349 }, { "name": "imu_accel_gyro", "l2": 17.070541, "mean_abs": 1.02759, "relative": 0.921014 }, { "name": "depth_confidence", "l2": 10.688617, "mean_abs": 0.197091, "relative": 0.000355 }, { "name": "video_fisheye_cam0", "l2": 5.121196, "mean_abs": 0.106025, "relative": 0.927339 }, { "name": "video_fisheye_cam1", "l2": 5.351353, "mean_abs": 0.112348, "relative": 0.884266 }, { "name": "video_fisheye_cam2", "l2": 5.710952, "mean_abs": 0.114973, "relative": 0.911669 }, { "name": "video_fisheye_cam3", "l2": 5.362764, "mean_abs": 0.110879, "relative": 0.89793 }, { "name": "video_stereo_left", "l2": 5.352339, "mean_abs": 0.115284, "relative": 0.869766 }, { "name": "video_stereo_right", "l2": 6.348262, "mean_abs": 0.139386, "relative": 0.894972 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.998035, "mean_abs": 0.077672, "relative": 0.715067 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 451, "start_frame": 2255, "end_frame": 2274, "center_frame": 2264, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.738919, "mean_abs": 0.244424, "relative": 0.905874 }, { "name": "hand_right_joints", "l2": 12.686049, "mean_abs": 0.302701, "relative": 1.0 }, { "name": "body_joints", "l2": 5.023132, "mean_abs": 0.072241, "relative": 0.67847 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021027, "mean_abs": 0.002934, "relative": 0.049621 }, { "name": "camera_rotation_matrix", "l2": 3.000596, "mean_abs": 0.206272, "relative": 0.921434 }, { "name": "imu_accel_gyro", "l2": 17.069157, "mean_abs": 1.050406, "relative": 0.92094 }, { "name": "depth_confidence", "l2": 10.744237, "mean_abs": 0.198576, "relative": 0.000357 }, { "name": "video_fisheye_cam0", "l2": 5.118303, "mean_abs": 0.106396, "relative": 0.926815 }, { "name": "video_fisheye_cam1", "l2": 5.354671, "mean_abs": 0.112894, "relative": 0.884814 }, { "name": "video_fisheye_cam2", "l2": 5.72064, "mean_abs": 0.115607, "relative": 0.913215 }, { "name": "video_fisheye_cam3", "l2": 5.408415, "mean_abs": 0.110746, "relative": 0.905574 }, { "name": "video_stereo_left", "l2": 5.342467, "mean_abs": 0.116397, "relative": 0.868161 }, { "name": "video_stereo_right", "l2": 6.345677, "mean_abs": 0.140681, "relative": 0.894607 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.125108, "mean_abs": 0.08474, "relative": 0.760544 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 452, "start_frame": 2260, "end_frame": 2279, "center_frame": 2269, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.724934, "mean_abs": 0.242118, "relative": 0.904694 }, { "name": "hand_right_joints", "l2": 12.65114, "mean_abs": 0.299734, "relative": 0.997248 }, { "name": "body_joints", "l2": 5.03821, "mean_abs": 0.072113, "relative": 0.680507 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.026708, "mean_abs": 0.003978, "relative": 0.063027 }, { "name": "camera_rotation_matrix", "l2": 3.000962, "mean_abs": 0.207948, "relative": 0.921547 }, { "name": "imu_accel_gyro", "l2": 17.095835, "mean_abs": 1.067286, "relative": 0.922379 }, { "name": "depth_confidence", "l2": 10.795813, "mean_abs": 0.200451, "relative": 0.000358 }, { "name": "video_fisheye_cam0", "l2": 5.119357, "mean_abs": 0.10659, "relative": 0.927006 }, { "name": "video_fisheye_cam1", "l2": 5.363963, "mean_abs": 0.113441, "relative": 0.88635 }, { "name": "video_fisheye_cam2", "l2": 5.732228, "mean_abs": 0.116191, "relative": 0.915065 }, { "name": "video_fisheye_cam3", "l2": 5.43995, "mean_abs": 0.110657, "relative": 0.910854 }, { "name": "video_stereo_left", "l2": 5.342393, "mean_abs": 0.117068, "relative": 0.868149 }, { "name": "video_stereo_right", "l2": 6.348963, "mean_abs": 0.141498, "relative": 0.895071 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.135664, "mean_abs": 0.08517, "relative": 0.764322 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 453, "start_frame": 2265, "end_frame": 2284, "center_frame": 2274, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.71593, "mean_abs": 0.241853, "relative": 0.903934 }, { "name": "hand_right_joints", "l2": 12.563112, "mean_abs": 0.29679, "relative": 0.990309 }, { "name": "body_joints", "l2": 5.05143, "mean_abs": 0.072001, "relative": 0.682293 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028973, "mean_abs": 0.004273, "relative": 0.068373 }, { "name": "camera_rotation_matrix", "l2": 3.001081, "mean_abs": 0.207823, "relative": 0.921583 }, { "name": "imu_accel_gyro", "l2": 17.123524, "mean_abs": 1.068927, "relative": 0.923873 }, { "name": "depth_confidence", "l2": 10.825921, "mean_abs": 0.200108, "relative": 0.000359 }, { "name": "video_fisheye_cam0", "l2": 5.120748, "mean_abs": 0.106465, "relative": 0.927258 }, { "name": "video_fisheye_cam1", "l2": 5.376597, "mean_abs": 0.11362, "relative": 0.888437 }, { "name": "video_fisheye_cam2", "l2": 5.739697, "mean_abs": 0.116141, "relative": 0.916257 }, { "name": "video_fisheye_cam3", "l2": 5.462073, "mean_abs": 0.110673, "relative": 0.914558 }, { "name": "video_stereo_left", "l2": 5.33257, "mean_abs": 0.116344, "relative": 0.866553 }, { "name": "video_stereo_right", "l2": 6.353309, "mean_abs": 0.140643, "relative": 0.895683 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.109239, "mean_abs": 0.084013, "relative": 0.754865 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 454, "start_frame": 2270, "end_frame": 2289, "center_frame": 2279, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.705371, "mean_abs": 0.241878, "relative": 0.903044 }, { "name": "hand_right_joints", "l2": 12.627964, "mean_abs": 0.297131, "relative": 0.995421 }, { "name": "body_joints", "l2": 5.04192, "mean_abs": 0.071921, "relative": 0.681008 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.032538, "mean_abs": 0.004846, "relative": 0.076785 }, { "name": "camera_rotation_matrix", "l2": 3.001773, "mean_abs": 0.207916, "relative": 0.921796 }, { "name": "imu_accel_gyro", "l2": 17.148106, "mean_abs": 1.068806, "relative": 0.925199 }, { "name": "depth_confidence", "l2": 10.84489, "mean_abs": 0.199705, "relative": 0.00036 }, { "name": "video_fisheye_cam0", "l2": 5.116762, "mean_abs": 0.106286, "relative": 0.926536 }, { "name": "video_fisheye_cam1", "l2": 5.38902, "mean_abs": 0.113957, "relative": 0.89049 }, { "name": "video_fisheye_cam2", "l2": 5.741674, "mean_abs": 0.115991, "relative": 0.916573 }, { "name": "video_fisheye_cam3", "l2": 5.462286, "mean_abs": 0.110682, "relative": 0.914594 }, { "name": "video_stereo_left", "l2": 5.324528, "mean_abs": 0.115629, "relative": 0.865246 }, { "name": "video_stereo_right", "l2": 6.353486, "mean_abs": 0.140034, "relative": 0.895708 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.101246, "mean_abs": 0.082723, "relative": 0.752004 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 455, "start_frame": 2275, "end_frame": 2294, "center_frame": 2284, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.700451, "mean_abs": 0.240986, "relative": 0.902629 }, { "name": "hand_right_joints", "l2": 12.603601, "mean_abs": 0.297526, "relative": 0.993501 }, { "name": "body_joints", "l2": 5.066763, "mean_abs": 0.073151, "relative": 0.684364 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.052677, "mean_abs": 0.006925, "relative": 0.12431 }, { "name": "camera_rotation_matrix", "l2": 3.0012, "mean_abs": 0.207994, "relative": 0.92162 }, { "name": "imu_accel_gyro", "l2": 17.11644, "mean_abs": 1.059271, "relative": 0.923491 }, { "name": "depth_confidence", "l2": 10.879614, "mean_abs": 0.199793, "relative": 0.000361 }, { "name": "video_fisheye_cam0", "l2": 5.12375, "mean_abs": 0.106485, "relative": 0.927802 }, { "name": "video_fisheye_cam1", "l2": 5.417164, "mean_abs": 0.114526, "relative": 0.895141 }, { "name": "video_fisheye_cam2", "l2": 5.74076, "mean_abs": 0.115948, "relative": 0.916427 }, { "name": "video_fisheye_cam3", "l2": 5.475042, "mean_abs": 0.111474, "relative": 0.91673 }, { "name": "video_stereo_left", "l2": 5.313611, "mean_abs": 0.114985, "relative": 0.863472 }, { "name": "video_stereo_right", "l2": 6.367578, "mean_abs": 0.139971, "relative": 0.897695 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.118597, "mean_abs": 0.086356, "relative": 0.758214 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 456, "start_frame": 2280, "end_frame": 2299, "center_frame": 2289, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.627836, "mean_abs": 0.238906, "relative": 0.896503 }, { "name": "hand_right_joints", "l2": 12.539124, "mean_abs": 0.297549, "relative": 0.988418 }, { "name": "body_joints", "l2": 5.161076, "mean_abs": 0.07685, "relative": 0.697102 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.078863, "mean_abs": 0.010054, "relative": 0.186106 }, { "name": "camera_rotation_matrix", "l2": 3.002966, "mean_abs": 0.209712, "relative": 0.922162 }, { "name": "imu_accel_gyro", "l2": 17.180801, "mean_abs": 1.076792, "relative": 0.926963 }, { "name": "depth_confidence", "l2": 10.947008, "mean_abs": 0.202248, "relative": 0.000363 }, { "name": "video_fisheye_cam0", "l2": 5.128179, "mean_abs": 0.106853, "relative": 0.928604 }, { "name": "video_fisheye_cam1", "l2": 5.460778, "mean_abs": 0.116179, "relative": 0.902347 }, { "name": "video_fisheye_cam2", "l2": 5.734554, "mean_abs": 0.116368, "relative": 0.915436 }, { "name": "video_fisheye_cam3", "l2": 5.500988, "mean_abs": 0.112802, "relative": 0.921074 }, { "name": "video_stereo_left", "l2": 5.328994, "mean_abs": 0.116674, "relative": 0.865972 }, { "name": "video_stereo_right", "l2": 6.405853, "mean_abs": 0.142331, "relative": 0.903091 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.095269, "mean_abs": 0.08505, "relative": 0.749865 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 457, "start_frame": 2285, "end_frame": 2304, "center_frame": 2294, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.603639, "mean_abs": 0.237824, "relative": 0.894462 }, { "name": "hand_right_joints", "l2": 12.490442, "mean_abs": 0.299913, "relative": 0.984581 }, { "name": "body_joints", "l2": 5.29189, "mean_abs": 0.080802, "relative": 0.714771 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.106318, "mean_abs": 0.013321, "relative": 0.250897 }, { "name": "camera_rotation_matrix", "l2": 3.003678, "mean_abs": 0.210244, "relative": 0.922381 }, { "name": "imu_accel_gyro", "l2": 17.136297, "mean_abs": 1.063061, "relative": 0.924562 }, { "name": "depth_confidence", "l2": 10.976484, "mean_abs": 0.20372, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.13192, "mean_abs": 0.106936, "relative": 0.929281 }, { "name": "video_fisheye_cam1", "l2": 5.509985, "mean_abs": 0.117657, "relative": 0.910478 }, { "name": "video_fisheye_cam2", "l2": 5.733106, "mean_abs": 0.116883, "relative": 0.915205 }, { "name": "video_fisheye_cam3", "l2": 5.530354, "mean_abs": 0.113707, "relative": 0.925991 }, { "name": "video_stereo_left", "l2": 5.346278, "mean_abs": 0.117893, "relative": 0.868781 }, { "name": "video_stereo_right", "l2": 6.44865, "mean_abs": 0.14406, "relative": 0.909124 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.992417, "mean_abs": 0.077279, "relative": 0.713056 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 458, "start_frame": 2290, "end_frame": 2309, "center_frame": 2299, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.55514, "mean_abs": 0.238346, "relative": 0.890371 }, { "name": "hand_right_joints", "l2": 12.088984, "mean_abs": 0.30143, "relative": 0.952935 }, { "name": "body_joints", "l2": 5.41929, "mean_abs": 0.084237, "relative": 0.731979 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.129821, "mean_abs": 0.016036, "relative": 0.306359 }, { "name": "camera_rotation_matrix", "l2": 3.004005, "mean_abs": 0.21135, "relative": 0.922481 }, { "name": "imu_accel_gyro", "l2": 17.20224, "mean_abs": 1.085902, "relative": 0.92812 }, { "name": "depth_confidence", "l2": 11.063176, "mean_abs": 0.206666, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.136285, "mean_abs": 0.107055, "relative": 0.930071 }, { "name": "video_fisheye_cam1", "l2": 5.563761, "mean_abs": 0.119513, "relative": 0.919365 }, { "name": "video_fisheye_cam2", "l2": 5.736238, "mean_abs": 0.117582, "relative": 0.915705 }, { "name": "video_fisheye_cam3", "l2": 5.566218, "mean_abs": 0.114432, "relative": 0.931996 }, { "name": "video_stereo_left", "l2": 5.363761, "mean_abs": 0.119516, "relative": 0.871622 }, { "name": "video_stereo_right", "l2": 6.501018, "mean_abs": 0.146945, "relative": 0.916507 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.008104, "mean_abs": 0.077561, "relative": 0.71867 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 459, "start_frame": 2295, "end_frame": 2314, "center_frame": 2304, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.446375, "mean_abs": 0.233926, "relative": 0.881196 }, { "name": "hand_right_joints", "l2": 11.620603, "mean_abs": 0.287528, "relative": 0.916014 }, { "name": "body_joints", "l2": 5.534634, "mean_abs": 0.086063, "relative": 0.747558 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.139393, "mean_abs": 0.019211, "relative": 0.328949 }, { "name": "camera_rotation_matrix", "l2": 3.001632, "mean_abs": 0.211085, "relative": 0.921753 }, { "name": "imu_accel_gyro", "l2": 17.376623, "mean_abs": 1.114801, "relative": 0.937529 }, { "name": "depth_confidence", "l2": 11.178244, "mean_abs": 0.208824, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.124732, "mean_abs": 0.106742, "relative": 0.927979 }, { "name": "video_fisheye_cam1", "l2": 5.595996, "mean_abs": 0.119922, "relative": 0.924691 }, { "name": "video_fisheye_cam2", "l2": 5.74051, "mean_abs": 0.11822, "relative": 0.916387 }, { "name": "video_fisheye_cam3", "l2": 5.553425, "mean_abs": 0.114024, "relative": 0.929854 }, { "name": "video_stereo_left", "l2": 5.359516, "mean_abs": 0.120054, "relative": 0.870932 }, { "name": "video_stereo_right", "l2": 6.517994, "mean_abs": 0.148835, "relative": 0.9189 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.113368, "mean_abs": 0.08146, "relative": 0.756343 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 460, "start_frame": 2300, "end_frame": 2319, "center_frame": 2309, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.381567, "mean_abs": 0.232591, "relative": 0.875729 }, { "name": "hand_right_joints", "l2": 11.428572, "mean_abs": 0.290245, "relative": 0.900877 }, { "name": "body_joints", "l2": 5.632154, "mean_abs": 0.086707, "relative": 0.760731 }, { "name": "body_contacts", "l2": 3.463894, "mean_abs": 0.081689, "relative": 0.999565 }, { "name": "camera_translation", "l2": 0.164681, "mean_abs": 0.024705, "relative": 0.388624 }, { "name": "camera_rotation_matrix", "l2": 3.003138, "mean_abs": 0.215624, "relative": 0.922215 }, { "name": "imu_accel_gyro", "l2": 17.252407, "mean_abs": 1.042817, "relative": 0.930827 }, { "name": "depth_confidence", "l2": 11.319099, "mean_abs": 0.212065, "relative": 0.000376 }, { "name": "video_fisheye_cam0", "l2": 5.11132, "mean_abs": 0.10726, "relative": 0.925551 }, { "name": "video_fisheye_cam1", "l2": 5.645714, "mean_abs": 0.12124, "relative": 0.932907 }, { "name": "video_fisheye_cam2", "l2": 5.762644, "mean_abs": 0.119494, "relative": 0.91992 }, { "name": "video_fisheye_cam3", "l2": 5.542371, "mean_abs": 0.114231, "relative": 0.928003 }, { "name": "video_stereo_left", "l2": 5.383205, "mean_abs": 0.12141, "relative": 0.874782 }, { "name": "video_stereo_right", "l2": 6.5469, "mean_abs": 0.150936, "relative": 0.922976 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.197474, "mean_abs": 0.088367, "relative": 0.786443 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 461, "start_frame": 2305, "end_frame": 2324, "center_frame": 2314, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.297675, "mean_abs": 0.230852, "relative": 0.868653 }, { "name": "hand_right_joints", "l2": 11.260067, "mean_abs": 0.290561, "relative": 0.887595 }, { "name": "body_joints", "l2": 5.744421, "mean_abs": 0.086706, "relative": 0.775894 }, { "name": "body_contacts", "l2": 3.463894, "mean_abs": 0.081689, "relative": 0.999565 }, { "name": "camera_translation", "l2": 0.204552, "mean_abs": 0.027901, "relative": 0.482714 }, { "name": "camera_rotation_matrix", "l2": 3.00941, "mean_abs": 0.221563, "relative": 0.924141 }, { "name": "imu_accel_gyro", "l2": 17.28698, "mean_abs": 1.07899, "relative": 0.932692 }, { "name": "depth_confidence", "l2": 11.448525, "mean_abs": 0.216699, "relative": 0.00038 }, { "name": "video_fisheye_cam0", "l2": 5.090224, "mean_abs": 0.107905, "relative": 0.921731 }, { "name": "video_fisheye_cam1", "l2": 5.649466, "mean_abs": 0.122162, "relative": 0.933527 }, { "name": "video_fisheye_cam2", "l2": 5.782752, "mean_abs": 0.120554, "relative": 0.92313 }, { "name": "video_fisheye_cam3", "l2": 5.499051, "mean_abs": 0.114324, "relative": 0.92075 }, { "name": "video_stereo_left", "l2": 5.413778, "mean_abs": 0.123103, "relative": 0.87975 }, { "name": "video_stereo_right", "l2": 6.571808, "mean_abs": 0.152404, "relative": 0.926487 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.125767, "mean_abs": 0.085772, "relative": 0.76078 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 462, "start_frame": 2310, "end_frame": 2329, "center_frame": 2319, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.157389, "mean_abs": 0.230999, "relative": 0.856819 }, { "name": "hand_right_joints", "l2": 10.886772, "mean_abs": 0.28181, "relative": 0.858169 }, { "name": "body_joints", "l2": 5.797554, "mean_abs": 0.087416, "relative": 0.783071 }, { "name": "body_contacts", "l2": 3.463894, "mean_abs": 0.081689, "relative": 0.999565 }, { "name": "camera_translation", "l2": 0.23248, "mean_abs": 0.031767, "relative": 0.54862 }, { "name": "camera_rotation_matrix", "l2": 3.041528, "mean_abs": 0.235518, "relative": 0.934004 }, { "name": "imu_accel_gyro", "l2": 17.692743, "mean_abs": 1.169744, "relative": 0.954584 }, { "name": "depth_confidence", "l2": 11.957361, "mean_abs": 0.231945, "relative": 0.000397 }, { "name": "video_fisheye_cam0", "l2": 5.069864, "mean_abs": 0.109181, "relative": 0.918044 }, { "name": "video_fisheye_cam1", "l2": 5.765573, "mean_abs": 0.127549, "relative": 0.952712 }, { "name": "video_fisheye_cam2", "l2": 5.835785, "mean_abs": 0.124411, "relative": 0.931596 }, { "name": "video_fisheye_cam3", "l2": 5.454037, "mean_abs": 0.115232, "relative": 0.913213 }, { "name": "video_stereo_left", "l2": 5.485401, "mean_abs": 0.129664, "relative": 0.891389 }, { "name": "video_stereo_right", "l2": 6.596042, "mean_abs": 0.158194, "relative": 0.929904 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.261553, "mean_abs": 0.092805, "relative": 0.809376 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 463, "start_frame": 2315, "end_frame": 2334, "center_frame": 2324, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.05347, "mean_abs": 0.229324, "relative": 0.848053 }, { "name": "hand_right_joints", "l2": 10.582804, "mean_abs": 0.266527, "relative": 0.834208 }, { "name": "body_joints", "l2": 5.843808, "mean_abs": 0.089576, "relative": 0.789318 }, { "name": "body_contacts", "l2": 3.463894, "mean_abs": 0.081689, "relative": 0.999565 }, { "name": "camera_translation", "l2": 0.194812, "mean_abs": 0.029322, "relative": 0.459728 }, { "name": "camera_rotation_matrix", "l2": 3.051211, "mean_abs": 0.242964, "relative": 0.936977 }, { "name": "imu_accel_gyro", "l2": 17.631456, "mean_abs": 1.111436, "relative": 0.951278 }, { "name": "depth_confidence", "l2": 13.349605, "mean_abs": 0.244524, "relative": 0.000443 }, { "name": "video_fisheye_cam0", "l2": 5.042864, "mean_abs": 0.108821, "relative": 0.913155 }, { "name": "video_fisheye_cam1", "l2": 5.910051, "mean_abs": 0.132438, "relative": 0.976586 }, { "name": "video_fisheye_cam2", "l2": 5.858438, "mean_abs": 0.12547, "relative": 0.935213 }, { "name": "video_fisheye_cam3", "l2": 5.444006, "mean_abs": 0.115199, "relative": 0.911533 }, { "name": "video_stereo_left", "l2": 5.488392, "mean_abs": 0.131309, "relative": 0.891875 }, { "name": "video_stereo_right", "l2": 6.584796, "mean_abs": 0.159472, "relative": 0.928318 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.058845, "mean_abs": 0.083122, "relative": 0.73683 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 464, "start_frame": 2320, "end_frame": 2339, "center_frame": 2329, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.982935, "mean_abs": 0.225968, "relative": 0.842103 }, { "name": "hand_right_joints", "l2": 10.514292, "mean_abs": 0.262463, "relative": 0.828807 }, { "name": "body_joints", "l2": 5.898449, "mean_abs": 0.090855, "relative": 0.796699 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.150433, "mean_abs": 0.023348, "relative": 0.355002 }, { "name": "camera_rotation_matrix", "l2": 3.04748, "mean_abs": 0.24468, "relative": 0.935832 }, { "name": "imu_accel_gyro", "l2": 17.704893, "mean_abs": 1.149672, "relative": 0.95524 }, { "name": "depth_confidence", "l2": 14.113056, "mean_abs": 0.250233, "relative": 0.000468 }, { "name": "video_fisheye_cam0", "l2": 5.027515, "mean_abs": 0.108222, "relative": 0.910375 }, { "name": "video_fisheye_cam1", "l2": 5.945685, "mean_abs": 0.133634, "relative": 0.982474 }, { "name": "video_fisheye_cam2", "l2": 5.876033, "mean_abs": 0.125601, "relative": 0.938021 }, { "name": "video_fisheye_cam3", "l2": 5.404827, "mean_abs": 0.113932, "relative": 0.904973 }, { "name": "video_stereo_left", "l2": 5.491837, "mean_abs": 0.130801, "relative": 0.892434 }, { "name": "video_stereo_right", "l2": 6.519186, "mean_abs": 0.157009, "relative": 0.919068 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.031021, "mean_abs": 0.079822, "relative": 0.726872 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 465, "start_frame": 2325, "end_frame": 2344, "center_frame": 2334, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.876571, "mean_abs": 0.215073, "relative": 0.833131 }, { "name": "hand_right_joints", "l2": 10.440259, "mean_abs": 0.255552, "relative": 0.822972 }, { "name": "body_joints", "l2": 5.909753, "mean_abs": 0.090809, "relative": 0.798226 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.12535, "mean_abs": 0.018002, "relative": 0.295809 }, { "name": "camera_rotation_matrix", "l2": 3.023178, "mean_abs": 0.236762, "relative": 0.928369 }, { "name": "imu_accel_gyro", "l2": 17.82019, "mean_abs": 1.129805, "relative": 0.961461 }, { "name": "depth_confidence", "l2": 14.416693, "mean_abs": 0.248911, "relative": 0.000478 }, { "name": "video_fisheye_cam0", "l2": 5.010891, "mean_abs": 0.106496, "relative": 0.907365 }, { "name": "video_fisheye_cam1", "l2": 5.959817, "mean_abs": 0.132165, "relative": 0.984809 }, { "name": "video_fisheye_cam2", "l2": 5.86741, "mean_abs": 0.123508, "relative": 0.936645 }, { "name": "video_fisheye_cam3", "l2": 5.416226, "mean_abs": 0.113104, "relative": 0.906882 }, { "name": "video_stereo_left", "l2": 5.406831, "mean_abs": 0.126224, "relative": 0.878621 }, { "name": "video_stereo_right", "l2": 6.408026, "mean_abs": 0.151621, "relative": 0.903397 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.069259, "mean_abs": 0.080583, "relative": 0.740557 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 466, "start_frame": 2330, "end_frame": 2349, "center_frame": 2339, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.85121, "mean_abs": 0.215567, "relative": 0.830992 }, { "name": "hand_right_joints", "l2": 10.346772, "mean_abs": 0.254645, "relative": 0.815602 }, { "name": "body_joints", "l2": 5.927042, "mean_abs": 0.089715, "relative": 0.800561 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.089839, "mean_abs": 0.013576, "relative": 0.212007 }, { "name": "camera_rotation_matrix", "l2": 3.003969, "mean_abs": 0.22383, "relative": 0.92247 }, { "name": "imu_accel_gyro", "l2": 16.86207, "mean_abs": 0.952015, "relative": 0.909767 }, { "name": "depth_confidence", "l2": 14.330976, "mean_abs": 0.244077, "relative": 0.000476 }, { "name": "video_fisheye_cam0", "l2": 5.017923, "mean_abs": 0.105228, "relative": 0.908639 }, { "name": "video_fisheye_cam1", "l2": 5.93145, "mean_abs": 0.128698, "relative": 0.980122 }, { "name": "video_fisheye_cam2", "l2": 5.849147, "mean_abs": 0.121322, "relative": 0.933729 }, { "name": "video_fisheye_cam3", "l2": 5.51499, "mean_abs": 0.113016, "relative": 0.923419 }, { "name": "video_stereo_left", "l2": 5.356643, "mean_abs": 0.123742, "relative": 0.870465 }, { "name": "video_stereo_right", "l2": 6.271426, "mean_abs": 0.146218, "relative": 0.884139 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.04684, "mean_abs": 0.079798, "relative": 0.732533 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 467, "start_frame": 2335, "end_frame": 2354, "center_frame": 2344, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.876008, "mean_abs": 0.216022, "relative": 0.833083 }, { "name": "hand_right_joints", "l2": 10.345356, "mean_abs": 0.255228, "relative": 0.815491 }, { "name": "body_joints", "l2": 5.942265, "mean_abs": 0.089427, "relative": 0.802617 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.062072, "mean_abs": 0.009675, "relative": 0.146482 }, { "name": "camera_rotation_matrix", "l2": 3.008995, "mean_abs": 0.228885, "relative": 0.924013 }, { "name": "imu_accel_gyro", "l2": 16.964735, "mean_abs": 0.95545, "relative": 0.915306 }, { "name": "depth_confidence", "l2": 14.458439, "mean_abs": 0.247075, "relative": 0.00048 }, { "name": "video_fisheye_cam0", "l2": 5.023979, "mean_abs": 0.105763, "relative": 0.909735 }, { "name": "video_fisheye_cam1", "l2": 5.900597, "mean_abs": 0.129587, "relative": 0.975024 }, { "name": "video_fisheye_cam2", "l2": 5.865955, "mean_abs": 0.122277, "relative": 0.936413 }, { "name": "video_fisheye_cam3", "l2": 5.537177, "mean_abs": 0.113614, "relative": 0.927134 }, { "name": "video_stereo_left", "l2": 5.378505, "mean_abs": 0.125025, "relative": 0.874018 }, { "name": "video_stereo_right", "l2": 6.257415, "mean_abs": 0.146876, "relative": 0.882164 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.055998, "mean_abs": 0.078517, "relative": 0.735811 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 468, "start_frame": 2340, "end_frame": 2359, "center_frame": 2349, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.89553, "mean_abs": 0.216661, "relative": 0.83473 }, { "name": "hand_right_joints", "l2": 10.357005, "mean_abs": 0.254792, "relative": 0.816409 }, { "name": "body_joints", "l2": 5.925651, "mean_abs": 0.089146, "relative": 0.800373 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.055965, "mean_abs": 0.008237, "relative": 0.132068 }, { "name": "camera_rotation_matrix", "l2": 3.013485, "mean_abs": 0.230945, "relative": 0.925392 }, { "name": "imu_accel_gyro", "l2": 16.953781, "mean_abs": 0.957051, "relative": 0.914715 }, { "name": "depth_confidence", "l2": 14.211188, "mean_abs": 0.246115, "relative": 0.000472 }, { "name": "video_fisheye_cam0", "l2": 5.039428, "mean_abs": 0.106195, "relative": 0.912533 }, { "name": "video_fisheye_cam1", "l2": 5.845738, "mean_abs": 0.128505, "relative": 0.965959 }, { "name": "video_fisheye_cam2", "l2": 5.866996, "mean_abs": 0.122642, "relative": 0.936579 }, { "name": "video_fisheye_cam3", "l2": 5.569751, "mean_abs": 0.114245, "relative": 0.932588 }, { "name": "video_stereo_left", "l2": 5.408026, "mean_abs": 0.125735, "relative": 0.878815 }, { "name": "video_stereo_right", "l2": 6.254278, "mean_abs": 0.146917, "relative": 0.881722 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.072558, "mean_abs": 0.080357, "relative": 0.741737 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011846, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 469, "start_frame": 2345, "end_frame": 2364, "center_frame": 2354, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.964392, "mean_abs": 0.216497, "relative": 0.840539 }, { "name": "hand_right_joints", "l2": 10.388518, "mean_abs": 0.255994, "relative": 0.818893 }, { "name": "body_joints", "l2": 5.914549, "mean_abs": 0.088965, "relative": 0.798873 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.055741, "mean_abs": 0.008631, "relative": 0.131542 }, { "name": "camera_rotation_matrix", "l2": 3.009989, "mean_abs": 0.22858, "relative": 0.924319 }, { "name": "imu_accel_gyro", "l2": 17.019758, "mean_abs": 0.993459, "relative": 0.918275 }, { "name": "depth_confidence", "l2": 12.457342, "mean_abs": 0.233368, "relative": 0.000413 }, { "name": "video_fisheye_cam0", "l2": 5.057654, "mean_abs": 0.106363, "relative": 0.915833 }, { "name": "video_fisheye_cam1", "l2": 5.719158, "mean_abs": 0.124337, "relative": 0.945043 }, { "name": "video_fisheye_cam2", "l2": 5.851699, "mean_abs": 0.121817, "relative": 0.934137 }, { "name": "video_fisheye_cam3", "l2": 5.570923, "mean_abs": 0.113845, "relative": 0.932784 }, { "name": "video_stereo_left", "l2": 5.390944, "mean_abs": 0.123973, "relative": 0.876039 }, { "name": "video_stereo_right", "l2": 6.224514, "mean_abs": 0.143939, "relative": 0.877526 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.126343, "mean_abs": 0.08295, "relative": 0.760986 }, { "name": "caption_objects_interaction_text", "l2": 1.878715, "mean_abs": 0.019599, "relative": 0.86266 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 470, "start_frame": 2350, "end_frame": 2369, "center_frame": 2359, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.993517, "mean_abs": 0.215056, "relative": 0.842996 }, { "name": "hand_right_joints", "l2": 10.40142, "mean_abs": 0.253991, "relative": 0.81991 }, { "name": "body_joints", "l2": 5.874997, "mean_abs": 0.088123, "relative": 0.793531 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.042488, "mean_abs": 0.006686, "relative": 0.100266 }, { "name": "camera_rotation_matrix", "l2": 3.002756, "mean_abs": 0.225407, "relative": 0.922098 }, { "name": "imu_accel_gyro", "l2": 17.586178, "mean_abs": 0.94426, "relative": 0.948835 }, { "name": "depth_confidence", "l2": 12.349762, "mean_abs": 0.228966, "relative": 0.00041 }, { "name": "video_fisheye_cam0", "l2": 5.073868, "mean_abs": 0.106352, "relative": 0.918769 }, { "name": "video_fisheye_cam1", "l2": 5.604675, "mean_abs": 0.120098, "relative": 0.926125 }, { "name": "video_fisheye_cam2", "l2": 5.835304, "mean_abs": 0.121011, "relative": 0.931519 }, { "name": "video_fisheye_cam3", "l2": 5.562167, "mean_abs": 0.112842, "relative": 0.931318 }, { "name": "video_stereo_left", "l2": 5.36273, "mean_abs": 0.121218, "relative": 0.871454 }, { "name": "video_stereo_right", "l2": 6.185539, "mean_abs": 0.140805, "relative": 0.872031 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.063235, "mean_abs": 0.082755, "relative": 0.738401 }, { "name": "caption_objects_interaction_text", "l2": 1.878715, "mean_abs": 0.020082, "relative": 0.86266 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 471, "start_frame": 2355, "end_frame": 2374, "center_frame": 2364, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.997437, "mean_abs": 0.215952, "relative": 0.843326 }, { "name": "hand_right_joints", "l2": 10.406895, "mean_abs": 0.256675, "relative": 0.820342 }, { "name": "body_joints", "l2": 5.842124, "mean_abs": 0.08744, "relative": 0.789091 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.058635, "mean_abs": 0.009, "relative": 0.13837 }, { "name": "camera_rotation_matrix", "l2": 3.006586, "mean_abs": 0.226483, "relative": 0.923274 }, { "name": "imu_accel_gyro", "l2": 17.281485, "mean_abs": 0.945343, "relative": 0.932396 }, { "name": "depth_confidence", "l2": 12.256351, "mean_abs": 0.228436, "relative": 0.000407 }, { "name": "video_fisheye_cam0", "l2": 5.084503, "mean_abs": 0.106813, "relative": 0.920695 }, { "name": "video_fisheye_cam1", "l2": 5.587009, "mean_abs": 0.120411, "relative": 0.923206 }, { "name": "video_fisheye_cam2", "l2": 5.804613, "mean_abs": 0.120863, "relative": 0.92662 }, { "name": "video_fisheye_cam3", "l2": 5.588268, "mean_abs": 0.113682, "relative": 0.935688 }, { "name": "video_stereo_left", "l2": 5.376594, "mean_abs": 0.122128, "relative": 0.873707 }, { "name": "video_stereo_right", "l2": 6.224545, "mean_abs": 0.142421, "relative": 0.87753 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.053191, "mean_abs": 0.07986, "relative": 0.734806 }, { "name": "caption_objects_interaction_text", "l2": 1.878715, "mean_abs": 0.020043, "relative": 0.86266 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 472, "start_frame": 2360, "end_frame": 2379, "center_frame": 2369, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.005702, "mean_abs": 0.217539, "relative": 0.844024 }, { "name": "hand_right_joints", "l2": 10.431998, "mean_abs": 0.260304, "relative": 0.822321 }, { "name": "body_joints", "l2": 5.824533, "mean_abs": 0.088576, "relative": 0.786715 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.077732, "mean_abs": 0.01045, "relative": 0.183436 }, { "name": "camera_rotation_matrix", "l2": 3.007697, "mean_abs": 0.223836, "relative": 0.923615 }, { "name": "imu_accel_gyro", "l2": 17.177942, "mean_abs": 0.943382, "relative": 0.926809 }, { "name": "depth_confidence", "l2": 12.153178, "mean_abs": 0.225468, "relative": 0.000403 }, { "name": "video_fisheye_cam0", "l2": 5.104214, "mean_abs": 0.10722, "relative": 0.924264 }, { "name": "video_fisheye_cam1", "l2": 5.560204, "mean_abs": 0.119552, "relative": 0.918777 }, { "name": "video_fisheye_cam2", "l2": 5.768468, "mean_abs": 0.120018, "relative": 0.92085 }, { "name": "video_fisheye_cam3", "l2": 5.627968, "mean_abs": 0.114656, "relative": 0.942335 }, { "name": "video_stereo_left", "l2": 5.36154, "mean_abs": 0.121522, "relative": 0.871261 }, { "name": "video_stereo_right", "l2": 6.232406, "mean_abs": 0.142067, "relative": 0.878638 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.125463, "mean_abs": 0.082873, "relative": 0.760671 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014509, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 473, "start_frame": 2365, "end_frame": 2384, "center_frame": 2374, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.024622, "mean_abs": 0.218405, "relative": 0.84562 }, { "name": "hand_right_joints", "l2": 10.401098, "mean_abs": 0.261745, "relative": 0.819885 }, { "name": "body_joints", "l2": 5.781171, "mean_abs": 0.089889, "relative": 0.780858 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.115344, "mean_abs": 0.014524, "relative": 0.272196 }, { "name": "camera_rotation_matrix", "l2": 3.00747, "mean_abs": 0.223598, "relative": 0.923545 }, { "name": "imu_accel_gyro", "l2": 17.158909, "mean_abs": 0.965167, "relative": 0.925782 }, { "name": "depth_confidence", "l2": 12.128344, "mean_abs": 0.224801, "relative": 0.000403 }, { "name": "video_fisheye_cam0", "l2": 5.122443, "mean_abs": 0.107657, "relative": 0.927565 }, { "name": "video_fisheye_cam1", "l2": 5.538056, "mean_abs": 0.119343, "relative": 0.915117 }, { "name": "video_fisheye_cam2", "l2": 5.747118, "mean_abs": 0.119492, "relative": 0.917442 }, { "name": "video_fisheye_cam3", "l2": 5.636546, "mean_abs": 0.115367, "relative": 0.943772 }, { "name": "video_stereo_left", "l2": 5.352513, "mean_abs": 0.120749, "relative": 0.869794 }, { "name": "video_stereo_right", "l2": 6.233654, "mean_abs": 0.14212, "relative": 0.878814 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.033881, "mean_abs": 0.078381, "relative": 0.727896 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014509, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 474, "start_frame": 2370, "end_frame": 2389, "center_frame": 2379, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.060526, "mean_abs": 0.221384, "relative": 0.848648 }, { "name": "hand_right_joints", "l2": 10.428903, "mean_abs": 0.263678, "relative": 0.822076 }, { "name": "body_joints", "l2": 5.761278, "mean_abs": 0.090374, "relative": 0.778171 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.137721, "mean_abs": 0.018048, "relative": 0.325001 }, { "name": "camera_rotation_matrix", "l2": 3.007915, "mean_abs": 0.223517, "relative": 0.923682 }, { "name": "imu_accel_gyro", "l2": 17.22419, "mean_abs": 1.00706, "relative": 0.929304 }, { "name": "depth_confidence", "l2": 12.048514, "mean_abs": 0.224934, "relative": 0.0004 }, { "name": "video_fisheye_cam0", "l2": 5.141273, "mean_abs": 0.108235, "relative": 0.930975 }, { "name": "video_fisheye_cam1", "l2": 5.538305, "mean_abs": 0.119856, "relative": 0.915158 }, { "name": "video_fisheye_cam2", "l2": 5.732718, "mean_abs": 0.119714, "relative": 0.915143 }, { "name": "video_fisheye_cam3", "l2": 5.654102, "mean_abs": 0.116415, "relative": 0.946711 }, { "name": "video_stereo_left", "l2": 5.377595, "mean_abs": 0.122182, "relative": 0.87387 }, { "name": "video_stereo_right", "l2": 6.273915, "mean_abs": 0.144484, "relative": 0.88449 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.04911, "mean_abs": 0.079109, "relative": 0.733346 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014509, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 475, "start_frame": 2375, "end_frame": 2394, "center_frame": 2384, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.062567, "mean_abs": 0.217385, "relative": 0.84882 }, { "name": "hand_right_joints", "l2": 10.437523, "mean_abs": 0.26086, "relative": 0.822756 }, { "name": "body_joints", "l2": 5.721784, "mean_abs": 0.089202, "relative": 0.772837 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.113938, "mean_abs": 0.015161, "relative": 0.268879 }, { "name": "camera_rotation_matrix", "l2": 3.002362, "mean_abs": 0.217795, "relative": 0.921977 }, { "name": "imu_accel_gyro", "l2": 17.251196, "mean_abs": 1.003271, "relative": 0.930761 }, { "name": "depth_confidence", "l2": 11.94388, "mean_abs": 0.21989, "relative": 0.000396 }, { "name": "video_fisheye_cam0", "l2": 5.156396, "mean_abs": 0.10804, "relative": 0.933713 }, { "name": "video_fisheye_cam1", "l2": 5.534628, "mean_abs": 0.118696, "relative": 0.914551 }, { "name": "video_fisheye_cam2", "l2": 5.710725, "mean_abs": 0.11874, "relative": 0.911632 }, { "name": "video_fisheye_cam3", "l2": 5.666611, "mean_abs": 0.1161, "relative": 0.948806 }, { "name": "video_stereo_left", "l2": 5.378169, "mean_abs": 0.121334, "relative": 0.873963 }, { "name": "video_stereo_right", "l2": 6.271615, "mean_abs": 0.143748, "relative": 0.884166 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.318784, "mean_abs": 0.092005, "relative": 0.829858 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014509, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 476, "start_frame": 2380, "end_frame": 2399, "center_frame": 2389, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.072392, "mean_abs": 0.212395, "relative": 0.849649 }, { "name": "hand_right_joints", "l2": 10.415021, "mean_abs": 0.256132, "relative": 0.820982 }, { "name": "body_joints", "l2": 5.674875, "mean_abs": 0.086773, "relative": 0.766501 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.066327, "mean_abs": 0.009644, "relative": 0.156522 }, { "name": "camera_rotation_matrix", "l2": 3.000305, "mean_abs": 0.211955, "relative": 0.921345 }, { "name": "imu_accel_gyro", "l2": 17.310837, "mean_abs": 0.99778, "relative": 0.933979 }, { "name": "depth_confidence", "l2": 11.859909, "mean_abs": 0.217005, "relative": 0.000394 }, { "name": "video_fisheye_cam0", "l2": 5.166637, "mean_abs": 0.107511, "relative": 0.935567 }, { "name": "video_fisheye_cam1", "l2": 5.525101, "mean_abs": 0.117419, "relative": 0.912976 }, { "name": "video_fisheye_cam2", "l2": 5.70037, "mean_abs": 0.117846, "relative": 0.909979 }, { "name": "video_fisheye_cam3", "l2": 5.654245, "mean_abs": 0.115488, "relative": 0.946735 }, { "name": "video_stereo_left", "l2": 5.366539, "mean_abs": 0.120954, "relative": 0.872073 }, { "name": "video_stereo_right", "l2": 6.258522, "mean_abs": 0.14266, "relative": 0.88232 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.318437, "mean_abs": 0.09542, "relative": 0.829734 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014509, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 477, "start_frame": 2385, "end_frame": 2404, "center_frame": 2394, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.14427, "mean_abs": 0.214079, "relative": 0.855712 }, { "name": "hand_right_joints", "l2": 10.423326, "mean_abs": 0.253397, "relative": 0.821637 }, { "name": "body_joints", "l2": 5.631135, "mean_abs": 0.085146, "relative": 0.760593 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.050049, "mean_abs": 0.007634, "relative": 0.118109 }, { "name": "camera_rotation_matrix", "l2": 3.004859, "mean_abs": 0.212501, "relative": 0.922743 }, { "name": "imu_accel_gyro", "l2": 17.190536, "mean_abs": 1.021243, "relative": 0.927489 }, { "name": "depth_confidence", "l2": 11.833479, "mean_abs": 0.215198, "relative": 0.000393 }, { "name": "video_fisheye_cam0", "l2": 5.167338, "mean_abs": 0.10784, "relative": 0.935694 }, { "name": "video_fisheye_cam1", "l2": 5.505558, "mean_abs": 0.117719, "relative": 0.909747 }, { "name": "video_fisheye_cam2", "l2": 5.690671, "mean_abs": 0.117633, "relative": 0.908431 }, { "name": "video_fisheye_cam3", "l2": 5.619081, "mean_abs": 0.114905, "relative": 0.940847 }, { "name": "video_stereo_left", "l2": 5.366295, "mean_abs": 0.120953, "relative": 0.872034 }, { "name": "video_stereo_right", "l2": 6.263, "mean_abs": 0.141536, "relative": 0.882951 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.293415, "mean_abs": 0.091984, "relative": 0.820779 }, { "name": "caption_objects_interaction_text", "l2": 1.96589, "mean_abs": 0.022035, "relative": 0.902689 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 478, "start_frame": 2390, "end_frame": 2409, "center_frame": 2399, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.176277, "mean_abs": 0.217772, "relative": 0.858412 }, { "name": "hand_right_joints", "l2": 10.470324, "mean_abs": 0.257505, "relative": 0.825342 }, { "name": "body_joints", "l2": 5.60858, "mean_abs": 0.084874, "relative": 0.757546 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.080846, "mean_abs": 0.010186, "relative": 0.190785 }, { "name": "camera_rotation_matrix", "l2": 3.002835, "mean_abs": 0.212656, "relative": 0.922122 }, { "name": "imu_accel_gyro", "l2": 17.121286, "mean_abs": 0.995785, "relative": 0.923752 }, { "name": "depth_confidence", "l2": 11.739947, "mean_abs": 0.212208, "relative": 0.00039 }, { "name": "video_fisheye_cam0", "l2": 5.162149, "mean_abs": 0.108124, "relative": 0.934755 }, { "name": "video_fisheye_cam1", "l2": 5.479992, "mean_abs": 0.117172, "relative": 0.905522 }, { "name": "video_fisheye_cam2", "l2": 5.69274, "mean_abs": 0.117404, "relative": 0.908761 }, { "name": "video_fisheye_cam3", "l2": 5.583231, "mean_abs": 0.114188, "relative": 0.934845 }, { "name": "video_stereo_left", "l2": 5.372365, "mean_abs": 0.120153, "relative": 0.87302 }, { "name": "video_stereo_right", "l2": 6.268866, "mean_abs": 0.140662, "relative": 0.883779 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.260977, "mean_abs": 0.091814, "relative": 0.80917 }, { "name": "caption_objects_interaction_text", "l2": 1.96589, "mean_abs": 0.022204, "relative": 0.902689 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 479, "start_frame": 2395, "end_frame": 2414, "center_frame": 2404, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "water bottle", "coffee mug", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.207141, "mean_abs": 0.220408, "relative": 0.861016 }, { "name": "hand_right_joints", "l2": 10.482659, "mean_abs": 0.258569, "relative": 0.826314 }, { "name": "body_joints", "l2": 5.575353, "mean_abs": 0.083525, "relative": 0.753058 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.084131, "mean_abs": 0.010567, "relative": 0.198537 }, { "name": "camera_rotation_matrix", "l2": 3.000575, "mean_abs": 0.212291, "relative": 0.921428 }, { "name": "imu_accel_gyro", "l2": 17.096882, "mean_abs": 0.975201, "relative": 0.922436 }, { "name": "depth_confidence", "l2": 11.721546, "mean_abs": 0.211718, "relative": 0.000389 }, { "name": "video_fisheye_cam0", "l2": 5.148758, "mean_abs": 0.107925, "relative": 0.93233 }, { "name": "video_fisheye_cam1", "l2": 5.450072, "mean_abs": 0.115962, "relative": 0.900578 }, { "name": "video_fisheye_cam2", "l2": 5.704377, "mean_abs": 0.117143, "relative": 0.910619 }, { "name": "video_fisheye_cam3", "l2": 5.562768, "mean_abs": 0.113732, "relative": 0.931418 }, { "name": "video_stereo_left", "l2": 5.374581, "mean_abs": 0.118839, "relative": 0.87338 }, { "name": "video_stereo_right", "l2": 6.262702, "mean_abs": 0.139683, "relative": 0.88291 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.098012, "mean_abs": 0.089972, "relative": 0.750847 }, { "name": "caption_objects_interaction_text", "l2": 1.96589, "mean_abs": 0.021721, "relative": 0.902689 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 480, "start_frame": 2400, "end_frame": 2419, "center_frame": 2409, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.219044, "mean_abs": 0.220923, "relative": 0.86202 }, { "name": "hand_right_joints", "l2": 10.497815, "mean_abs": 0.259685, "relative": 0.827509 }, { "name": "body_joints", "l2": 5.565909, "mean_abs": 0.083453, "relative": 0.751783 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.070453, "mean_abs": 0.008616, "relative": 0.166259 }, { "name": "camera_rotation_matrix", "l2": 2.999358, "mean_abs": 0.21235, "relative": 0.921054 }, { "name": "imu_accel_gyro", "l2": 17.019299, "mean_abs": 0.975898, "relative": 0.91825 }, { "name": "depth_confidence", "l2": 11.728744, "mean_abs": 0.211183, "relative": 0.000389 }, { "name": "video_fisheye_cam0", "l2": 5.133596, "mean_abs": 0.107435, "relative": 0.929584 }, { "name": "video_fisheye_cam1", "l2": 5.41899, "mean_abs": 0.114971, "relative": 0.895442 }, { "name": "video_fisheye_cam2", "l2": 5.70656, "mean_abs": 0.116877, "relative": 0.910967 }, { "name": "video_fisheye_cam3", "l2": 5.534963, "mean_abs": 0.113015, "relative": 0.926763 }, { "name": "video_stereo_left", "l2": 5.367607, "mean_abs": 0.118333, "relative": 0.872247 }, { "name": "video_stereo_right", "l2": 6.248813, "mean_abs": 0.139286, "relative": 0.880952 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.082069, "mean_abs": 0.084998, "relative": 0.745141 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 481, "start_frame": 2405, "end_frame": 2424, "center_frame": 2414, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.209648, "mean_abs": 0.220405, "relative": 0.861227 }, { "name": "hand_right_joints", "l2": 10.521742, "mean_abs": 0.258506, "relative": 0.829395 }, { "name": "body_joints", "l2": 5.554218, "mean_abs": 0.083415, "relative": 0.750204 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.050741, "mean_abs": 0.005926, "relative": 0.119742 }, { "name": "camera_rotation_matrix", "l2": 3.000238, "mean_abs": 0.211869, "relative": 0.921325 }, { "name": "imu_accel_gyro", "l2": 17.021425, "mean_abs": 0.97078, "relative": 0.918365 }, { "name": "depth_confidence", "l2": 11.563837, "mean_abs": 0.210106, "relative": 0.000384 }, { "name": "video_fisheye_cam0", "l2": 5.120082, "mean_abs": 0.106807, "relative": 0.927137 }, { "name": "video_fisheye_cam1", "l2": 5.39519, "mean_abs": 0.1138, "relative": 0.89151 }, { "name": "video_fisheye_cam2", "l2": 5.713669, "mean_abs": 0.116648, "relative": 0.912102 }, { "name": "video_fisheye_cam3", "l2": 5.49874, "mean_abs": 0.111981, "relative": 0.920698 }, { "name": "video_stereo_left", "l2": 5.369181, "mean_abs": 0.117794, "relative": 0.872503 }, { "name": "video_stereo_right", "l2": 6.239191, "mean_abs": 0.138243, "relative": 0.879595 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.048552, "mean_abs": 0.077592, "relative": 0.733146 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 482, "start_frame": 2410, "end_frame": 2429, "center_frame": 2419, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.2085, "mean_abs": 0.21906, "relative": 0.861131 }, { "name": "hand_right_joints", "l2": 10.532602, "mean_abs": 0.25716, "relative": 0.830251 }, { "name": "body_joints", "l2": 5.538293, "mean_abs": 0.083191, "relative": 0.748053 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.027178, "mean_abs": 0.003272, "relative": 0.064137 }, { "name": "camera_rotation_matrix", "l2": 3.000579, "mean_abs": 0.211691, "relative": 0.921429 }, { "name": "imu_accel_gyro", "l2": 17.057446, "mean_abs": 0.961194, "relative": 0.920308 }, { "name": "depth_confidence", "l2": 11.741345, "mean_abs": 0.211397, "relative": 0.00039 }, { "name": "video_fisheye_cam0", "l2": 5.107123, "mean_abs": 0.106343, "relative": 0.924791 }, { "name": "video_fisheye_cam1", "l2": 5.376674, "mean_abs": 0.113389, "relative": 0.88845 }, { "name": "video_fisheye_cam2", "l2": 5.71521, "mean_abs": 0.116588, "relative": 0.912348 }, { "name": "video_fisheye_cam3", "l2": 5.469717, "mean_abs": 0.111413, "relative": 0.915838 }, { "name": "video_stereo_left", "l2": 5.36469, "mean_abs": 0.117423, "relative": 0.871773 }, { "name": "video_stereo_right", "l2": 6.222274, "mean_abs": 0.137699, "relative": 0.87721 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.040375, "mean_abs": 0.077518, "relative": 0.73022 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 483, "start_frame": 2415, "end_frame": 2434, "center_frame": 2424, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.199606, "mean_abs": 0.216553, "relative": 0.86038 }, { "name": "hand_right_joints", "l2": 10.548427, "mean_abs": 0.25506, "relative": 0.831498 }, { "name": "body_joints", "l2": 5.522121, "mean_abs": 0.082601, "relative": 0.745868 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010519, "mean_abs": 0.001448, "relative": 0.024823 }, { "name": "camera_rotation_matrix", "l2": 2.999944, "mean_abs": 0.210308, "relative": 0.921234 }, { "name": "imu_accel_gyro", "l2": 17.15134, "mean_abs": 0.953582, "relative": 0.925374 }, { "name": "depth_confidence", "l2": 11.813204, "mean_abs": 0.210844, "relative": 0.000392 }, { "name": "video_fisheye_cam0", "l2": 5.098188, "mean_abs": 0.105706, "relative": 0.923173 }, { "name": "video_fisheye_cam1", "l2": 5.367389, "mean_abs": 0.112647, "relative": 0.886916 }, { "name": "video_fisheye_cam2", "l2": 5.715479, "mean_abs": 0.116085, "relative": 0.912391 }, { "name": "video_fisheye_cam3", "l2": 5.452664, "mean_abs": 0.110345, "relative": 0.912983 }, { "name": "video_stereo_left", "l2": 5.355548, "mean_abs": 0.115849, "relative": 0.870287 }, { "name": "video_stereo_right", "l2": 6.200334, "mean_abs": 0.135787, "relative": 0.874117 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.066059, "mean_abs": 0.080242, "relative": 0.739412 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 484, "start_frame": 2420, "end_frame": 2439, "center_frame": 2429, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.202747, "mean_abs": 0.216679, "relative": 0.860645 }, { "name": "hand_right_joints", "l2": 10.566034, "mean_abs": 0.255331, "relative": 0.832886 }, { "name": "body_joints", "l2": 5.503728, "mean_abs": 0.08263, "relative": 0.743384 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007505, "mean_abs": 0.000939, "relative": 0.01771 }, { "name": "camera_rotation_matrix", "l2": 2.999892, "mean_abs": 0.21044, "relative": 0.921218 }, { "name": "imu_accel_gyro", "l2": 17.158712, "mean_abs": 0.946651, "relative": 0.925772 }, { "name": "depth_confidence", "l2": 11.817564, "mean_abs": 0.210919, "relative": 0.000392 }, { "name": "video_fisheye_cam0", "l2": 5.097637, "mean_abs": 0.105711, "relative": 0.923073 }, { "name": "video_fisheye_cam1", "l2": 5.365334, "mean_abs": 0.112592, "relative": 0.886576 }, { "name": "video_fisheye_cam2", "l2": 5.713769, "mean_abs": 0.116001, "relative": 0.912118 }, { "name": "video_fisheye_cam3", "l2": 5.468483, "mean_abs": 0.11072, "relative": 0.915632 }, { "name": "video_stereo_left", "l2": 5.354695, "mean_abs": 0.115928, "relative": 0.870149 }, { "name": "video_stereo_right", "l2": 6.197109, "mean_abs": 0.135834, "relative": 0.873662 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.075124, "mean_abs": 0.08261, "relative": 0.742656 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 485, "start_frame": 2425, "end_frame": 2444, "center_frame": 2434, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.207361, "mean_abs": 0.216115, "relative": 0.861034 }, { "name": "hand_right_joints", "l2": 10.569656, "mean_abs": 0.255152, "relative": 0.833172 }, { "name": "body_joints", "l2": 5.489153, "mean_abs": 0.082246, "relative": 0.741415 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010516, "mean_abs": 0.001483, "relative": 0.024817 }, { "name": "camera_rotation_matrix", "l2": 2.999741, "mean_abs": 0.210995, "relative": 0.921172 }, { "name": "imu_accel_gyro", "l2": 17.114645, "mean_abs": 0.943227, "relative": 0.923394 }, { "name": "depth_confidence", "l2": 11.844188, "mean_abs": 0.211753, "relative": 0.000393 }, { "name": "video_fisheye_cam0", "l2": 5.095109, "mean_abs": 0.105706, "relative": 0.922615 }, { "name": "video_fisheye_cam1", "l2": 5.355922, "mean_abs": 0.112529, "relative": 0.885021 }, { "name": "video_fisheye_cam2", "l2": 5.711566, "mean_abs": 0.11611, "relative": 0.911767 }, { "name": "video_fisheye_cam3", "l2": 5.468438, "mean_abs": 0.110949, "relative": 0.915624 }, { "name": "video_stereo_left", "l2": 5.346263, "mean_abs": 0.116066, "relative": 0.868778 }, { "name": "video_stereo_right", "l2": 6.182194, "mean_abs": 0.135845, "relative": 0.87156 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.966755, "mean_abs": 0.077951, "relative": 0.703872 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 486, "start_frame": 2430, "end_frame": 2449, "center_frame": 2439, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.225577, "mean_abs": 0.21638, "relative": 0.862571 }, { "name": "hand_right_joints", "l2": 10.572222, "mean_abs": 0.254387, "relative": 0.833374 }, { "name": "body_joints", "l2": 5.466319, "mean_abs": 0.081812, "relative": 0.738331 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009812, "mean_abs": 0.001493, "relative": 0.023154 }, { "name": "camera_rotation_matrix", "l2": 2.999957, "mean_abs": 0.212376, "relative": 0.921238 }, { "name": "imu_accel_gyro", "l2": 17.077938, "mean_abs": 0.944928, "relative": 0.921414 }, { "name": "depth_confidence", "l2": 11.836544, "mean_abs": 0.212247, "relative": 0.000393 }, { "name": "video_fisheye_cam0", "l2": 5.091321, "mean_abs": 0.105833, "relative": 0.921929 }, { "name": "video_fisheye_cam1", "l2": 5.349447, "mean_abs": 0.112971, "relative": 0.883951 }, { "name": "video_fisheye_cam2", "l2": 5.713619, "mean_abs": 0.116477, "relative": 0.912094 }, { "name": "video_fisheye_cam3", "l2": 5.47584, "mean_abs": 0.111109, "relative": 0.916863 }, { "name": "video_stereo_left", "l2": 5.338959, "mean_abs": 0.116466, "relative": 0.867592 }, { "name": "video_stereo_right", "l2": 6.169263, "mean_abs": 0.136071, "relative": 0.869737 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.02214, "mean_abs": 0.079446, "relative": 0.723694 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 487, "start_frame": 2435, "end_frame": 2454, "center_frame": 2444, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.243021, "mean_abs": 0.216581, "relative": 0.864043 }, { "name": "hand_right_joints", "l2": 10.5714, "mean_abs": 0.253604, "relative": 0.833309 }, { "name": "body_joints", "l2": 5.434362, "mean_abs": 0.081268, "relative": 0.734015 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01323, "mean_abs": 0.001628, "relative": 0.03122 }, { "name": "camera_rotation_matrix", "l2": 3.000916, "mean_abs": 0.213895, "relative": 0.921532 }, { "name": "imu_accel_gyro", "l2": 17.076429, "mean_abs": 0.936421, "relative": 0.921332 }, { "name": "depth_confidence", "l2": 11.784391, "mean_abs": 0.212835, "relative": 0.000391 }, { "name": "video_fisheye_cam0", "l2": 5.0909, "mean_abs": 0.106203, "relative": 0.921853 }, { "name": "video_fisheye_cam1", "l2": 5.34934, "mean_abs": 0.113698, "relative": 0.883933 }, { "name": "video_fisheye_cam2", "l2": 5.718198, "mean_abs": 0.117024, "relative": 0.912825 }, { "name": "video_fisheye_cam3", "l2": 5.481987, "mean_abs": 0.111475, "relative": 0.917893 }, { "name": "video_stereo_left", "l2": 5.337483, "mean_abs": 0.117392, "relative": 0.867352 }, { "name": "video_stereo_right", "l2": 6.167159, "mean_abs": 0.137282, "relative": 0.86944 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.007842, "mean_abs": 0.07811, "relative": 0.718577 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 488, "start_frame": 2440, "end_frame": 2459, "center_frame": 2449, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.250892, "mean_abs": 0.216676, "relative": 0.864706 }, { "name": "hand_right_joints", "l2": 10.572814, "mean_abs": 0.25309, "relative": 0.833421 }, { "name": "body_joints", "l2": 5.395349, "mean_abs": 0.08103, "relative": 0.728745 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012913, "mean_abs": 0.001526, "relative": 0.030474 }, { "name": "camera_rotation_matrix", "l2": 3.001003, "mean_abs": 0.213775, "relative": 0.921559 }, { "name": "imu_accel_gyro", "l2": 17.104691, "mean_abs": 0.946142, "relative": 0.922857 }, { "name": "depth_confidence", "l2": 11.743988, "mean_abs": 0.212442, "relative": 0.00039 }, { "name": "video_fisheye_cam0", "l2": 5.087615, "mean_abs": 0.105995, "relative": 0.921258 }, { "name": "video_fisheye_cam1", "l2": 5.341492, "mean_abs": 0.113282, "relative": 0.882636 }, { "name": "video_fisheye_cam2", "l2": 5.719128, "mean_abs": 0.11688, "relative": 0.912974 }, { "name": "video_fisheye_cam3", "l2": 5.477235, "mean_abs": 0.111256, "relative": 0.917097 }, { "name": "video_stereo_left", "l2": 5.331059, "mean_abs": 0.116936, "relative": 0.866308 }, { "name": "video_stereo_right", "l2": 6.155046, "mean_abs": 0.136568, "relative": 0.867732 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.055074, "mean_abs": 0.079157, "relative": 0.73548 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 489, "start_frame": 2445, "end_frame": 2464, "center_frame": 2454, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.270334, "mean_abs": 0.215457, "relative": 0.866347 }, { "name": "hand_right_joints", "l2": 10.575789, "mean_abs": 0.251868, "relative": 0.833655 }, { "name": "body_joints", "l2": 5.377371, "mean_abs": 0.080883, "relative": 0.726317 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010075, "mean_abs": 0.001309, "relative": 0.023775 }, { "name": "camera_rotation_matrix", "l2": 3.000417, "mean_abs": 0.212607, "relative": 0.921379 }, { "name": "imu_accel_gyro", "l2": 17.050785, "mean_abs": 0.942117, "relative": 0.919949 }, { "name": "depth_confidence", "l2": 11.850442, "mean_abs": 0.212164, "relative": 0.000393 }, { "name": "video_fisheye_cam0", "l2": 5.084634, "mean_abs": 0.105751, "relative": 0.920719 }, { "name": "video_fisheye_cam1", "l2": 5.339418, "mean_abs": 0.112666, "relative": 0.882294 }, { "name": "video_fisheye_cam2", "l2": 5.719133, "mean_abs": 0.11648, "relative": 0.912975 }, { "name": "video_fisheye_cam3", "l2": 5.476041, "mean_abs": 0.111328, "relative": 0.916897 }, { "name": "video_stereo_left", "l2": 5.326252, "mean_abs": 0.116, "relative": 0.865527 }, { "name": "video_stereo_right", "l2": 6.145154, "mean_abs": 0.135509, "relative": 0.866338 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.068179, "mean_abs": 0.079631, "relative": 0.74017 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 490, "start_frame": 2450, "end_frame": 2469, "center_frame": 2459, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.291962, "mean_abs": 0.214678, "relative": 0.868171 }, { "name": "hand_right_joints", "l2": 10.584257, "mean_abs": 0.250444, "relative": 0.834323 }, { "name": "body_joints", "l2": 5.354946, "mean_abs": 0.080475, "relative": 0.723288 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013797, "mean_abs": 0.001819, "relative": 0.032558 }, { "name": "camera_rotation_matrix", "l2": 3.000124, "mean_abs": 0.212983, "relative": 0.921289 }, { "name": "imu_accel_gyro", "l2": 17.048267, "mean_abs": 0.935268, "relative": 0.919813 }, { "name": "depth_confidence", "l2": 11.850105, "mean_abs": 0.213087, "relative": 0.000393 }, { "name": "video_fisheye_cam0", "l2": 5.079004, "mean_abs": 0.105519, "relative": 0.919699 }, { "name": "video_fisheye_cam1", "l2": 5.318337, "mean_abs": 0.111735, "relative": 0.87881 }, { "name": "video_fisheye_cam2", "l2": 5.716401, "mean_abs": 0.116224, "relative": 0.912538 }, { "name": "video_fisheye_cam3", "l2": 5.459348, "mean_abs": 0.110307, "relative": 0.914102 }, { "name": "video_stereo_left", "l2": 5.314338, "mean_abs": 0.115411, "relative": 0.863591 }, { "name": "video_stereo_right", "l2": 6.120454, "mean_abs": 0.13414, "relative": 0.862856 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.081473, "mean_abs": 0.080472, "relative": 0.744928 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 491, "start_frame": 2455, "end_frame": 2474, "center_frame": 2464, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.276955, "mean_abs": 0.213754, "relative": 0.866905 }, { "name": "hand_right_joints", "l2": 10.580458, "mean_abs": 0.249197, "relative": 0.834023 }, { "name": "body_joints", "l2": 5.337405, "mean_abs": 0.079564, "relative": 0.720919 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013582, "mean_abs": 0.001965, "relative": 0.032052 }, { "name": "camera_rotation_matrix", "l2": 3.000036, "mean_abs": 0.212958, "relative": 0.921262 }, { "name": "imu_accel_gyro", "l2": 17.051966, "mean_abs": 0.939612, "relative": 0.920012 }, { "name": "depth_confidence", "l2": 11.825532, "mean_abs": 0.212829, "relative": 0.000392 }, { "name": "video_fisheye_cam0", "l2": 5.075632, "mean_abs": 0.105409, "relative": 0.919088 }, { "name": "video_fisheye_cam1", "l2": 5.328452, "mean_abs": 0.112087, "relative": 0.880482 }, { "name": "video_fisheye_cam2", "l2": 5.716477, "mean_abs": 0.116171, "relative": 0.912551 }, { "name": "video_fisheye_cam3", "l2": 5.475722, "mean_abs": 0.110875, "relative": 0.916844 }, { "name": "video_stereo_left", "l2": 5.310313, "mean_abs": 0.115569, "relative": 0.862936 }, { "name": "video_stereo_right", "l2": 6.134671, "mean_abs": 0.134864, "relative": 0.86486 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.085632, "mean_abs": 0.079168, "relative": 0.746417 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 492, "start_frame": 2460, "end_frame": 2479, "center_frame": 2469, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.249075, "mean_abs": 0.211877, "relative": 0.864553 }, { "name": "hand_right_joints", "l2": 10.573627, "mean_abs": 0.248202, "relative": 0.833485 }, { "name": "body_joints", "l2": 5.327661, "mean_abs": 0.078538, "relative": 0.719603 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018696, "mean_abs": 0.002374, "relative": 0.04412 }, { "name": "camera_rotation_matrix", "l2": 2.999645, "mean_abs": 0.213432, "relative": 0.921142 }, { "name": "imu_accel_gyro", "l2": 17.047335, "mean_abs": 0.934479, "relative": 0.919762 }, { "name": "depth_confidence", "l2": 11.795437, "mean_abs": 0.212088, "relative": 0.000391 }, { "name": "video_fisheye_cam0", "l2": 5.076098, "mean_abs": 0.105421, "relative": 0.919173 }, { "name": "video_fisheye_cam1", "l2": 5.34129, "mean_abs": 0.112368, "relative": 0.882603 }, { "name": "video_fisheye_cam2", "l2": 5.716424, "mean_abs": 0.116273, "relative": 0.912542 }, { "name": "video_fisheye_cam3", "l2": 5.513834, "mean_abs": 0.111856, "relative": 0.923225 }, { "name": "video_stereo_left", "l2": 5.309995, "mean_abs": 0.115539, "relative": 0.862885 }, { "name": "video_stereo_right", "l2": 6.143872, "mean_abs": 0.135193, "relative": 0.866157 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.06687, "mean_abs": 0.081135, "relative": 0.739702 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 493, "start_frame": 2465, "end_frame": 2484, "center_frame": 2474, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.234076, "mean_abs": 0.211065, "relative": 0.863288 }, { "name": "hand_right_joints", "l2": 10.575193, "mean_abs": 0.247885, "relative": 0.833608 }, { "name": "body_joints", "l2": 5.329516, "mean_abs": 0.078349, "relative": 0.719853 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.022354, "mean_abs": 0.003091, "relative": 0.052753 }, { "name": "camera_rotation_matrix", "l2": 3.000038, "mean_abs": 0.21436, "relative": 0.921263 }, { "name": "imu_accel_gyro", "l2": 17.048244, "mean_abs": 0.944611, "relative": 0.919811 }, { "name": "depth_confidence", "l2": 11.627805, "mean_abs": 0.211606, "relative": 0.000386 }, { "name": "video_fisheye_cam0", "l2": 5.074823, "mean_abs": 0.105639, "relative": 0.918942 }, { "name": "video_fisheye_cam1", "l2": 5.353781, "mean_abs": 0.112914, "relative": 0.884667 }, { "name": "video_fisheye_cam2", "l2": 5.711166, "mean_abs": 0.116573, "relative": 0.911703 }, { "name": "video_fisheye_cam3", "l2": 5.540926, "mean_abs": 0.112393, "relative": 0.927761 }, { "name": "video_stereo_left", "l2": 5.312383, "mean_abs": 0.116321, "relative": 0.863273 }, { "name": "video_stereo_right", "l2": 6.152023, "mean_abs": 0.135839, "relative": 0.867306 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.983491, "mean_abs": 0.076168, "relative": 0.709862 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 494, "start_frame": 2470, "end_frame": 2489, "center_frame": 2479, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.223571, "mean_abs": 0.209322, "relative": 0.862402 }, { "name": "hand_right_joints", "l2": 10.577284, "mean_abs": 0.247641, "relative": 0.833773 }, { "name": "body_joints", "l2": 5.340963, "mean_abs": 0.079137, "relative": 0.721399 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028427, "mean_abs": 0.003955, "relative": 0.067083 }, { "name": "camera_rotation_matrix", "l2": 3.000865, "mean_abs": 0.215997, "relative": 0.921517 }, { "name": "imu_accel_gyro", "l2": 17.078581, "mean_abs": 0.9455, "relative": 0.921448 }, { "name": "depth_confidence", "l2": 11.633418, "mean_abs": 0.21219, "relative": 0.000386 }, { "name": "video_fisheye_cam0", "l2": 5.077455, "mean_abs": 0.106031, "relative": 0.919418 }, { "name": "video_fisheye_cam1", "l2": 5.375032, "mean_abs": 0.1139, "relative": 0.888179 }, { "name": "video_fisheye_cam2", "l2": 5.709113, "mean_abs": 0.117085, "relative": 0.911375 }, { "name": "video_fisheye_cam3", "l2": 5.575068, "mean_abs": 0.113005, "relative": 0.933478 }, { "name": "video_stereo_left", "l2": 5.316365, "mean_abs": 0.11726, "relative": 0.86392 }, { "name": "video_stereo_right", "l2": 6.172559, "mean_abs": 0.137202, "relative": 0.870201 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.034447, "mean_abs": 0.078924, "relative": 0.728098 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 495, "start_frame": 2475, "end_frame": 2494, "center_frame": 2484, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.1785, "mean_abs": 0.207525, "relative": 0.8586 }, { "name": "hand_right_joints", "l2": 10.583909, "mean_abs": 0.246961, "relative": 0.834295 }, { "name": "body_joints", "l2": 5.361928, "mean_abs": 0.079912, "relative": 0.724231 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.027396, "mean_abs": 0.003744, "relative": 0.06465 }, { "name": "camera_rotation_matrix", "l2": 3.001006, "mean_abs": 0.215566, "relative": 0.92156 }, { "name": "imu_accel_gyro", "l2": 17.064924, "mean_abs": 0.940819, "relative": 0.920711 }, { "name": "depth_confidence", "l2": 12.041613, "mean_abs": 0.215323, "relative": 0.0004 }, { "name": "video_fisheye_cam0", "l2": 5.080944, "mean_abs": 0.106074, "relative": 0.92005 }, { "name": "video_fisheye_cam1", "l2": 5.378806, "mean_abs": 0.11388, "relative": 0.888802 }, { "name": "video_fisheye_cam2", "l2": 5.701201, "mean_abs": 0.116874, "relative": 0.910112 }, { "name": "video_fisheye_cam3", "l2": 5.603024, "mean_abs": 0.113318, "relative": 0.938159 }, { "name": "video_stereo_left", "l2": 5.304093, "mean_abs": 0.116787, "relative": 0.861926 }, { "name": "video_stereo_right", "l2": 6.158606, "mean_abs": 0.136833, "relative": 0.868234 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.00604, "mean_abs": 0.076391, "relative": 0.717931 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 496, "start_frame": 2480, "end_frame": 2499, "center_frame": 2489, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.174478, "mean_abs": 0.207591, "relative": 0.858261 }, { "name": "hand_right_joints", "l2": 10.586747, "mean_abs": 0.24668, "relative": 0.834519 }, { "name": "body_joints", "l2": 5.386939, "mean_abs": 0.081142, "relative": 0.727609 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.047168, "mean_abs": 0.005301, "relative": 0.111309 }, { "name": "camera_rotation_matrix", "l2": 3.001086, "mean_abs": 0.214243, "relative": 0.921585 }, { "name": "imu_accel_gyro", "l2": 17.043049, "mean_abs": 0.93812, "relative": 0.919531 }, { "name": "depth_confidence", "l2": 12.05134, "mean_abs": 0.216484, "relative": 0.0004 }, { "name": "video_fisheye_cam0", "l2": 5.090171, "mean_abs": 0.106308, "relative": 0.921721 }, { "name": "video_fisheye_cam1", "l2": 5.369315, "mean_abs": 0.113717, "relative": 0.887234 }, { "name": "video_fisheye_cam2", "l2": 5.684806, "mean_abs": 0.116449, "relative": 0.907495 }, { "name": "video_fisheye_cam3", "l2": 5.622695, "mean_abs": 0.113087, "relative": 0.941452 }, { "name": "video_stereo_left", "l2": 5.271637, "mean_abs": 0.116113, "relative": 0.856652 }, { "name": "video_stereo_right", "l2": 6.106082, "mean_abs": 0.136127, "relative": 0.86083 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.018965, "mean_abs": 0.079014, "relative": 0.722557 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 497, "start_frame": 2485, "end_frame": 2504, "center_frame": 2494, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.141036, "mean_abs": 0.208756, "relative": 0.85544 }, { "name": "hand_right_joints", "l2": 10.550799, "mean_abs": 0.247182, "relative": 0.831685 }, { "name": "body_joints", "l2": 5.409032, "mean_abs": 0.082315, "relative": 0.730594 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.096464, "mean_abs": 0.010181, "relative": 0.227641 }, { "name": "camera_rotation_matrix", "l2": 3.00215, "mean_abs": 0.2136, "relative": 0.921912 }, { "name": "imu_accel_gyro", "l2": 17.06233, "mean_abs": 0.936416, "relative": 0.920571 }, { "name": "depth_confidence", "l2": 12.050948, "mean_abs": 0.217364, "relative": 0.0004 }, { "name": "video_fisheye_cam0", "l2": 5.100411, "mean_abs": 0.106617, "relative": 0.923575 }, { "name": "video_fisheye_cam1", "l2": 5.335065, "mean_abs": 0.11368, "relative": 0.881574 }, { "name": "video_fisheye_cam2", "l2": 5.671967, "mean_abs": 0.11649, "relative": 0.905445 }, { "name": "video_fisheye_cam3", "l2": 5.619789, "mean_abs": 0.113662, "relative": 0.940966 }, { "name": "video_stereo_left", "l2": 5.239435, "mean_abs": 0.115528, "relative": 0.851419 }, { "name": "video_stereo_right", "l2": 6.048122, "mean_abs": 0.135794, "relative": 0.852658 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.005675, "mean_abs": 0.079514, "relative": 0.717801 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 498, "start_frame": 2490, "end_frame": 2509, "center_frame": 2499, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.125031, "mean_abs": 0.209373, "relative": 0.854089 }, { "name": "hand_right_joints", "l2": 10.536393, "mean_abs": 0.247692, "relative": 0.83055 }, { "name": "body_joints", "l2": 5.418552, "mean_abs": 0.082222, "relative": 0.731879 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.13621, "mean_abs": 0.014876, "relative": 0.321437 }, { "name": "camera_rotation_matrix", "l2": 3.001229, "mean_abs": 0.214149, "relative": 0.921629 }, { "name": "imu_accel_gyro", "l2": 17.129204, "mean_abs": 0.974209, "relative": 0.92418 }, { "name": "depth_confidence", "l2": 11.990373, "mean_abs": 0.217615, "relative": 0.000398 }, { "name": "video_fisheye_cam0", "l2": 5.116924, "mean_abs": 0.107096, "relative": 0.926565 }, { "name": "video_fisheye_cam1", "l2": 5.311704, "mean_abs": 0.113228, "relative": 0.877714 }, { "name": "video_fisheye_cam2", "l2": 5.663502, "mean_abs": 0.116342, "relative": 0.904094 }, { "name": "video_fisheye_cam3", "l2": 5.648064, "mean_abs": 0.114841, "relative": 0.9457 }, { "name": "video_stereo_left", "l2": 5.220002, "mean_abs": 0.115061, "relative": 0.848261 }, { "name": "video_stereo_right", "l2": 6.013878, "mean_abs": 0.135463, "relative": 0.847831 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.990752, "mean_abs": 0.076081, "relative": 0.71246 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 499, "start_frame": 2495, "end_frame": 2514, "center_frame": 2504, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.076002, "mean_abs": 0.208135, "relative": 0.849954 }, { "name": "hand_right_joints", "l2": 10.514225, "mean_abs": 0.246622, "relative": 0.828802 }, { "name": "body_joints", "l2": 5.421669, "mean_abs": 0.081024, "relative": 0.7323 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.143125, "mean_abs": 0.016104, "relative": 0.337755 }, { "name": "camera_rotation_matrix", "l2": 3.000121, "mean_abs": 0.213876, "relative": 0.921288 }, { "name": "imu_accel_gyro", "l2": 17.169224, "mean_abs": 1.010635, "relative": 0.926339 }, { "name": "depth_confidence", "l2": 11.5904, "mean_abs": 0.213992, "relative": 0.000385 }, { "name": "video_fisheye_cam0", "l2": 5.13222, "mean_abs": 0.106945, "relative": 0.929335 }, { "name": "video_fisheye_cam1", "l2": 5.286262, "mean_abs": 0.112752, "relative": 0.87351 }, { "name": "video_fisheye_cam2", "l2": 5.658628, "mean_abs": 0.116279, "relative": 0.903316 }, { "name": "video_fisheye_cam3", "l2": 5.655666, "mean_abs": 0.114939, "relative": 0.946973 }, { "name": "video_stereo_left", "l2": 5.198188, "mean_abs": 0.114744, "relative": 0.844716 }, { "name": "video_stereo_right", "l2": 5.975327, "mean_abs": 0.135137, "relative": 0.842396 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.046332, "mean_abs": 0.07635, "relative": 0.732351 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 500, "start_frame": 2500, "end_frame": 2519, "center_frame": 2509, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.053538, "mean_abs": 0.206349, "relative": 0.848059 }, { "name": "hand_right_joints", "l2": 10.510642, "mean_abs": 0.245173, "relative": 0.82852 }, { "name": "body_joints", "l2": 5.407053, "mean_abs": 0.079271, "relative": 0.730326 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.102987, "mean_abs": 0.012039, "relative": 0.243036 }, { "name": "camera_rotation_matrix", "l2": 3.001307, "mean_abs": 0.213432, "relative": 0.921653 }, { "name": "imu_accel_gyro", "l2": 17.14452, "mean_abs": 1.002598, "relative": 0.925006 }, { "name": "depth_confidence", "l2": 11.471609, "mean_abs": 0.212152, "relative": 0.000381 }, { "name": "video_fisheye_cam0", "l2": 5.152956, "mean_abs": 0.10715, "relative": 0.93309 }, { "name": "video_fisheye_cam1", "l2": 5.263189, "mean_abs": 0.111602, "relative": 0.869698 }, { "name": "video_fisheye_cam2", "l2": 5.655632, "mean_abs": 0.115772, "relative": 0.902838 }, { "name": "video_fisheye_cam3", "l2": 5.670205, "mean_abs": 0.114779, "relative": 0.949407 }, { "name": "video_stereo_left", "l2": 5.174305, "mean_abs": 0.113959, "relative": 0.840835 }, { "name": "video_stereo_right", "l2": 5.934937, "mean_abs": 0.133709, "relative": 0.836702 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.045336, "mean_abs": 0.077263, "relative": 0.731995 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 501, "start_frame": 2505, "end_frame": 2524, "center_frame": 2514, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.054251, "mean_abs": 0.207196, "relative": 0.848119 }, { "name": "hand_right_joints", "l2": 10.50885, "mean_abs": 0.244913, "relative": 0.828378 }, { "name": "body_joints", "l2": 5.370038, "mean_abs": 0.079365, "relative": 0.725327 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.043146, "mean_abs": 0.005255, "relative": 0.101817 }, { "name": "camera_rotation_matrix", "l2": 3.000669, "mean_abs": 0.211164, "relative": 0.921457 }, { "name": "imu_accel_gyro", "l2": 17.153009, "mean_abs": 0.993671, "relative": 0.925464 }, { "name": "depth_confidence", "l2": 11.40801, "mean_abs": 0.21057, "relative": 0.000379 }, { "name": "video_fisheye_cam0", "l2": 5.166244, "mean_abs": 0.107062, "relative": 0.935496 }, { "name": "video_fisheye_cam1", "l2": 5.262722, "mean_abs": 0.111337, "relative": 0.86962 }, { "name": "video_fisheye_cam2", "l2": 5.652294, "mean_abs": 0.114987, "relative": 0.902305 }, { "name": "video_fisheye_cam3", "l2": 5.678832, "mean_abs": 0.113982, "relative": 0.950852 }, { "name": "video_stereo_left", "l2": 5.160864, "mean_abs": 0.112665, "relative": 0.838651 }, { "name": "video_stereo_right", "l2": 5.92624, "mean_abs": 0.132234, "relative": 0.835476 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.042922, "mean_abs": 0.076733, "relative": 0.731131 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 502, "start_frame": 2510, "end_frame": 2529, "center_frame": 2519, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.061293, "mean_abs": 0.208022, "relative": 0.848713 }, { "name": "hand_right_joints", "l2": 10.525682, "mean_abs": 0.244482, "relative": 0.829705 }, { "name": "body_joints", "l2": 5.343117, "mean_abs": 0.07944, "relative": 0.72169 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01338, "mean_abs": 0.002227, "relative": 0.031575 }, { "name": "camera_rotation_matrix", "l2": 2.999412, "mean_abs": 0.20928, "relative": 0.921071 }, { "name": "imu_accel_gyro", "l2": 17.120804, "mean_abs": 0.983905, "relative": 0.923726 }, { "name": "depth_confidence", "l2": 11.368686, "mean_abs": 0.208995, "relative": 0.000377 }, { "name": "video_fisheye_cam0", "l2": 5.179399, "mean_abs": 0.107197, "relative": 0.937878 }, { "name": "video_fisheye_cam1", "l2": 5.284253, "mean_abs": 0.11214, "relative": 0.873178 }, { "name": "video_fisheye_cam2", "l2": 5.661348, "mean_abs": 0.114915, "relative": 0.90375 }, { "name": "video_fisheye_cam3", "l2": 5.675647, "mean_abs": 0.113697, "relative": 0.950319 }, { "name": "video_stereo_left", "l2": 5.165213, "mean_abs": 0.112037, "relative": 0.839357 }, { "name": "video_stereo_right", "l2": 5.949871, "mean_abs": 0.132317, "relative": 0.838807 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.985841, "mean_abs": 0.074461, "relative": 0.710703 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 503, "start_frame": 2515, "end_frame": 2534, "center_frame": 2524, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.100987, "mean_abs": 0.208445, "relative": 0.852061 }, { "name": "hand_right_joints", "l2": 10.524915, "mean_abs": 0.243626, "relative": 0.829645 }, { "name": "body_joints", "l2": 5.315346, "mean_abs": 0.078445, "relative": 0.717939 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.024721, "mean_abs": 0.003405, "relative": 0.058339 }, { "name": "camera_rotation_matrix", "l2": 2.999968, "mean_abs": 0.209189, "relative": 0.921241 }, { "name": "imu_accel_gyro", "l2": 17.095612, "mean_abs": 0.977579, "relative": 0.922367 }, { "name": "depth_confidence", "l2": 11.327352, "mean_abs": 0.208166, "relative": 0.000376 }, { "name": "video_fisheye_cam0", "l2": 5.194834, "mean_abs": 0.107322, "relative": 0.940673 }, { "name": "video_fisheye_cam1", "l2": 5.322556, "mean_abs": 0.112744, "relative": 0.879508 }, { "name": "video_fisheye_cam2", "l2": 5.685158, "mean_abs": 0.1153, "relative": 0.907551 }, { "name": "video_fisheye_cam3", "l2": 5.678614, "mean_abs": 0.113602, "relative": 0.950815 }, { "name": "video_stereo_left", "l2": 5.189132, "mean_abs": 0.112182, "relative": 0.843244 }, { "name": "video_stereo_right", "l2": 6.007555, "mean_abs": 0.133173, "relative": 0.846939 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.002107, "mean_abs": 0.075488, "relative": 0.716524 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 504, "start_frame": 2520, "end_frame": 2539, "center_frame": 2529, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.128826, "mean_abs": 0.207668, "relative": 0.85441 }, { "name": "hand_right_joints", "l2": 10.517291, "mean_abs": 0.242427, "relative": 0.829044 }, { "name": "body_joints", "l2": 5.290796, "mean_abs": 0.077226, "relative": 0.714624 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030118, "mean_abs": 0.003877, "relative": 0.071075 }, { "name": "camera_rotation_matrix", "l2": 2.999937, "mean_abs": 0.209723, "relative": 0.921232 }, { "name": "imu_accel_gyro", "l2": 17.065264, "mean_abs": 0.96548, "relative": 0.92073 }, { "name": "depth_confidence", "l2": 11.393883, "mean_abs": 0.208662, "relative": 0.000378 }, { "name": "video_fisheye_cam0", "l2": 5.209951, "mean_abs": 0.10774, "relative": 0.943411 }, { "name": "video_fisheye_cam1", "l2": 5.362397, "mean_abs": 0.113359, "relative": 0.886091 }, { "name": "video_fisheye_cam2", "l2": 5.708302, "mean_abs": 0.116166, "relative": 0.911246 }, { "name": "video_fisheye_cam3", "l2": 5.634285, "mean_abs": 0.112715, "relative": 0.943393 }, { "name": "video_stereo_left", "l2": 5.211067, "mean_abs": 0.112924, "relative": 0.846809 }, { "name": "video_stereo_right", "l2": 6.057268, "mean_abs": 0.13398, "relative": 0.853948 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.091473, "mean_abs": 0.080942, "relative": 0.748507 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 505, "start_frame": 2525, "end_frame": 2544, "center_frame": 2534, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.144279, "mean_abs": 0.20867, "relative": 0.855713 }, { "name": "hand_right_joints", "l2": 10.546675, "mean_abs": 0.244304, "relative": 0.83136 }, { "name": "body_joints", "l2": 5.267977, "mean_abs": 0.076152, "relative": 0.711541 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.03656, "mean_abs": 0.004168, "relative": 0.086276 }, { "name": "camera_rotation_matrix", "l2": 2.999905, "mean_abs": 0.210415, "relative": 0.921222 }, { "name": "imu_accel_gyro", "l2": 17.064566, "mean_abs": 0.958983, "relative": 0.920692 }, { "name": "depth_confidence", "l2": 11.419841, "mean_abs": 0.209289, "relative": 0.000379 }, { "name": "video_fisheye_cam0", "l2": 5.223563, "mean_abs": 0.108368, "relative": 0.945876 }, { "name": "video_fisheye_cam1", "l2": 5.412876, "mean_abs": 0.113964, "relative": 0.894432 }, { "name": "video_fisheye_cam2", "l2": 5.737384, "mean_abs": 0.117029, "relative": 0.915888 }, { "name": "video_fisheye_cam3", "l2": 5.645055, "mean_abs": 0.113089, "relative": 0.945196 }, { "name": "video_stereo_left", "l2": 5.240816, "mean_abs": 0.113849, "relative": 0.851643 }, { "name": "video_stereo_right", "l2": 6.116049, "mean_abs": 0.135073, "relative": 0.862235 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.05785, "mean_abs": 0.078445, "relative": 0.736474 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 506, "start_frame": 2530, "end_frame": 2549, "center_frame": 2539, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.14051, "mean_abs": 0.209004, "relative": 0.855395 }, { "name": "hand_right_joints", "l2": 10.557631, "mean_abs": 0.245086, "relative": 0.832224 }, { "name": "body_joints", "l2": 5.284914, "mean_abs": 0.077162, "relative": 0.713829 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.040266, "mean_abs": 0.00449, "relative": 0.095022 }, { "name": "camera_rotation_matrix", "l2": 3.000381, "mean_abs": 0.210539, "relative": 0.921368 }, { "name": "imu_accel_gyro", "l2": 17.087103, "mean_abs": 0.967909, "relative": 0.921908 }, { "name": "depth_confidence", "l2": 11.445836, "mean_abs": 0.209975, "relative": 0.00038 }, { "name": "video_fisheye_cam0", "l2": 5.236189, "mean_abs": 0.10854, "relative": 0.948162 }, { "name": "video_fisheye_cam1", "l2": 5.435504, "mean_abs": 0.114525, "relative": 0.898171 }, { "name": "video_fisheye_cam2", "l2": 5.760515, "mean_abs": 0.117594, "relative": 0.91958 }, { "name": "video_fisheye_cam3", "l2": 5.646096, "mean_abs": 0.113131, "relative": 0.945371 }, { "name": "video_stereo_left", "l2": 5.262634, "mean_abs": 0.114242, "relative": 0.855189 }, { "name": "video_stereo_right", "l2": 6.145669, "mean_abs": 0.135568, "relative": 0.86641 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.075976, "mean_abs": 0.078116, "relative": 0.742961 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 507, "start_frame": 2535, "end_frame": 2554, "center_frame": 2544, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.140566, "mean_abs": 0.208415, "relative": 0.8554 }, { "name": "hand_right_joints", "l2": 10.58145, "mean_abs": 0.245502, "relative": 0.834101 }, { "name": "body_joints", "l2": 5.305205, "mean_abs": 0.077849, "relative": 0.71657 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.035088, "mean_abs": 0.004133, "relative": 0.082804 }, { "name": "camera_rotation_matrix", "l2": 3.000088, "mean_abs": 0.210326, "relative": 0.921278 }, { "name": "imu_accel_gyro", "l2": 17.084208, "mean_abs": 0.960526, "relative": 0.921752 }, { "name": "depth_confidence", "l2": 11.463669, "mean_abs": 0.210327, "relative": 0.00038 }, { "name": "video_fisheye_cam0", "l2": 5.24666, "mean_abs": 0.108582, "relative": 0.950058 }, { "name": "video_fisheye_cam1", "l2": 5.456897, "mean_abs": 0.114882, "relative": 0.901706 }, { "name": "video_fisheye_cam2", "l2": 5.779653, "mean_abs": 0.117693, "relative": 0.922636 }, { "name": "video_fisheye_cam3", "l2": 5.647894, "mean_abs": 0.113312, "relative": 0.945672 }, { "name": "video_stereo_left", "l2": 5.278296, "mean_abs": 0.114331, "relative": 0.857734 }, { "name": "video_stereo_right", "l2": 6.173644, "mean_abs": 0.136203, "relative": 0.870354 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.054112, "mean_abs": 0.078389, "relative": 0.735136 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 508, "start_frame": 2540, "end_frame": 2559, "center_frame": 2549, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.13926, "mean_abs": 0.207905, "relative": 0.85529 }, { "name": "hand_right_joints", "l2": 10.59551, "mean_abs": 0.24574, "relative": 0.83521 }, { "name": "body_joints", "l2": 5.326313, "mean_abs": 0.078511, "relative": 0.719421 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030918, "mean_abs": 0.003825, "relative": 0.072962 }, { "name": "camera_rotation_matrix", "l2": 2.999238, "mean_abs": 0.210558, "relative": 0.921017 }, { "name": "imu_accel_gyro", "l2": 17.083561, "mean_abs": 0.958114, "relative": 0.921717 }, { "name": "depth_confidence", "l2": 11.460224, "mean_abs": 0.20929, "relative": 0.00038 }, { "name": "video_fisheye_cam0", "l2": 5.260074, "mean_abs": 0.108917, "relative": 0.952487 }, { "name": "video_fisheye_cam1", "l2": 5.490284, "mean_abs": 0.115562, "relative": 0.907223 }, { "name": "video_fisheye_cam2", "l2": 5.803683, "mean_abs": 0.117929, "relative": 0.926472 }, { "name": "video_fisheye_cam3", "l2": 5.668428, "mean_abs": 0.11352, "relative": 0.94911 }, { "name": "video_stereo_left", "l2": 5.302442, "mean_abs": 0.115157, "relative": 0.861657 }, { "name": "video_stereo_right", "l2": 6.214859, "mean_abs": 0.137323, "relative": 0.876165 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.958684, "mean_abs": 0.074202, "relative": 0.700983 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 509, "start_frame": 2545, "end_frame": 2564, "center_frame": 2554, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.134769, "mean_abs": 0.208239, "relative": 0.854911 }, { "name": "hand_right_joints", "l2": 10.609924, "mean_abs": 0.247543, "relative": 0.836346 }, { "name": "body_joints", "l2": 5.352887, "mean_abs": 0.07902, "relative": 0.72301 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.032868, "mean_abs": 0.003707, "relative": 0.077563 }, { "name": "camera_rotation_matrix", "l2": 2.999853, "mean_abs": 0.210159, "relative": 0.921206 }, { "name": "imu_accel_gyro", "l2": 17.080647, "mean_abs": 0.945007, "relative": 0.92156 }, { "name": "depth_confidence", "l2": 11.475858, "mean_abs": 0.209373, "relative": 0.000381 }, { "name": "video_fisheye_cam0", "l2": 5.270891, "mean_abs": 0.108931, "relative": 0.954446 }, { "name": "video_fisheye_cam1", "l2": 5.529171, "mean_abs": 0.115814, "relative": 0.913649 }, { "name": "video_fisheye_cam2", "l2": 5.828177, "mean_abs": 0.118339, "relative": 0.930382 }, { "name": "video_fisheye_cam3", "l2": 5.672506, "mean_abs": 0.11341, "relative": 0.949793 }, { "name": "video_stereo_left", "l2": 5.326203, "mean_abs": 0.115341, "relative": 0.865519 }, { "name": "video_stereo_right", "l2": 6.256821, "mean_abs": 0.137659, "relative": 0.88208 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.406888, "mean_abs": 0.098847, "relative": 0.861389 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 510, "start_frame": 2550, "end_frame": 2569, "center_frame": 2559, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.120636, "mean_abs": 0.207588, "relative": 0.853719 }, { "name": "hand_right_joints", "l2": 10.621666, "mean_abs": 0.247874, "relative": 0.837271 }, { "name": "body_joints", "l2": 5.381475, "mean_abs": 0.079533, "relative": 0.726871 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.043189, "mean_abs": 0.004796, "relative": 0.101919 }, { "name": "camera_rotation_matrix", "l2": 2.999703, "mean_abs": 0.211237, "relative": 0.92116 }, { "name": "imu_accel_gyro", "l2": 17.098583, "mean_abs": 0.952149, "relative": 0.922527 }, { "name": "depth_confidence", "l2": 11.478657, "mean_abs": 0.208835, "relative": 0.000381 }, { "name": "video_fisheye_cam0", "l2": 5.278915, "mean_abs": 0.109255, "relative": 0.955899 }, { "name": "video_fisheye_cam1", "l2": 5.553591, "mean_abs": 0.116564, "relative": 0.917684 }, { "name": "video_fisheye_cam2", "l2": 5.831504, "mean_abs": 0.118615, "relative": 0.930913 }, { "name": "video_fisheye_cam3", "l2": 5.671795, "mean_abs": 0.113445, "relative": 0.949674 }, { "name": "video_stereo_left", "l2": 5.334732, "mean_abs": 0.115638, "relative": 0.866905 }, { "name": "video_stereo_right", "l2": 6.285494, "mean_abs": 0.13839, "relative": 0.886123 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.245373, "mean_abs": 0.090288, "relative": 0.803585 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 511, "start_frame": 2555, "end_frame": 2574, "center_frame": 2564, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.130676, "mean_abs": 0.207975, "relative": 0.854566 }, { "name": "hand_right_joints", "l2": 10.62022, "mean_abs": 0.248177, "relative": 0.837157 }, { "name": "body_joints", "l2": 5.406235, "mean_abs": 0.080105, "relative": 0.730216 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.051922, "mean_abs": 0.005751, "relative": 0.12253 }, { "name": "camera_rotation_matrix", "l2": 3.0008, "mean_abs": 0.212459, "relative": 0.921497 }, { "name": "imu_accel_gyro", "l2": 17.090158, "mean_abs": 0.951328, "relative": 0.922073 }, { "name": "depth_confidence", "l2": 12.031081, "mean_abs": 0.21384, "relative": 0.000399 }, { "name": "video_fisheye_cam0", "l2": 5.284187, "mean_abs": 0.109561, "relative": 0.956853 }, { "name": "video_fisheye_cam1", "l2": 5.575504, "mean_abs": 0.117207, "relative": 0.921305 }, { "name": "video_fisheye_cam2", "l2": 5.83726, "mean_abs": 0.118921, "relative": 0.931832 }, { "name": "video_fisheye_cam3", "l2": 5.682544, "mean_abs": 0.114046, "relative": 0.951473 }, { "name": "video_stereo_left", "l2": 5.343528, "mean_abs": 0.116314, "relative": 0.868334 }, { "name": "video_stereo_right", "l2": 6.313778, "mean_abs": 0.139721, "relative": 0.89011 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.263438, "mean_abs": 0.089513, "relative": 0.810051 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 512, "start_frame": 2560, "end_frame": 2579, "center_frame": 2569, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.12751, "mean_abs": 0.208447, "relative": 0.854299 }, { "name": "hand_right_joints", "l2": 10.617238, "mean_abs": 0.247975, "relative": 0.836922 }, { "name": "body_joints", "l2": 5.429792, "mean_abs": 0.080547, "relative": 0.733398 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.056248, "mean_abs": 0.006726, "relative": 0.132737 }, { "name": "camera_rotation_matrix", "l2": 3.001923, "mean_abs": 0.213007, "relative": 0.921842 }, { "name": "imu_accel_gyro", "l2": 17.09099, "mean_abs": 0.954996, "relative": 0.922118 }, { "name": "depth_confidence", "l2": 12.012136, "mean_abs": 0.213791, "relative": 0.000399 }, { "name": "video_fisheye_cam0", "l2": 5.284009, "mean_abs": 0.109584, "relative": 0.956821 }, { "name": "video_fisheye_cam1", "l2": 5.599956, "mean_abs": 0.117873, "relative": 0.925345 }, { "name": "video_fisheye_cam2", "l2": 5.838955, "mean_abs": 0.119167, "relative": 0.932102 }, { "name": "video_fisheye_cam3", "l2": 5.694652, "mean_abs": 0.114679, "relative": 0.953501 }, { "name": "video_stereo_left", "l2": 5.350111, "mean_abs": 0.1168, "relative": 0.869404 }, { "name": "video_stereo_right", "l2": 6.340661, "mean_abs": 0.140736, "relative": 0.8939 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.196449, "mean_abs": 0.087166, "relative": 0.786076 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 513, "start_frame": 2565, "end_frame": 2584, "center_frame": 2574, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.118966, "mean_abs": 0.207671, "relative": 0.853578 }, { "name": "hand_right_joints", "l2": 10.612978, "mean_abs": 0.246987, "relative": 0.836587 }, { "name": "body_joints", "l2": 5.438821, "mean_abs": 0.080014, "relative": 0.734617 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.051286, "mean_abs": 0.006636, "relative": 0.121027 }, { "name": "camera_rotation_matrix", "l2": 3.001154, "mean_abs": 0.213887, "relative": 0.921606 }, { "name": "imu_accel_gyro", "l2": 17.071686, "mean_abs": 0.934577, "relative": 0.921076 }, { "name": "depth_confidence", "l2": 12.125335, "mean_abs": 0.215174, "relative": 0.000402 }, { "name": "video_fisheye_cam0", "l2": 5.282621, "mean_abs": 0.109566, "relative": 0.95657 }, { "name": "video_fisheye_cam1", "l2": 5.624093, "mean_abs": 0.118351, "relative": 0.929334 }, { "name": "video_fisheye_cam2", "l2": 5.82444, "mean_abs": 0.119206, "relative": 0.929785 }, { "name": "video_fisheye_cam3", "l2": 5.678018, "mean_abs": 0.114601, "relative": 0.950716 }, { "name": "video_stereo_left", "l2": 5.341852, "mean_abs": 0.116869, "relative": 0.868062 }, { "name": "video_stereo_right", "l2": 6.359255, "mean_abs": 0.141512, "relative": 0.896522 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.441786, "mean_abs": 0.09849, "relative": 0.873879 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 514, "start_frame": 2570, "end_frame": 2589, "center_frame": 2579, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.128588, "mean_abs": 0.210327, "relative": 0.85439 }, { "name": "hand_right_joints", "l2": 10.589688, "mean_abs": 0.249484, "relative": 0.834751 }, { "name": "body_joints", "l2": 5.418927, "mean_abs": 0.080454, "relative": 0.73193 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.058863, "mean_abs": 0.008201, "relative": 0.138908 }, { "name": "camera_rotation_matrix", "l2": 3.003709, "mean_abs": 0.219313, "relative": 0.92239 }, { "name": "imu_accel_gyro", "l2": 17.070967, "mean_abs": 0.969292, "relative": 0.921037 }, { "name": "depth_confidence", "l2": 12.261752, "mean_abs": 0.221017, "relative": 0.000407 }, { "name": "video_fisheye_cam0", "l2": 5.28202, "mean_abs": 0.110217, "relative": 0.956461 }, { "name": "video_fisheye_cam1", "l2": 5.670577, "mean_abs": 0.12073, "relative": 0.937015 }, { "name": "video_fisheye_cam2", "l2": 5.823143, "mean_abs": 0.120431, "relative": 0.929578 }, { "name": "video_fisheye_cam3", "l2": 5.667325, "mean_abs": 0.114909, "relative": 0.948925 }, { "name": "video_stereo_left", "l2": 5.338916, "mean_abs": 0.119958, "relative": 0.867585 }, { "name": "video_stereo_right", "l2": 6.390554, "mean_abs": 0.145413, "relative": 0.900934 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.170051, "mean_abs": 0.082326, "relative": 0.776629 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 515, "start_frame": 2575, "end_frame": 2594, "center_frame": 2584, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.137355, "mean_abs": 0.217033, "relative": 0.855129 }, { "name": "hand_right_joints", "l2": 10.592965, "mean_abs": 0.256426, "relative": 0.835009 }, { "name": "body_joints", "l2": 5.407574, "mean_abs": 0.081425, "relative": 0.730397 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.072998, "mean_abs": 0.010232, "relative": 0.172266 }, { "name": "camera_rotation_matrix", "l2": 3.012364, "mean_abs": 0.227452, "relative": 0.925048 }, { "name": "imu_accel_gyro", "l2": 17.159472, "mean_abs": 0.987426, "relative": 0.925813 }, { "name": "depth_confidence", "l2": 12.438596, "mean_abs": 0.228819, "relative": 0.000413 }, { "name": "video_fisheye_cam0", "l2": 5.276768, "mean_abs": 0.111058, "relative": 0.95551 }, { "name": "video_fisheye_cam1", "l2": 5.716407, "mean_abs": 0.124096, "relative": 0.944588 }, { "name": "video_fisheye_cam2", "l2": 5.842598, "mean_abs": 0.122677, "relative": 0.932684 }, { "name": "video_fisheye_cam3", "l2": 5.645267, "mean_abs": 0.115633, "relative": 0.945232 }, { "name": "video_stereo_left", "l2": 5.349849, "mean_abs": 0.123163, "relative": 0.869361 }, { "name": "video_stereo_right", "l2": 6.401025, "mean_abs": 0.148628, "relative": 0.90241 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.037086, "mean_abs": 0.077218, "relative": 0.729043 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 516, "start_frame": 2580, "end_frame": 2599, "center_frame": 2589, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.141949, "mean_abs": 0.21904, "relative": 0.855517 }, { "name": "hand_right_joints", "l2": 10.58445, "mean_abs": 0.257991, "relative": 0.834338 }, { "name": "body_joints", "l2": 5.398247, "mean_abs": 0.081507, "relative": 0.729137 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.082584, "mean_abs": 0.011681, "relative": 0.194885 }, { "name": "camera_rotation_matrix", "l2": 3.015472, "mean_abs": 0.230377, "relative": 0.926003 }, { "name": "imu_accel_gyro", "l2": 17.305634, "mean_abs": 0.989468, "relative": 0.933699 }, { "name": "depth_confidence", "l2": 12.530097, "mean_abs": 0.232155, "relative": 0.000416 }, { "name": "video_fisheye_cam0", "l2": 5.269701, "mean_abs": 0.111124, "relative": 0.95423 }, { "name": "video_fisheye_cam1", "l2": 5.749249, "mean_abs": 0.125534, "relative": 0.950015 }, { "name": "video_fisheye_cam2", "l2": 5.845875, "mean_abs": 0.123383, "relative": 0.933207 }, { "name": "video_fisheye_cam3", "l2": 5.616382, "mean_abs": 0.11543, "relative": 0.940395 }, { "name": "video_stereo_left", "l2": 5.345094, "mean_abs": 0.12394, "relative": 0.868588 }, { "name": "video_stereo_right", "l2": 6.392759, "mean_abs": 0.150055, "relative": 0.901245 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.075286, "mean_abs": 0.077517, "relative": 0.742714 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 517, "start_frame": 2585, "end_frame": 2604, "center_frame": 2594, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.134659, "mean_abs": 0.218695, "relative": 0.854902 }, { "name": "hand_right_joints", "l2": 10.568542, "mean_abs": 0.256671, "relative": 0.833084 }, { "name": "body_joints", "l2": 5.381266, "mean_abs": 0.080745, "relative": 0.726843 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.08526, "mean_abs": 0.012022, "relative": 0.201201 }, { "name": "camera_rotation_matrix", "l2": 3.01524, "mean_abs": 0.231271, "relative": 0.925931 }, { "name": "imu_accel_gyro", "l2": 17.303612, "mean_abs": 0.965671, "relative": 0.933589 }, { "name": "depth_confidence", "l2": 12.253923, "mean_abs": 0.231167, "relative": 0.000407 }, { "name": "video_fisheye_cam0", "l2": 5.263812, "mean_abs": 0.110917, "relative": 0.953164 }, { "name": "video_fisheye_cam1", "l2": 5.779546, "mean_abs": 0.126067, "relative": 0.955021 }, { "name": "video_fisheye_cam2", "l2": 5.858764, "mean_abs": 0.123439, "relative": 0.935265 }, { "name": "video_fisheye_cam3", "l2": 5.619167, "mean_abs": 0.115638, "relative": 0.940862 }, { "name": "video_stereo_left", "l2": 5.339374, "mean_abs": 0.123374, "relative": 0.867659 }, { "name": "video_stereo_right", "l2": 6.382557, "mean_abs": 0.14976, "relative": 0.899807 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.077517, "mean_abs": 0.079183, "relative": 0.743512 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 518, "start_frame": 2590, "end_frame": 2609, "center_frame": 2599, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.105964, "mean_abs": 0.214686, "relative": 0.852481 }, { "name": "hand_right_joints", "l2": 10.545696, "mean_abs": 0.253224, "relative": 0.831283 }, { "name": "body_joints", "l2": 5.366399, "mean_abs": 0.079647, "relative": 0.724835 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.065623, "mean_abs": 0.009068, "relative": 0.154862 }, { "name": "camera_rotation_matrix", "l2": 3.006659, "mean_abs": 0.227567, "relative": 0.923296 }, { "name": "imu_accel_gyro", "l2": 17.232916, "mean_abs": 0.91336, "relative": 0.929775 }, { "name": "depth_confidence", "l2": 12.423309, "mean_abs": 0.230706, "relative": 0.000412 }, { "name": "video_fisheye_cam0", "l2": 5.248086, "mean_abs": 0.109605, "relative": 0.950316 }, { "name": "video_fisheye_cam1", "l2": 5.7845, "mean_abs": 0.124392, "relative": 0.95584 }, { "name": "video_fisheye_cam2", "l2": 5.849712, "mean_abs": 0.121459, "relative": 0.933819 }, { "name": "video_fisheye_cam3", "l2": 5.600397, "mean_abs": 0.114488, "relative": 0.937719 }, { "name": "video_stereo_left", "l2": 5.27562, "mean_abs": 0.119942, "relative": 0.857299 }, { "name": "video_stereo_right", "l2": 6.272019, "mean_abs": 0.144535, "relative": 0.884223 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.056051, "mean_abs": 0.074778, "relative": 0.73583 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 519, "start_frame": 2595, "end_frame": 2614, "center_frame": 2604, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.086268, "mean_abs": 0.212195, "relative": 0.85082 }, { "name": "hand_right_joints", "l2": 10.515597, "mean_abs": 0.251666, "relative": 0.82891 }, { "name": "body_joints", "l2": 5.374482, "mean_abs": 0.080064, "relative": 0.725927 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.049798, "mean_abs": 0.006601, "relative": 0.117516 }, { "name": "camera_rotation_matrix", "l2": 3.001045, "mean_abs": 0.223666, "relative": 0.921572 }, { "name": "imu_accel_gyro", "l2": 17.143221, "mean_abs": 0.824679, "relative": 0.924936 }, { "name": "depth_confidence", "l2": 12.527545, "mean_abs": 0.22931, "relative": 0.000416 }, { "name": "video_fisheye_cam0", "l2": 5.247807, "mean_abs": 0.108809, "relative": 0.950266 }, { "name": "video_fisheye_cam1", "l2": 5.785913, "mean_abs": 0.123283, "relative": 0.956073 }, { "name": "video_fisheye_cam2", "l2": 5.854506, "mean_abs": 0.120009, "relative": 0.934585 }, { "name": "video_fisheye_cam3", "l2": 5.598161, "mean_abs": 0.11381, "relative": 0.937344 }, { "name": "video_stereo_left", "l2": 5.248856, "mean_abs": 0.116933, "relative": 0.85295 }, { "name": "video_stereo_right", "l2": 6.172813, "mean_abs": 0.139305, "relative": 0.870237 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.033725, "mean_abs": 0.078526, "relative": 0.72784 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 520, "start_frame": 2600, "end_frame": 2619, "center_frame": 2609, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.081683, "mean_abs": 0.21202, "relative": 0.850433 }, { "name": "hand_right_joints", "l2": 10.501872, "mean_abs": 0.251196, "relative": 0.827828 }, { "name": "body_joints", "l2": 5.384513, "mean_abs": 0.080422, "relative": 0.727282 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.044875, "mean_abs": 0.006288, "relative": 0.105899 }, { "name": "camera_rotation_matrix", "l2": 2.999876, "mean_abs": 0.222913, "relative": 0.921213 }, { "name": "imu_accel_gyro", "l2": 17.149624, "mean_abs": 0.806221, "relative": 0.925281 }, { "name": "depth_confidence", "l2": 12.549476, "mean_abs": 0.229782, "relative": 0.000416 }, { "name": "video_fisheye_cam0", "l2": 5.253472, "mean_abs": 0.109031, "relative": 0.951292 }, { "name": "video_fisheye_cam1", "l2": 5.781544, "mean_abs": 0.122974, "relative": 0.955351 }, { "name": "video_fisheye_cam2", "l2": 5.861107, "mean_abs": 0.12003, "relative": 0.935639 }, { "name": "video_fisheye_cam3", "l2": 5.60206, "mean_abs": 0.11417, "relative": 0.937997 }, { "name": "video_stereo_left", "l2": 5.245856, "mean_abs": 0.116545, "relative": 0.852462 }, { "name": "video_stereo_right", "l2": 6.121415, "mean_abs": 0.137369, "relative": 0.862991 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.115308, "mean_abs": 0.08127, "relative": 0.757037 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 521, "start_frame": 2605, "end_frame": 2624, "center_frame": 2614, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.084435, "mean_abs": 0.211483, "relative": 0.850665 }, { "name": "hand_right_joints", "l2": 10.488429, "mean_abs": 0.250917, "relative": 0.826769 }, { "name": "body_joints", "l2": 5.391517, "mean_abs": 0.08081, "relative": 0.728228 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.040984, "mean_abs": 0.005649, "relative": 0.096715 }, { "name": "camera_rotation_matrix", "l2": 3.000006, "mean_abs": 0.22189, "relative": 0.921253 }, { "name": "imu_accel_gyro", "l2": 17.147694, "mean_abs": 0.799823, "relative": 0.925177 }, { "name": "depth_confidence", "l2": 12.605762, "mean_abs": 0.229948, "relative": 0.000418 }, { "name": "video_fisheye_cam0", "l2": 5.258377, "mean_abs": 0.108898, "relative": 0.95218 }, { "name": "video_fisheye_cam1", "l2": 5.779366, "mean_abs": 0.122944, "relative": 0.954991 }, { "name": "video_fisheye_cam2", "l2": 5.858707, "mean_abs": 0.119724, "relative": 0.935255 }, { "name": "video_fisheye_cam3", "l2": 5.609381, "mean_abs": 0.114206, "relative": 0.939223 }, { "name": "video_stereo_left", "l2": 5.237389, "mean_abs": 0.115794, "relative": 0.851086 }, { "name": "video_stereo_right", "l2": 6.087395, "mean_abs": 0.136705, "relative": 0.858195 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.047624, "mean_abs": 0.07961, "relative": 0.732814 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 522, "start_frame": 2610, "end_frame": 2629, "center_frame": 2619, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.07556, "mean_abs": 0.208854, "relative": 0.849916 }, { "name": "hand_right_joints", "l2": 10.463374, "mean_abs": 0.248259, "relative": 0.824794 }, { "name": "body_joints", "l2": 5.403267, "mean_abs": 0.080734, "relative": 0.729815 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.031848, "mean_abs": 0.004288, "relative": 0.075157 }, { "name": "camera_rotation_matrix", "l2": 3.000171, "mean_abs": 0.219793, "relative": 0.921304 }, { "name": "imu_accel_gyro", "l2": 17.117388, "mean_abs": 0.797473, "relative": 0.923542 }, { "name": "depth_confidence", "l2": 12.646764, "mean_abs": 0.229788, "relative": 0.00042 }, { "name": "video_fisheye_cam0", "l2": 5.264715, "mean_abs": 0.108546, "relative": 0.953327 }, { "name": "video_fisheye_cam1", "l2": 5.753928, "mean_abs": 0.121576, "relative": 0.950788 }, { "name": "video_fisheye_cam2", "l2": 5.854142, "mean_abs": 0.119113, "relative": 0.934527 }, { "name": "video_fisheye_cam3", "l2": 5.602935, "mean_abs": 0.113385, "relative": 0.938144 }, { "name": "video_stereo_left", "l2": 5.22928, "mean_abs": 0.114563, "relative": 0.849768 }, { "name": "video_stereo_right", "l2": 6.036097, "mean_abs": 0.134291, "relative": 0.850963 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.966009, "mean_abs": 0.075866, "relative": 0.703605 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 523, "start_frame": 2615, "end_frame": 2634, "center_frame": 2624, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.06802, "mean_abs": 0.206876, "relative": 0.84928 }, { "name": "hand_right_joints", "l2": 10.454199, "mean_abs": 0.246692, "relative": 0.824071 }, { "name": "body_joints", "l2": 5.415735, "mean_abs": 0.081091, "relative": 0.731499 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01798, "mean_abs": 0.002773, "relative": 0.042431 }, { "name": "camera_rotation_matrix", "l2": 2.999829, "mean_abs": 0.220355, "relative": 0.921199 }, { "name": "imu_accel_gyro", "l2": 17.130959, "mean_abs": 0.803135, "relative": 0.924274 }, { "name": "depth_confidence", "l2": 12.65946, "mean_abs": 0.230692, "relative": 0.00042 }, { "name": "video_fisheye_cam0", "l2": 5.27897, "mean_abs": 0.109011, "relative": 0.955909 }, { "name": "video_fisheye_cam1", "l2": 5.720422, "mean_abs": 0.120872, "relative": 0.945251 }, { "name": "video_fisheye_cam2", "l2": 5.862103, "mean_abs": 0.119486, "relative": 0.935798 }, { "name": "video_fisheye_cam3", "l2": 5.623372, "mean_abs": 0.113727, "relative": 0.941566 }, { "name": "video_stereo_left", "l2": 5.242647, "mean_abs": 0.115495, "relative": 0.851941 }, { "name": "video_stereo_right", "l2": 6.005013, "mean_abs": 0.134013, "relative": 0.846581 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.113394, "mean_abs": 0.081996, "relative": 0.756352 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 524, "start_frame": 2620, "end_frame": 2639, "center_frame": 2629, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.06646, "mean_abs": 0.20615, "relative": 0.849149 }, { "name": "hand_right_joints", "l2": 10.46921, "mean_abs": 0.245974, "relative": 0.825254 }, { "name": "body_joints", "l2": 5.431415, "mean_abs": 0.081592, "relative": 0.733617 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01884, "mean_abs": 0.002647, "relative": 0.04446 }, { "name": "camera_rotation_matrix", "l2": 3.001035, "mean_abs": 0.222321, "relative": 0.921569 }, { "name": "imu_accel_gyro", "l2": 17.138502, "mean_abs": 0.823409, "relative": 0.924681 }, { "name": "depth_confidence", "l2": 12.634106, "mean_abs": 0.231574, "relative": 0.000419 }, { "name": "video_fisheye_cam0", "l2": 5.290945, "mean_abs": 0.109612, "relative": 0.958077 }, { "name": "video_fisheye_cam1", "l2": 5.69294, "mean_abs": 0.121233, "relative": 0.94071 }, { "name": "video_fisheye_cam2", "l2": 5.862863, "mean_abs": 0.12015, "relative": 0.935919 }, { "name": "video_fisheye_cam3", "l2": 5.623573, "mean_abs": 0.114545, "relative": 0.941599 }, { "name": "video_stereo_left", "l2": 5.25563, "mean_abs": 0.117259, "relative": 0.85405 }, { "name": "video_stereo_right", "l2": 6.002517, "mean_abs": 0.13527, "relative": 0.846229 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.108515, "mean_abs": 0.082326, "relative": 0.754606 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 525, "start_frame": 2625, "end_frame": 2644, "center_frame": 2634, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.069351, "mean_abs": 0.206318, "relative": 0.849393 }, { "name": "hand_right_joints", "l2": 10.497635, "mean_abs": 0.248242, "relative": 0.827494 }, { "name": "body_joints", "l2": 5.445682, "mean_abs": 0.081743, "relative": 0.735544 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.039546, "mean_abs": 0.005181, "relative": 0.093323 }, { "name": "camera_rotation_matrix", "l2": 3.006209, "mean_abs": 0.225248, "relative": 0.923158 }, { "name": "imu_accel_gyro", "l2": 17.191214, "mean_abs": 0.883511, "relative": 0.927525 }, { "name": "depth_confidence", "l2": 12.573387, "mean_abs": 0.233613, "relative": 0.000417 }, { "name": "video_fisheye_cam0", "l2": 5.301874, "mean_abs": 0.110349, "relative": 0.960056 }, { "name": "video_fisheye_cam1", "l2": 5.63411, "mean_abs": 0.121299, "relative": 0.930989 }, { "name": "video_fisheye_cam2", "l2": 5.850559, "mean_abs": 0.120959, "relative": 0.933955 }, { "name": "video_fisheye_cam3", "l2": 5.595694, "mean_abs": 0.115141, "relative": 0.936931 }, { "name": "video_stereo_left", "l2": 5.261498, "mean_abs": 0.119346, "relative": 0.855004 }, { "name": "video_stereo_right", "l2": 5.991392, "mean_abs": 0.137192, "relative": 0.844661 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.125486, "mean_abs": 0.084506, "relative": 0.76068 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 526, "start_frame": 2630, "end_frame": 2649, "center_frame": 2639, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.091657, "mean_abs": 0.206872, "relative": 0.851274 }, { "name": "hand_right_joints", "l2": 10.498424, "mean_abs": 0.248798, "relative": 0.827557 }, { "name": "body_joints", "l2": 5.441505, "mean_abs": 0.08157, "relative": 0.73498 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.062973, "mean_abs": 0.008792, "relative": 0.148607 }, { "name": "camera_rotation_matrix", "l2": 3.005827, "mean_abs": 0.226133, "relative": 0.923041 }, { "name": "imu_accel_gyro", "l2": 17.128139, "mean_abs": 0.875782, "relative": 0.924122 }, { "name": "depth_confidence", "l2": 12.476117, "mean_abs": 0.232181, "relative": 0.000414 }, { "name": "video_fisheye_cam0", "l2": 5.312643, "mean_abs": 0.110763, "relative": 0.962006 }, { "name": "video_fisheye_cam1", "l2": 5.610084, "mean_abs": 0.121298, "relative": 0.927019 }, { "name": "video_fisheye_cam2", "l2": 5.848057, "mean_abs": 0.12116, "relative": 0.933555 }, { "name": "video_fisheye_cam3", "l2": 5.597124, "mean_abs": 0.115665, "relative": 0.937171 }, { "name": "video_stereo_left", "l2": 5.266028, "mean_abs": 0.11959, "relative": 0.85574 }, { "name": "video_stereo_right", "l2": 6.001844, "mean_abs": 0.137823, "relative": 0.846134 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.013941, "mean_abs": 0.079812, "relative": 0.720759 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 527, "start_frame": 2635, "end_frame": 2654, "center_frame": 2644, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.108666, "mean_abs": 0.205528, "relative": 0.852709 }, { "name": "hand_right_joints", "l2": 10.485238, "mean_abs": 0.246777, "relative": 0.826517 }, { "name": "body_joints", "l2": 5.436527, "mean_abs": 0.081809, "relative": 0.734307 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.088743, "mean_abs": 0.012466, "relative": 0.209421 }, { "name": "camera_rotation_matrix", "l2": 3.003846, "mean_abs": 0.224445, "relative": 0.922432 }, { "name": "imu_accel_gyro", "l2": 17.118073, "mean_abs": 0.879657, "relative": 0.923579 }, { "name": "depth_confidence", "l2": 12.364914, "mean_abs": 0.229704, "relative": 0.00041 }, { "name": "video_fisheye_cam0", "l2": 5.321686, "mean_abs": 0.110859, "relative": 0.963643 }, { "name": "video_fisheye_cam1", "l2": 5.596018, "mean_abs": 0.120856, "relative": 0.924695 }, { "name": "video_fisheye_cam2", "l2": 5.843752, "mean_abs": 0.120925, "relative": 0.932868 }, { "name": "video_fisheye_cam3", "l2": 5.619262, "mean_abs": 0.116227, "relative": 0.940878 }, { "name": "video_stereo_left", "l2": 5.275175, "mean_abs": 0.118942, "relative": 0.857226 }, { "name": "video_stereo_right", "l2": 6.011709, "mean_abs": 0.137583, "relative": 0.847525 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.036159, "mean_abs": 0.081967, "relative": 0.728711 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 528, "start_frame": 2640, "end_frame": 2659, "center_frame": 2649, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.122753, "mean_abs": 0.202647, "relative": 0.853897 }, { "name": "hand_right_joints", "l2": 10.479954, "mean_abs": 0.24475, "relative": 0.826101 }, { "name": "body_joints", "l2": 5.429231, "mean_abs": 0.081588, "relative": 0.733322 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.085453, "mean_abs": 0.012173, "relative": 0.201658 }, { "name": "camera_rotation_matrix", "l2": 3.002542, "mean_abs": 0.220929, "relative": 0.922032 }, { "name": "imu_accel_gyro", "l2": 17.108397, "mean_abs": 0.879954, "relative": 0.923057 }, { "name": "depth_confidence", "l2": 12.313774, "mean_abs": 0.226976, "relative": 0.000409 }, { "name": "video_fisheye_cam0", "l2": 5.32669, "mean_abs": 0.110569, "relative": 0.96455 }, { "name": "video_fisheye_cam1", "l2": 5.567435, "mean_abs": 0.119461, "relative": 0.919972 }, { "name": "video_fisheye_cam2", "l2": 5.830728, "mean_abs": 0.120339, "relative": 0.930789 }, { "name": "video_fisheye_cam3", "l2": 5.594979, "mean_abs": 0.115304, "relative": 0.936812 }, { "name": "video_stereo_left", "l2": 5.267714, "mean_abs": 0.117309, "relative": 0.856014 }, { "name": "video_stereo_right", "l2": 6.005825, "mean_abs": 0.135729, "relative": 0.846695 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.045457, "mean_abs": 0.08057, "relative": 0.732038 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 529, "start_frame": 2645, "end_frame": 2664, "center_frame": 2654, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.133097, "mean_abs": 0.199596, "relative": 0.85477 }, { "name": "hand_right_joints", "l2": 10.457232, "mean_abs": 0.242052, "relative": 0.82431 }, { "name": "body_joints", "l2": 5.421045, "mean_abs": 0.081519, "relative": 0.732216 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.063998, "mean_abs": 0.008969, "relative": 0.151025 }, { "name": "camera_rotation_matrix", "l2": 3.000109, "mean_abs": 0.214302, "relative": 0.921285 }, { "name": "imu_accel_gyro", "l2": 17.136667, "mean_abs": 0.876236, "relative": 0.924582 }, { "name": "depth_confidence", "l2": 12.26232, "mean_abs": 0.222775, "relative": 0.000407 }, { "name": "video_fisheye_cam0", "l2": 5.312865, "mean_abs": 0.110132, "relative": 0.962046 }, { "name": "video_fisheye_cam1", "l2": 5.545299, "mean_abs": 0.117535, "relative": 0.916314 }, { "name": "video_fisheye_cam2", "l2": 5.815644, "mean_abs": 0.118943, "relative": 0.928381 }, { "name": "video_fisheye_cam3", "l2": 5.612772, "mean_abs": 0.114527, "relative": 0.939791 }, { "name": "video_stereo_left", "l2": 5.266923, "mean_abs": 0.115337, "relative": 0.855885 }, { "name": "video_stereo_right", "l2": 6.000803, "mean_abs": 0.13305, "relative": 0.845987 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.103401, "mean_abs": 0.085482, "relative": 0.752776 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 530, "start_frame": 2650, "end_frame": 2669, "center_frame": 2659, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.150966, "mean_abs": 0.197848, "relative": 0.856277 }, { "name": "hand_right_joints", "l2": 10.421474, "mean_abs": 0.238996, "relative": 0.821491 }, { "name": "body_joints", "l2": 5.420487, "mean_abs": 0.081231, "relative": 0.732141 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.034733, "mean_abs": 0.005219, "relative": 0.081966 }, { "name": "camera_rotation_matrix", "l2": 2.999692, "mean_abs": 0.21308, "relative": 0.921157 }, { "name": "imu_accel_gyro", "l2": 17.135406, "mean_abs": 0.861281, "relative": 0.924514 }, { "name": "depth_confidence", "l2": 12.266821, "mean_abs": 0.222274, "relative": 0.000407 }, { "name": "video_fisheye_cam0", "l2": 5.297508, "mean_abs": 0.110095, "relative": 0.959265 }, { "name": "video_fisheye_cam1", "l2": 5.543613, "mean_abs": 0.117085, "relative": 0.916035 }, { "name": "video_fisheye_cam2", "l2": 5.811813, "mean_abs": 0.11879, "relative": 0.92777 }, { "name": "video_fisheye_cam3", "l2": 5.612016, "mean_abs": 0.114274, "relative": 0.939664 }, { "name": "video_stereo_left", "l2": 5.269829, "mean_abs": 0.115154, "relative": 0.856358 }, { "name": "video_stereo_right", "l2": 6.009564, "mean_abs": 0.13272, "relative": 0.847223 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.103054, "mean_abs": 0.084202, "relative": 0.752651 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 531, "start_frame": 2655, "end_frame": 2674, "center_frame": 2664, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.175749, "mean_abs": 0.197437, "relative": 0.858368 }, { "name": "hand_right_joints", "l2": 10.417538, "mean_abs": 0.238858, "relative": 0.821181 }, { "name": "body_joints", "l2": 5.423027, "mean_abs": 0.080997, "relative": 0.732484 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012262, "mean_abs": 0.001814, "relative": 0.028938 }, { "name": "camera_rotation_matrix", "l2": 2.999484, "mean_abs": 0.212523, "relative": 0.921093 }, { "name": "imu_accel_gyro", "l2": 17.152258, "mean_abs": 0.876018, "relative": 0.925423 }, { "name": "depth_confidence", "l2": 12.276098, "mean_abs": 0.222168, "relative": 0.000407 }, { "name": "video_fisheye_cam0", "l2": 5.263772, "mean_abs": 0.109294, "relative": 0.953157 }, { "name": "video_fisheye_cam1", "l2": 5.531743, "mean_abs": 0.116593, "relative": 0.914074 }, { "name": "video_fisheye_cam2", "l2": 5.798147, "mean_abs": 0.118598, "relative": 0.925588 }, { "name": "video_fisheye_cam3", "l2": 5.591918, "mean_abs": 0.11359, "relative": 0.936299 }, { "name": "video_stereo_left", "l2": 5.256955, "mean_abs": 0.114888, "relative": 0.854266 }, { "name": "video_stereo_right", "l2": 5.991908, "mean_abs": 0.132128, "relative": 0.844733 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.068708, "mean_abs": 0.082492, "relative": 0.740359 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 532, "start_frame": 2660, "end_frame": 2679, "center_frame": 2669, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.17457, "mean_abs": 0.197599, "relative": 0.858268 }, { "name": "hand_right_joints", "l2": 10.419747, "mean_abs": 0.239158, "relative": 0.821355 }, { "name": "body_joints", "l2": 5.436693, "mean_abs": 0.08104, "relative": 0.73433 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01443, "mean_abs": 0.002054, "relative": 0.034052 }, { "name": "camera_rotation_matrix", "l2": 2.999854, "mean_abs": 0.212062, "relative": 0.921206 }, { "name": "imu_accel_gyro", "l2": 17.139429, "mean_abs": 0.86001, "relative": 0.924731 }, { "name": "depth_confidence", "l2": 12.259767, "mean_abs": 0.221691, "relative": 0.000407 }, { "name": "video_fisheye_cam0", "l2": 5.235858, "mean_abs": 0.108739, "relative": 0.948102 }, { "name": "video_fisheye_cam1", "l2": 5.501557, "mean_abs": 0.115743, "relative": 0.909086 }, { "name": "video_fisheye_cam2", "l2": 5.776081, "mean_abs": 0.118044, "relative": 0.922065 }, { "name": "video_fisheye_cam3", "l2": 5.541978, "mean_abs": 0.112572, "relative": 0.927937 }, { "name": "video_stereo_left", "l2": 5.23819, "mean_abs": 0.114141, "relative": 0.851216 }, { "name": "video_stereo_right", "l2": 5.963741, "mean_abs": 0.130946, "relative": 0.840762 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.060311, "mean_abs": 0.082625, "relative": 0.737354 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 533, "start_frame": 2665, "end_frame": 2684, "center_frame": 2674, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.170579, "mean_abs": 0.198747, "relative": 0.857932 }, { "name": "hand_right_joints", "l2": 10.43584, "mean_abs": 0.240436, "relative": 0.822623 }, { "name": "body_joints", "l2": 5.443841, "mean_abs": 0.081167, "relative": 0.735295 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.022492, "mean_abs": 0.003159, "relative": 0.053079 }, { "name": "camera_rotation_matrix", "l2": 2.999427, "mean_abs": 0.212744, "relative": 0.921075 }, { "name": "imu_accel_gyro", "l2": 17.119762, "mean_abs": 0.872893, "relative": 0.92367 }, { "name": "depth_confidence", "l2": 12.19486, "mean_abs": 0.221435, "relative": 0.000405 }, { "name": "video_fisheye_cam0", "l2": 5.194973, "mean_abs": 0.10761, "relative": 0.940699 }, { "name": "video_fisheye_cam1", "l2": 5.483466, "mean_abs": 0.115646, "relative": 0.906096 }, { "name": "video_fisheye_cam2", "l2": 5.75602, "mean_abs": 0.118121, "relative": 0.918863 }, { "name": "video_fisheye_cam3", "l2": 5.518592, "mean_abs": 0.112077, "relative": 0.924022 }, { "name": "video_stereo_left", "l2": 5.23386, "mean_abs": 0.114486, "relative": 0.850513 }, { "name": "video_stereo_right", "l2": 5.965539, "mean_abs": 0.131511, "relative": 0.841016 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.031112, "mean_abs": 0.084213, "relative": 0.726905 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 534, "start_frame": 2670, "end_frame": 2689, "center_frame": 2679, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.172927, "mean_abs": 0.200095, "relative": 0.85813 }, { "name": "hand_right_joints", "l2": 10.474648, "mean_abs": 0.243245, "relative": 0.825682 }, { "name": "body_joints", "l2": 5.457002, "mean_abs": 0.081917, "relative": 0.737073 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.024481, "mean_abs": 0.003336, "relative": 0.057773 }, { "name": "camera_rotation_matrix", "l2": 2.999888, "mean_abs": 0.212715, "relative": 0.921217 }, { "name": "imu_accel_gyro", "l2": 17.077057, "mean_abs": 0.867719, "relative": 0.921366 }, { "name": "depth_confidence", "l2": 12.18029, "mean_abs": 0.221835, "relative": 0.000404 }, { "name": "video_fisheye_cam0", "l2": 5.172132, "mean_abs": 0.107111, "relative": 0.936562 }, { "name": "video_fisheye_cam1", "l2": 5.468163, "mean_abs": 0.115366, "relative": 0.903568 }, { "name": "video_fisheye_cam2", "l2": 5.732662, "mean_abs": 0.117642, "relative": 0.915134 }, { "name": "video_fisheye_cam3", "l2": 5.532377, "mean_abs": 0.112427, "relative": 0.92633 }, { "name": "video_stereo_left", "l2": 5.226879, "mean_abs": 0.114107, "relative": 0.849378 }, { "name": "video_stereo_right", "l2": 5.961437, "mean_abs": 0.131312, "relative": 0.840438 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.068795, "mean_abs": 0.084433, "relative": 0.740391 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 535, "start_frame": 2675, "end_frame": 2694, "center_frame": 2684, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.163691, "mean_abs": 0.200585, "relative": 0.857351 }, { "name": "hand_right_joints", "l2": 10.478334, "mean_abs": 0.244055, "relative": 0.825973 }, { "name": "body_joints", "l2": 5.465932, "mean_abs": 0.082144, "relative": 0.738279 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018829, "mean_abs": 0.00274, "relative": 0.044434 }, { "name": "camera_rotation_matrix", "l2": 3.000284, "mean_abs": 0.212807, "relative": 0.921338 }, { "name": "imu_accel_gyro", "l2": 17.035204, "mean_abs": 0.866552, "relative": 0.919108 }, { "name": "depth_confidence", "l2": 12.160773, "mean_abs": 0.22109, "relative": 0.000404 }, { "name": "video_fisheye_cam0", "l2": 5.149597, "mean_abs": 0.106653, "relative": 0.932482 }, { "name": "video_fisheye_cam1", "l2": 5.456929, "mean_abs": 0.115295, "relative": 0.901711 }, { "name": "video_fisheye_cam2", "l2": 5.713995, "mean_abs": 0.117355, "relative": 0.912154 }, { "name": "video_fisheye_cam3", "l2": 5.536983, "mean_abs": 0.112545, "relative": 0.927101 }, { "name": "video_stereo_left", "l2": 5.222482, "mean_abs": 0.114191, "relative": 0.848664 }, { "name": "video_stereo_right", "l2": 5.95946, "mean_abs": 0.131195, "relative": 0.840159 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.064418, "mean_abs": 0.083322, "relative": 0.738824 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 536, "start_frame": 2680, "end_frame": 2699, "center_frame": 2689, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.165439, "mean_abs": 0.200776, "relative": 0.857498 }, { "name": "hand_right_joints", "l2": 10.478712, "mean_abs": 0.244208, "relative": 0.826003 }, { "name": "body_joints", "l2": 5.473245, "mean_abs": 0.08254, "relative": 0.739267 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015951, "mean_abs": 0.00243, "relative": 0.037642 }, { "name": "camera_rotation_matrix", "l2": 3.000999, "mean_abs": 0.212253, "relative": 0.921558 }, { "name": "imu_accel_gyro", "l2": 17.045076, "mean_abs": 0.874924, "relative": 0.919641 }, { "name": "depth_confidence", "l2": 12.134687, "mean_abs": 0.220156, "relative": 0.000403 }, { "name": "video_fisheye_cam0", "l2": 5.130357, "mean_abs": 0.106081, "relative": 0.928998 }, { "name": "video_fisheye_cam1", "l2": 5.444664, "mean_abs": 0.114847, "relative": 0.899685 }, { "name": "video_fisheye_cam2", "l2": 5.68999, "mean_abs": 0.116634, "relative": 0.908322 }, { "name": "video_fisheye_cam3", "l2": 5.545204, "mean_abs": 0.112455, "relative": 0.928478 }, { "name": "video_stereo_left", "l2": 5.213883, "mean_abs": 0.1137, "relative": 0.847266 }, { "name": "video_stereo_right", "l2": 5.959137, "mean_abs": 0.130914, "relative": 0.840113 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.176846, "mean_abs": 0.089642, "relative": 0.77906 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 537, "start_frame": 2685, "end_frame": 2704, "center_frame": 2694, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.152086, "mean_abs": 0.201414, "relative": 0.856372 }, { "name": "hand_right_joints", "l2": 10.493284, "mean_abs": 0.244739, "relative": 0.827152 }, { "name": "body_joints", "l2": 5.481464, "mean_abs": 0.082894, "relative": 0.740377 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012092, "mean_abs": 0.001653, "relative": 0.028535 }, { "name": "camera_rotation_matrix", "l2": 2.999881, "mean_abs": 0.211095, "relative": 0.921215 }, { "name": "imu_accel_gyro", "l2": 17.097479, "mean_abs": 0.869375, "relative": 0.922468 }, { "name": "depth_confidence", "l2": 12.134041, "mean_abs": 0.21962, "relative": 0.000403 }, { "name": "video_fisheye_cam0", "l2": 5.106544, "mean_abs": 0.105589, "relative": 0.924686 }, { "name": "video_fisheye_cam1", "l2": 5.427248, "mean_abs": 0.114173, "relative": 0.896807 }, { "name": "video_fisheye_cam2", "l2": 5.667025, "mean_abs": 0.115975, "relative": 0.904656 }, { "name": "video_fisheye_cam3", "l2": 5.538442, "mean_abs": 0.112318, "relative": 0.927345 }, { "name": "video_stereo_left", "l2": 5.205609, "mean_abs": 0.113044, "relative": 0.845922 }, { "name": "video_stereo_right", "l2": 5.953213, "mean_abs": 0.130328, "relative": 0.839278 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.16664, "mean_abs": 0.08873, "relative": 0.775408 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 538, "start_frame": 2690, "end_frame": 2709, "center_frame": 2699, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.143789, "mean_abs": 0.200967, "relative": 0.855672 }, { "name": "hand_right_joints", "l2": 10.492774, "mean_abs": 0.244458, "relative": 0.827111 }, { "name": "body_joints", "l2": 5.478229, "mean_abs": 0.083027, "relative": 0.73994 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010242, "mean_abs": 0.001377, "relative": 0.024169 }, { "name": "camera_rotation_matrix", "l2": 3.000417, "mean_abs": 0.210925, "relative": 0.92138 }, { "name": "imu_accel_gyro", "l2": 17.108202, "mean_abs": 0.867093, "relative": 0.923046 }, { "name": "depth_confidence", "l2": 12.15481, "mean_abs": 0.219639, "relative": 0.000403 }, { "name": "video_fisheye_cam0", "l2": 5.09111, "mean_abs": 0.105168, "relative": 0.921891 }, { "name": "video_fisheye_cam1", "l2": 5.435169, "mean_abs": 0.114601, "relative": 0.898116 }, { "name": "video_fisheye_cam2", "l2": 5.650772, "mean_abs": 0.115554, "relative": 0.902062 }, { "name": "video_fisheye_cam3", "l2": 5.535461, "mean_abs": 0.112187, "relative": 0.926846 }, { "name": "video_stereo_left", "l2": 5.197762, "mean_abs": 0.112996, "relative": 0.844647 }, { "name": "video_stereo_right", "l2": 5.967328, "mean_abs": 0.131237, "relative": 0.841268 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.184699, "mean_abs": 0.088527, "relative": 0.781871 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 539, "start_frame": 2695, "end_frame": 2714, "center_frame": 2704, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.140034, "mean_abs": 0.200961, "relative": 0.855355 }, { "name": "hand_right_joints", "l2": 10.48736, "mean_abs": 0.24422, "relative": 0.826685 }, { "name": "body_joints", "l2": 5.467224, "mean_abs": 0.082604, "relative": 0.738453 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01009, "mean_abs": 0.001418, "relative": 0.023811 }, { "name": "camera_rotation_matrix", "l2": 3.000353, "mean_abs": 0.21026, "relative": 0.92136 }, { "name": "imu_accel_gyro", "l2": 17.110228, "mean_abs": 0.871482, "relative": 0.923156 }, { "name": "depth_confidence", "l2": 12.146878, "mean_abs": 0.218923, "relative": 0.000403 }, { "name": "video_fisheye_cam0", "l2": 5.075786, "mean_abs": 0.104758, "relative": 0.919116 }, { "name": "video_fisheye_cam1", "l2": 5.436443, "mean_abs": 0.114515, "relative": 0.898326 }, { "name": "video_fisheye_cam2", "l2": 5.6368, "mean_abs": 0.115113, "relative": 0.899831 }, { "name": "video_fisheye_cam3", "l2": 5.533299, "mean_abs": 0.11227, "relative": 0.926484 }, { "name": "video_stereo_left", "l2": 5.187429, "mean_abs": 0.112525, "relative": 0.842968 }, { "name": "video_stereo_right", "l2": 5.96897, "mean_abs": 0.131091, "relative": 0.8415 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.161789, "mean_abs": 0.087806, "relative": 0.773672 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 540, "start_frame": 2700, "end_frame": 2719, "center_frame": 2709, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.12461, "mean_abs": 0.200108, "relative": 0.854054 }, { "name": "hand_right_joints", "l2": 10.496192, "mean_abs": 0.243999, "relative": 0.827381 }, { "name": "body_joints", "l2": 5.46877, "mean_abs": 0.082532, "relative": 0.738662 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008212, "mean_abs": 0.001224, "relative": 0.01938 }, { "name": "camera_rotation_matrix", "l2": 3.000076, "mean_abs": 0.21016, "relative": 0.921275 }, { "name": "imu_accel_gyro", "l2": 17.126337, "mean_abs": 0.868657, "relative": 0.924025 }, { "name": "depth_confidence", "l2": 12.146987, "mean_abs": 0.218872, "relative": 0.000403 }, { "name": "video_fisheye_cam0", "l2": 5.06481, "mean_abs": 0.104407, "relative": 0.917129 }, { "name": "video_fisheye_cam1", "l2": 5.440755, "mean_abs": 0.114625, "relative": 0.899039 }, { "name": "video_fisheye_cam2", "l2": 5.621791, "mean_abs": 0.114608, "relative": 0.897435 }, { "name": "video_fisheye_cam3", "l2": 5.526777, "mean_abs": 0.11236, "relative": 0.925392 }, { "name": "video_stereo_left", "l2": 5.182574, "mean_abs": 0.112364, "relative": 0.842179 }, { "name": "video_stereo_right", "l2": 5.979016, "mean_abs": 0.131436, "relative": 0.842916 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.196822, "mean_abs": 0.092374, "relative": 0.78621 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 541, "start_frame": 2705, "end_frame": 2724, "center_frame": 2714, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.144121, "mean_abs": 0.200774, "relative": 0.8557 }, { "name": "hand_right_joints", "l2": 10.536429, "mean_abs": 0.245827, "relative": 0.830553 }, { "name": "body_joints", "l2": 5.460062, "mean_abs": 0.08256, "relative": 0.737486 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007423, "mean_abs": 0.00117, "relative": 0.017517 }, { "name": "camera_rotation_matrix", "l2": 2.999737, "mean_abs": 0.211618, "relative": 0.92117 }, { "name": "imu_accel_gyro", "l2": 17.13699, "mean_abs": 0.878146, "relative": 0.9246 }, { "name": "depth_confidence", "l2": 12.085879, "mean_abs": 0.219266, "relative": 0.000401 }, { "name": "video_fisheye_cam0", "l2": 5.05532, "mean_abs": 0.104511, "relative": 0.91541 }, { "name": "video_fisheye_cam1", "l2": 5.438327, "mean_abs": 0.114826, "relative": 0.898638 }, { "name": "video_fisheye_cam2", "l2": 5.609719, "mean_abs": 0.114873, "relative": 0.895508 }, { "name": "video_fisheye_cam3", "l2": 5.524426, "mean_abs": 0.112559, "relative": 0.924998 }, { "name": "video_stereo_left", "l2": 5.180657, "mean_abs": 0.113395, "relative": 0.841867 }, { "name": "video_stereo_right", "l2": 5.985944, "mean_abs": 0.132054, "relative": 0.843893 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.107759, "mean_abs": 0.085489, "relative": 0.754335 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 542, "start_frame": 2710, "end_frame": 2729, "center_frame": 2719, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.14827, "mean_abs": 0.200465, "relative": 0.85605 }, { "name": "hand_right_joints", "l2": 10.545537, "mean_abs": 0.245217, "relative": 0.83127 }, { "name": "body_joints", "l2": 5.44836, "mean_abs": 0.082206, "relative": 0.735906 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009122, "mean_abs": 0.001354, "relative": 0.021527 }, { "name": "camera_rotation_matrix", "l2": 3.001227, "mean_abs": 0.21364, "relative": 0.921628 }, { "name": "imu_accel_gyro", "l2": 17.111319, "mean_abs": 0.883145, "relative": 0.923215 }, { "name": "depth_confidence", "l2": 12.025134, "mean_abs": 0.219027, "relative": 0.000399 }, { "name": "video_fisheye_cam0", "l2": 5.046962, "mean_abs": 0.104848, "relative": 0.913897 }, { "name": "video_fisheye_cam1", "l2": 5.448215, "mean_abs": 0.115545, "relative": 0.900271 }, { "name": "video_fisheye_cam2", "l2": 5.60992, "mean_abs": 0.115366, "relative": 0.89554 }, { "name": "video_fisheye_cam3", "l2": 5.522873, "mean_abs": 0.112592, "relative": 0.924738 }, { "name": "video_stereo_left", "l2": 5.195397, "mean_abs": 0.114536, "relative": 0.844262 }, { "name": "video_stereo_right", "l2": 6.020198, "mean_abs": 0.133691, "relative": 0.848722 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.130832, "mean_abs": 0.087807, "relative": 0.762593 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 543, "start_frame": 2715, "end_frame": 2734, "center_frame": 2724, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.145049, "mean_abs": 0.200396, "relative": 0.855778 }, { "name": "hand_right_joints", "l2": 10.552917, "mean_abs": 0.245304, "relative": 0.831852 }, { "name": "body_joints", "l2": 5.441441, "mean_abs": 0.082031, "relative": 0.734971 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011952, "mean_abs": 0.001866, "relative": 0.028206 }, { "name": "camera_rotation_matrix", "l2": 3.001393, "mean_abs": 0.213862, "relative": 0.921679 }, { "name": "imu_accel_gyro", "l2": 17.113586, "mean_abs": 0.883581, "relative": 0.923337 }, { "name": "depth_confidence", "l2": 11.976065, "mean_abs": 0.218536, "relative": 0.000397 }, { "name": "video_fisheye_cam0", "l2": 5.037519, "mean_abs": 0.104707, "relative": 0.912187 }, { "name": "video_fisheye_cam1", "l2": 5.43545, "mean_abs": 0.115283, "relative": 0.898162 }, { "name": "video_fisheye_cam2", "l2": 5.605121, "mean_abs": 0.115278, "relative": 0.894774 }, { "name": "video_fisheye_cam3", "l2": 5.504262, "mean_abs": 0.112244, "relative": 0.921622 }, { "name": "video_stereo_left", "l2": 5.193652, "mean_abs": 0.114388, "relative": 0.843979 }, { "name": "video_stereo_right", "l2": 6.02041, "mean_abs": 0.133643, "relative": 0.848751 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.162538, "mean_abs": 0.089305, "relative": 0.77394 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 544, "start_frame": 2720, "end_frame": 2739, "center_frame": 2729, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.146052, "mean_abs": 0.20052, "relative": 0.855863 }, { "name": "hand_right_joints", "l2": 10.57291, "mean_abs": 0.245559, "relative": 0.833428 }, { "name": "body_joints", "l2": 5.430772, "mean_abs": 0.08187, "relative": 0.73353 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013663, "mean_abs": 0.002118, "relative": 0.032242 }, { "name": "camera_rotation_matrix", "l2": 3.000796, "mean_abs": 0.213152, "relative": 0.921496 }, { "name": "imu_accel_gyro", "l2": 17.123018, "mean_abs": 0.889494, "relative": 0.923846 }, { "name": "depth_confidence", "l2": 11.921078, "mean_abs": 0.21734, "relative": 0.000396 }, { "name": "video_fisheye_cam0", "l2": 5.031394, "mean_abs": 0.104375, "relative": 0.911078 }, { "name": "video_fisheye_cam1", "l2": 5.425917, "mean_abs": 0.114576, "relative": 0.896587 }, { "name": "video_fisheye_cam2", "l2": 5.599174, "mean_abs": 0.114826, "relative": 0.893825 }, { "name": "video_fisheye_cam3", "l2": 5.501517, "mean_abs": 0.111673, "relative": 0.921163 }, { "name": "video_stereo_left", "l2": 5.196003, "mean_abs": 0.113705, "relative": 0.844361 }, { "name": "video_stereo_right", "l2": 6.023642, "mean_abs": 0.132828, "relative": 0.849207 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.11913, "mean_abs": 0.085369, "relative": 0.758405 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 545, "start_frame": 2725, "end_frame": 2744, "center_frame": 2734, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.147967, "mean_abs": 0.200818, "relative": 0.856024 }, { "name": "hand_right_joints", "l2": 10.575026, "mean_abs": 0.245485, "relative": 0.833595 }, { "name": "body_joints", "l2": 5.421442, "mean_abs": 0.081605, "relative": 0.73227 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015448, "mean_abs": 0.002096, "relative": 0.036456 }, { "name": "camera_rotation_matrix", "l2": 2.999957, "mean_abs": 0.211593, "relative": 0.921238 }, { "name": "imu_accel_gyro", "l2": 17.06497, "mean_abs": 0.885482, "relative": 0.920714 }, { "name": "depth_confidence", "l2": 11.910622, "mean_abs": 0.216297, "relative": 0.000395 }, { "name": "video_fisheye_cam0", "l2": 5.02541, "mean_abs": 0.103952, "relative": 0.909994 }, { "name": "video_fisheye_cam1", "l2": 5.408747, "mean_abs": 0.113575, "relative": 0.89375 }, { "name": "video_fisheye_cam2", "l2": 5.592907, "mean_abs": 0.114233, "relative": 0.892824 }, { "name": "video_fisheye_cam3", "l2": 5.497848, "mean_abs": 0.111276, "relative": 0.920548 }, { "name": "video_stereo_left", "l2": 5.195218, "mean_abs": 0.112518, "relative": 0.844233 }, { "name": "video_stereo_right", "l2": 6.018072, "mean_abs": 0.131395, "relative": 0.848422 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.264961, "mean_abs": 0.089937, "relative": 0.810595 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 546, "start_frame": 2730, "end_frame": 2749, "center_frame": 2739, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.145488, "mean_abs": 0.20054, "relative": 0.855815 }, { "name": "hand_right_joints", "l2": 10.568258, "mean_abs": 0.244469, "relative": 0.833061 }, { "name": "body_joints", "l2": 5.409637, "mean_abs": 0.081184, "relative": 0.730675 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015224, "mean_abs": 0.001973, "relative": 0.035928 }, { "name": "camera_rotation_matrix", "l2": 3.000168, "mean_abs": 0.210918, "relative": 0.921303 }, { "name": "imu_accel_gyro", "l2": 17.063011, "mean_abs": 0.878153, "relative": 0.920608 }, { "name": "depth_confidence", "l2": 11.935328, "mean_abs": 0.21621, "relative": 0.000396 }, { "name": "video_fisheye_cam0", "l2": 5.026556, "mean_abs": 0.103987, "relative": 0.910202 }, { "name": "video_fisheye_cam1", "l2": 5.408712, "mean_abs": 0.113265, "relative": 0.893744 }, { "name": "video_fisheye_cam2", "l2": 5.585729, "mean_abs": 0.11391, "relative": 0.891679 }, { "name": "video_fisheye_cam3", "l2": 5.497025, "mean_abs": 0.111012, "relative": 0.92041 }, { "name": "video_stereo_left", "l2": 5.189264, "mean_abs": 0.111931, "relative": 0.843266 }, { "name": "video_stereo_right", "l2": 6.022094, "mean_abs": 0.131135, "relative": 0.848989 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.196894, "mean_abs": 0.089881, "relative": 0.786235 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 547, "start_frame": 2735, "end_frame": 2754, "center_frame": 2744, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.156448, "mean_abs": 0.199835, "relative": 0.85674 }, { "name": "hand_right_joints", "l2": 10.560698, "mean_abs": 0.24385, "relative": 0.832465 }, { "name": "body_joints", "l2": 5.404127, "mean_abs": 0.08113, "relative": 0.729931 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013354, "mean_abs": 0.001914, "relative": 0.031513 }, { "name": "camera_rotation_matrix", "l2": 3.000078, "mean_abs": 0.210935, "relative": 0.921275 }, { "name": "imu_accel_gyro", "l2": 17.076511, "mean_abs": 0.886287, "relative": 0.921337 }, { "name": "depth_confidence", "l2": 12.110124, "mean_abs": 0.217321, "relative": 0.000402 }, { "name": "video_fisheye_cam0", "l2": 5.033138, "mean_abs": 0.104013, "relative": 0.911394 }, { "name": "video_fisheye_cam1", "l2": 5.422446, "mean_abs": 0.113459, "relative": 0.896013 }, { "name": "video_fisheye_cam2", "l2": 5.576319, "mean_abs": 0.113703, "relative": 0.890176 }, { "name": "video_fisheye_cam3", "l2": 5.52005, "mean_abs": 0.111031, "relative": 0.924266 }, { "name": "video_stereo_left", "l2": 5.181699, "mean_abs": 0.111594, "relative": 0.842036 }, { "name": "video_stereo_right", "l2": 6.034843, "mean_abs": 0.131388, "relative": 0.850786 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.152109, "mean_abs": 0.087923, "relative": 0.770208 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 548, "start_frame": 2740, "end_frame": 2759, "center_frame": 2749, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.166814, "mean_abs": 0.199051, "relative": 0.857614 }, { "name": "hand_right_joints", "l2": 10.558542, "mean_abs": 0.243309, "relative": 0.832296 }, { "name": "body_joints", "l2": 5.400278, "mean_abs": 0.080977, "relative": 0.729411 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009592, "mean_abs": 0.001248, "relative": 0.022636 }, { "name": "camera_rotation_matrix", "l2": 2.999641, "mean_abs": 0.210987, "relative": 0.921141 }, { "name": "imu_accel_gyro", "l2": 17.06461, "mean_abs": 0.885462, "relative": 0.920694 }, { "name": "depth_confidence", "l2": 12.120481, "mean_abs": 0.217374, "relative": 0.000402 }, { "name": "video_fisheye_cam0", "l2": 5.036851, "mean_abs": 0.104127, "relative": 0.912066 }, { "name": "video_fisheye_cam1", "l2": 5.432968, "mean_abs": 0.113925, "relative": 0.897752 }, { "name": "video_fisheye_cam2", "l2": 5.565984, "mean_abs": 0.113424, "relative": 0.888526 }, { "name": "video_fisheye_cam3", "l2": 5.518991, "mean_abs": 0.110941, "relative": 0.924089 }, { "name": "video_stereo_left", "l2": 5.175439, "mean_abs": 0.111487, "relative": 0.841019 }, { "name": "video_stereo_right", "l2": 6.047126, "mean_abs": 0.131938, "relative": 0.852518 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.148344, "mean_abs": 0.088593, "relative": 0.76886 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012627, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 549, "start_frame": 2745, "end_frame": 2764, "center_frame": 2754, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.177142, "mean_abs": 0.199005, "relative": 0.858485 }, { "name": "hand_right_joints", "l2": 10.553831, "mean_abs": 0.242482, "relative": 0.831924 }, { "name": "body_joints", "l2": 5.402136, "mean_abs": 0.081065, "relative": 0.729662 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003967, "mean_abs": 0.000553, "relative": 0.009362 }, { "name": "camera_rotation_matrix", "l2": 2.999725, "mean_abs": 0.211246, "relative": 0.921167 }, { "name": "imu_accel_gyro", "l2": 17.073282, "mean_abs": 0.880382, "relative": 0.921162 }, { "name": "depth_confidence", "l2": 12.120882, "mean_abs": 0.217711, "relative": 0.000402 }, { "name": "video_fisheye_cam0", "l2": 5.04104, "mean_abs": 0.104172, "relative": 0.912825 }, { "name": "video_fisheye_cam1", "l2": 5.44936, "mean_abs": 0.114237, "relative": 0.900461 }, { "name": "video_fisheye_cam2", "l2": 5.559329, "mean_abs": 0.113449, "relative": 0.887464 }, { "name": "video_fisheye_cam3", "l2": 5.519541, "mean_abs": 0.111136, "relative": 0.924181 }, { "name": "video_stereo_left", "l2": 5.169479, "mean_abs": 0.111705, "relative": 0.840051 }, { "name": "video_stereo_right", "l2": 6.062125, "mean_abs": 0.132251, "relative": 0.854632 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.203723, "mean_abs": 0.091096, "relative": 0.788679 }, { "name": "caption_objects_interaction_text", "l2": 1.91385, "mean_abs": 0.017798, "relative": 0.878793 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 550, "start_frame": 2750, "end_frame": 2769, "center_frame": 2759, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.181327, "mean_abs": 0.199596, "relative": 0.858838 }, { "name": "hand_right_joints", "l2": 10.558069, "mean_abs": 0.242708, "relative": 0.832258 }, { "name": "body_joints", "l2": 5.403825, "mean_abs": 0.081087, "relative": 0.72989 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006891, "mean_abs": 0.001086, "relative": 0.016263 }, { "name": "camera_rotation_matrix", "l2": 3.00039, "mean_abs": 0.211499, "relative": 0.921371 }, { "name": "imu_accel_gyro", "l2": 17.066267, "mean_abs": 0.881457, "relative": 0.920784 }, { "name": "depth_confidence", "l2": 12.139873, "mean_abs": 0.217802, "relative": 0.000403 }, { "name": "video_fisheye_cam0", "l2": 5.045531, "mean_abs": 0.104122, "relative": 0.913638 }, { "name": "video_fisheye_cam1", "l2": 5.471479, "mean_abs": 0.1147, "relative": 0.904116 }, { "name": "video_fisheye_cam2", "l2": 5.554572, "mean_abs": 0.113401, "relative": 0.886705 }, { "name": "video_fisheye_cam3", "l2": 5.527483, "mean_abs": 0.111665, "relative": 0.92551 }, { "name": "video_stereo_left", "l2": 5.168519, "mean_abs": 0.111905, "relative": 0.839895 }, { "name": "video_stereo_right", "l2": 6.078128, "mean_abs": 0.132819, "relative": 0.856888 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.182277, "mean_abs": 0.088222, "relative": 0.781004 }, { "name": "caption_objects_interaction_text", "l2": 1.91385, "mean_abs": 0.017957, "relative": 0.878793 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 551, "start_frame": 2755, "end_frame": 2774, "center_frame": 2764, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.182006, "mean_abs": 0.199656, "relative": 0.858896 }, { "name": "hand_right_joints", "l2": 10.570797, "mean_abs": 0.242948, "relative": 0.833262 }, { "name": "body_joints", "l2": 5.404802, "mean_abs": 0.081026, "relative": 0.730022 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007709, "mean_abs": 0.001183, "relative": 0.018193 }, { "name": "camera_rotation_matrix", "l2": 3.0005, "mean_abs": 0.21139, "relative": 0.921405 }, { "name": "imu_accel_gyro", "l2": 17.059795, "mean_abs": 0.877634, "relative": 0.920435 }, { "name": "depth_confidence", "l2": 11.995552, "mean_abs": 0.216893, "relative": 0.000398 }, { "name": "video_fisheye_cam0", "l2": 5.043931, "mean_abs": 0.104066, "relative": 0.913348 }, { "name": "video_fisheye_cam1", "l2": 5.475876, "mean_abs": 0.114709, "relative": 0.904842 }, { "name": "video_fisheye_cam2", "l2": 5.552613, "mean_abs": 0.113378, "relative": 0.886392 }, { "name": "video_fisheye_cam3", "l2": 5.531203, "mean_abs": 0.111512, "relative": 0.926133 }, { "name": "video_stereo_left", "l2": 5.168581, "mean_abs": 0.111759, "relative": 0.839905 }, { "name": "video_stereo_right", "l2": 6.080058, "mean_abs": 0.13278, "relative": 0.857161 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.170979, "mean_abs": 0.087871, "relative": 0.776961 }, { "name": "caption_objects_interaction_text", "l2": 1.91385, "mean_abs": 0.01769, "relative": 0.878793 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 552, "start_frame": 2760, "end_frame": 2779, "center_frame": 2769, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.182062, "mean_abs": 0.200087, "relative": 0.8589 }, { "name": "hand_right_joints", "l2": 10.562559, "mean_abs": 0.243168, "relative": 0.832612 }, { "name": "body_joints", "l2": 5.409367, "mean_abs": 0.081204, "relative": 0.730639 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008678, "mean_abs": 0.001332, "relative": 0.020478 }, { "name": "camera_rotation_matrix", "l2": 2.999972, "mean_abs": 0.210668, "relative": 0.921243 }, { "name": "imu_accel_gyro", "l2": 17.059532, "mean_abs": 0.874126, "relative": 0.920421 }, { "name": "depth_confidence", "l2": 12.001274, "mean_abs": 0.216493, "relative": 0.000398 }, { "name": "video_fisheye_cam0", "l2": 5.040951, "mean_abs": 0.103808, "relative": 0.912808 }, { "name": "video_fisheye_cam1", "l2": 5.476223, "mean_abs": 0.11444, "relative": 0.9049 }, { "name": "video_fisheye_cam2", "l2": 5.547101, "mean_abs": 0.113, "relative": 0.885512 }, { "name": "video_fisheye_cam3", "l2": 5.532103, "mean_abs": 0.111246, "relative": 0.926284 }, { "name": "video_stereo_left", "l2": 5.163354, "mean_abs": 0.111165, "relative": 0.839055 }, { "name": "video_stereo_right", "l2": 6.077723, "mean_abs": 0.132226, "relative": 0.856831 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.14428, "mean_abs": 0.085714, "relative": 0.767406 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011979, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 553, "start_frame": 2765, "end_frame": 2784, "center_frame": 2774, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.175637, "mean_abs": 0.199734, "relative": 0.858358 }, { "name": "hand_right_joints", "l2": 10.558095, "mean_abs": 0.24279, "relative": 0.83226 }, { "name": "body_joints", "l2": 5.413245, "mean_abs": 0.080992, "relative": 0.731163 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007588, "mean_abs": 0.001108, "relative": 0.017906 }, { "name": "camera_rotation_matrix", "l2": 2.999928, "mean_abs": 0.21052, "relative": 0.921229 }, { "name": "imu_accel_gyro", "l2": 17.06719, "mean_abs": 0.874147, "relative": 0.920834 }, { "name": "depth_confidence", "l2": 12.006639, "mean_abs": 0.216768, "relative": 0.000398 }, { "name": "video_fisheye_cam0", "l2": 5.039258, "mean_abs": 0.103754, "relative": 0.912502 }, { "name": "video_fisheye_cam1", "l2": 5.482563, "mean_abs": 0.114414, "relative": 0.905947 }, { "name": "video_fisheye_cam2", "l2": 5.544569, "mean_abs": 0.112909, "relative": 0.885108 }, { "name": "video_fisheye_cam3", "l2": 5.54078, "mean_abs": 0.111467, "relative": 0.927737 }, { "name": "video_stereo_left", "l2": 5.165238, "mean_abs": 0.111072, "relative": 0.839361 }, { "name": "video_stereo_right", "l2": 6.085616, "mean_abs": 0.132074, "relative": 0.857944 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.133296, "mean_abs": 0.088355, "relative": 0.763474 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011979, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 554, "start_frame": 2770, "end_frame": 2789, "center_frame": 2779, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.181484, "mean_abs": 0.200018, "relative": 0.858852 }, { "name": "hand_right_joints", "l2": 10.553598, "mean_abs": 0.242666, "relative": 0.831906 }, { "name": "body_joints", "l2": 5.417678, "mean_abs": 0.081091, "relative": 0.731761 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004379, "mean_abs": 0.000609, "relative": 0.010333 }, { "name": "camera_rotation_matrix", "l2": 2.999934, "mean_abs": 0.210635, "relative": 0.921231 }, { "name": "imu_accel_gyro", "l2": 17.067974, "mean_abs": 0.871845, "relative": 0.920876 }, { "name": "depth_confidence", "l2": 12.004549, "mean_abs": 0.216767, "relative": 0.000398 }, { "name": "video_fisheye_cam0", "l2": 5.036077, "mean_abs": 0.103762, "relative": 0.911926 }, { "name": "video_fisheye_cam1", "l2": 5.48244, "mean_abs": 0.114507, "relative": 0.905927 }, { "name": "video_fisheye_cam2", "l2": 5.541294, "mean_abs": 0.112886, "relative": 0.884585 }, { "name": "video_fisheye_cam3", "l2": 5.534604, "mean_abs": 0.111432, "relative": 0.926703 }, { "name": "video_stereo_left", "l2": 5.162332, "mean_abs": 0.111174, "relative": 0.838889 }, { "name": "video_stereo_right", "l2": 6.087205, "mean_abs": 0.132374, "relative": 0.858168 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.301273, "mean_abs": 0.09384, "relative": 0.823591 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011979, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 555, "start_frame": 2775, "end_frame": 2794, "center_frame": 2784, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.177168, "mean_abs": 0.199914, "relative": 0.858488 }, { "name": "hand_right_joints", "l2": 10.54939, "mean_abs": 0.243223, "relative": 0.831574 }, { "name": "body_joints", "l2": 5.420849, "mean_abs": 0.081521, "relative": 0.73219 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00442, "mean_abs": 0.000632, "relative": 0.01043 }, { "name": "camera_rotation_matrix", "l2": 2.999798, "mean_abs": 0.210471, "relative": 0.921189 }, { "name": "imu_accel_gyro", "l2": 17.085522, "mean_abs": 0.873569, "relative": 0.921823 }, { "name": "depth_confidence", "l2": 11.994023, "mean_abs": 0.21663, "relative": 0.000398 }, { "name": "video_fisheye_cam0", "l2": 5.032668, "mean_abs": 0.103686, "relative": 0.911308 }, { "name": "video_fisheye_cam1", "l2": 5.47493, "mean_abs": 0.114325, "relative": 0.904686 }, { "name": "video_fisheye_cam2", "l2": 5.534973, "mean_abs": 0.11283, "relative": 0.883576 }, { "name": "video_fisheye_cam3", "l2": 5.522764, "mean_abs": 0.110827, "relative": 0.92472 }, { "name": "video_stereo_left", "l2": 5.155272, "mean_abs": 0.111075, "relative": 0.837742 }, { "name": "video_stereo_right", "l2": 6.076698, "mean_abs": 0.132083, "relative": 0.856687 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.15205, "mean_abs": 0.084119, "relative": 0.770186 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011979, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 556, "start_frame": 2780, "end_frame": 2799, "center_frame": 2789, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.178515, "mean_abs": 0.199741, "relative": 0.858601 }, { "name": "hand_right_joints", "l2": 10.552706, "mean_abs": 0.243225, "relative": 0.831836 }, { "name": "body_joints", "l2": 5.418971, "mean_abs": 0.081515, "relative": 0.731936 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.0053, "mean_abs": 0.000792, "relative": 0.012507 }, { "name": "camera_rotation_matrix", "l2": 2.999881, "mean_abs": 0.210537, "relative": 0.921215 }, { "name": "imu_accel_gyro", "l2": 17.085585, "mean_abs": 0.87597, "relative": 0.921826 }, { "name": "depth_confidence", "l2": 11.967622, "mean_abs": 0.216822, "relative": 0.000397 }, { "name": "video_fisheye_cam0", "l2": 5.037126, "mean_abs": 0.10388, "relative": 0.912116 }, { "name": "video_fisheye_cam1", "l2": 5.467153, "mean_abs": 0.114363, "relative": 0.903401 }, { "name": "video_fisheye_cam2", "l2": 5.531132, "mean_abs": 0.112772, "relative": 0.882963 }, { "name": "video_fisheye_cam3", "l2": 5.515015, "mean_abs": 0.111041, "relative": 0.923423 }, { "name": "video_stereo_left", "l2": 5.15, "mean_abs": 0.111159, "relative": 0.836885 }, { "name": "video_stereo_right", "l2": 6.064289, "mean_abs": 0.13217, "relative": 0.854938 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.21118, "mean_abs": 0.087442, "relative": 0.791348 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011979, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 557, "start_frame": 2785, "end_frame": 2804, "center_frame": 2794, "action": "Hold gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.175863, "mean_abs": 0.199662, "relative": 0.858377 }, { "name": "hand_right_joints", "l2": 10.552855, "mean_abs": 0.243219, "relative": 0.831847 }, { "name": "body_joints", "l2": 5.416861, "mean_abs": 0.081494, "relative": 0.731651 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005198, "mean_abs": 0.000862, "relative": 0.012268 }, { "name": "camera_rotation_matrix", "l2": 2.999954, "mean_abs": 0.210542, "relative": 0.921237 }, { "name": "imu_accel_gyro", "l2": 17.084724, "mean_abs": 0.873441, "relative": 0.92178 }, { "name": "depth_confidence", "l2": 11.966084, "mean_abs": 0.216971, "relative": 0.000397 }, { "name": "video_fisheye_cam0", "l2": 5.044116, "mean_abs": 0.104213, "relative": 0.913382 }, { "name": "video_fisheye_cam1", "l2": 5.457963, "mean_abs": 0.114106, "relative": 0.901882 }, { "name": "video_fisheye_cam2", "l2": 5.526459, "mean_abs": 0.112647, "relative": 0.882217 }, { "name": "video_fisheye_cam3", "l2": 5.507327, "mean_abs": 0.111008, "relative": 0.922136 }, { "name": "video_stereo_left", "l2": 5.142741, "mean_abs": 0.110983, "relative": 0.835706 }, { "name": "video_stereo_right", "l2": 6.050505, "mean_abs": 0.131816, "relative": 0.852994 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.131693, "mean_abs": 0.084245, "relative": 0.762901 }, { "name": "caption_objects_interaction_text", "l2": 1.962123, "mean_abs": 0.017329, "relative": 0.900959 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 558, "start_frame": 2790, "end_frame": 2809, "center_frame": 2799, "action": "", "subtask": "", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.175958, "mean_abs": 0.19934, "relative": 0.858385 }, { "name": "hand_right_joints", "l2": 10.560799, "mean_abs": 0.243825, "relative": 0.832473 }, { "name": "body_joints", "l2": 5.422042, "mean_abs": 0.0816, "relative": 0.732351 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00671, "mean_abs": 0.001053, "relative": 0.015835 }, { "name": "camera_rotation_matrix", "l2": 2.999985, "mean_abs": 0.210376, "relative": 0.921247 }, { "name": "imu_accel_gyro", "l2": 17.10293, "mean_abs": 0.873777, "relative": 0.922762 }, { "name": "depth_confidence", "l2": 11.95448, "mean_abs": 0.216355, "relative": 0.000397 }, { "name": "video_fisheye_cam0", "l2": 5.046973, "mean_abs": 0.104143, "relative": 0.913899 }, { "name": "video_fisheye_cam1", "l2": 5.444832, "mean_abs": 0.114136, "relative": 0.899713 }, { "name": "video_fisheye_cam2", "l2": 5.518873, "mean_abs": 0.11252, "relative": 0.881006 }, { "name": "video_fisheye_cam3", "l2": 5.491286, "mean_abs": 0.111018, "relative": 0.91945 }, { "name": "video_stereo_left", "l2": 5.131878, "mean_abs": 0.11089, "relative": 0.83394 }, { "name": "video_stereo_right", "l2": 6.036272, "mean_abs": 0.131789, "relative": 0.850988 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.15085, "mean_abs": 0.087297, "relative": 0.769757 }, { "name": "caption_objects_interaction_text", "l2": 1.962123, "mean_abs": 0.017409, "relative": 0.900959 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 559, "start_frame": 2795, "end_frame": 2814, "center_frame": 2804, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "mug", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.176538, "mean_abs": 0.199579, "relative": 0.858434 }, { "name": "hand_right_joints", "l2": 10.563196, "mean_abs": 0.243526, "relative": 0.832662 }, { "name": "body_joints", "l2": 5.422179, "mean_abs": 0.081695, "relative": 0.732369 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00415, "mean_abs": 0.000656, "relative": 0.009792 }, { "name": "camera_rotation_matrix", "l2": 3.000148, "mean_abs": 0.210266, "relative": 0.921297 }, { "name": "imu_accel_gyro", "l2": 17.086878, "mean_abs": 0.873379, "relative": 0.921896 }, { "name": "depth_confidence", "l2": 11.960469, "mean_abs": 0.216529, "relative": 0.000397 }, { "name": "video_fisheye_cam0", "l2": 5.048974, "mean_abs": 0.104082, "relative": 0.914261 }, { "name": "video_fisheye_cam1", "l2": 5.43387, "mean_abs": 0.114048, "relative": 0.897901 }, { "name": "video_fisheye_cam2", "l2": 5.513419, "mean_abs": 0.112406, "relative": 0.880135 }, { "name": "video_fisheye_cam3", "l2": 5.473868, "mean_abs": 0.111103, "relative": 0.916533 }, { "name": "video_stereo_left", "l2": 5.123687, "mean_abs": 0.110699, "relative": 0.832609 }, { "name": "video_stereo_right", "l2": 6.025071, "mean_abs": 0.131505, "relative": 0.849409 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.109719, "mean_abs": 0.084283, "relative": 0.755037 }, { "name": "caption_objects_interaction_text", "l2": 1.962123, "mean_abs": 0.017007, "relative": 0.900959 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 560, "start_frame": 2800, "end_frame": 2819, "center_frame": 2809, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.167461, "mean_abs": 0.199839, "relative": 0.857669 }, { "name": "hand_right_joints", "l2": 10.559084, "mean_abs": 0.24334, "relative": 0.832338 }, { "name": "body_joints", "l2": 5.433531, "mean_abs": 0.081951, "relative": 0.733903 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004042, "mean_abs": 0.000606, "relative": 0.00954 }, { "name": "camera_rotation_matrix", "l2": 3.000076, "mean_abs": 0.210282, "relative": 0.921275 }, { "name": "imu_accel_gyro", "l2": 17.08457, "mean_abs": 0.873876, "relative": 0.921771 }, { "name": "depth_confidence", "l2": 11.961847, "mean_abs": 0.216626, "relative": 0.000397 }, { "name": "video_fisheye_cam0", "l2": 5.052832, "mean_abs": 0.1042, "relative": 0.91496 }, { "name": "video_fisheye_cam1", "l2": 5.421475, "mean_abs": 0.114, "relative": 0.895853 }, { "name": "video_fisheye_cam2", "l2": 5.506345, "mean_abs": 0.112284, "relative": 0.879006 }, { "name": "video_fisheye_cam3", "l2": 5.449286, "mean_abs": 0.111277, "relative": 0.912417 }, { "name": "video_stereo_left", "l2": 5.115296, "mean_abs": 0.110646, "relative": 0.831246 }, { "name": "video_stereo_right", "l2": 6.012425, "mean_abs": 0.13163, "relative": 0.847626 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.050107, "mean_abs": 0.084763, "relative": 0.733702 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 561, "start_frame": 2805, "end_frame": 2824, "center_frame": 2814, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.168796, "mean_abs": 0.200366, "relative": 0.857781 }, { "name": "hand_right_joints", "l2": 10.554046, "mean_abs": 0.244537, "relative": 0.831941 }, { "name": "body_joints", "l2": 5.447695, "mean_abs": 0.082603, "relative": 0.735816 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004699, "mean_abs": 0.000718, "relative": 0.011089 }, { "name": "camera_rotation_matrix", "l2": 3.000067, "mean_abs": 0.210405, "relative": 0.921272 }, { "name": "imu_accel_gyro", "l2": 17.078745, "mean_abs": 0.874808, "relative": 0.921457 }, { "name": "depth_confidence", "l2": 11.966755, "mean_abs": 0.216763, "relative": 0.000397 }, { "name": "video_fisheye_cam0", "l2": 5.045477, "mean_abs": 0.103898, "relative": 0.913628 }, { "name": "video_fisheye_cam1", "l2": 5.405224, "mean_abs": 0.113247, "relative": 0.893168 }, { "name": "video_fisheye_cam2", "l2": 5.497051, "mean_abs": 0.111987, "relative": 0.877522 }, { "name": "video_fisheye_cam3", "l2": 5.423745, "mean_abs": 0.109622, "relative": 0.908141 }, { "name": "video_stereo_left", "l2": 5.102043, "mean_abs": 0.110164, "relative": 0.829092 }, { "name": "video_stereo_right", "l2": 5.995011, "mean_abs": 0.130776, "relative": 0.845171 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.148807, "mean_abs": 0.088938, "relative": 0.769026 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 562, "start_frame": 2810, "end_frame": 2829, "center_frame": 2819, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.16454, "mean_abs": 0.201328, "relative": 0.857422 }, { "name": "hand_right_joints", "l2": 10.542298, "mean_abs": 0.244473, "relative": 0.831015 }, { "name": "body_joints", "l2": 5.463964, "mean_abs": 0.083051, "relative": 0.738013 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004125, "mean_abs": 0.000681, "relative": 0.009733 }, { "name": "camera_rotation_matrix", "l2": 3.000096, "mean_abs": 0.210321, "relative": 0.921281 }, { "name": "imu_accel_gyro", "l2": 17.078596, "mean_abs": 0.86999, "relative": 0.921449 }, { "name": "depth_confidence", "l2": 11.995807, "mean_abs": 0.216953, "relative": 0.000398 }, { "name": "video_fisheye_cam0", "l2": 5.04119, "mean_abs": 0.103755, "relative": 0.912852 }, { "name": "video_fisheye_cam1", "l2": 5.398249, "mean_abs": 0.113011, "relative": 0.892015 }, { "name": "video_fisheye_cam2", "l2": 5.49188, "mean_abs": 0.111823, "relative": 0.876697 }, { "name": "video_fisheye_cam3", "l2": 5.433848, "mean_abs": 0.110186, "relative": 0.909832 }, { "name": "video_stereo_left", "l2": 5.09523, "mean_abs": 0.109956, "relative": 0.827985 }, { "name": "video_stereo_right", "l2": 5.985642, "mean_abs": 0.130582, "relative": 0.84385 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.99596, "mean_abs": 0.078801, "relative": 0.714324 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 563, "start_frame": 2815, "end_frame": 2834, "center_frame": 2824, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.164026, "mean_abs": 0.20198, "relative": 0.857379 }, { "name": "hand_right_joints", "l2": 10.535235, "mean_abs": 0.245623, "relative": 0.830458 }, { "name": "body_joints", "l2": 5.488928, "mean_abs": 0.083886, "relative": 0.741385 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005216, "mean_abs": 0.000817, "relative": 0.01231 }, { "name": "camera_rotation_matrix", "l2": 2.999949, "mean_abs": 0.210553, "relative": 0.921236 }, { "name": "imu_accel_gyro", "l2": 17.100071, "mean_abs": 0.871531, "relative": 0.922608 }, { "name": "depth_confidence", "l2": 11.99164, "mean_abs": 0.217404, "relative": 0.000398 }, { "name": "video_fisheye_cam0", "l2": 5.039945, "mean_abs": 0.10395, "relative": 0.912626 }, { "name": "video_fisheye_cam1", "l2": 5.399892, "mean_abs": 0.113504, "relative": 0.892287 }, { "name": "video_fisheye_cam2", "l2": 5.489706, "mean_abs": 0.112013, "relative": 0.87635 }, { "name": "video_fisheye_cam3", "l2": 5.520393, "mean_abs": 0.112703, "relative": 0.924323 }, { "name": "video_stereo_left", "l2": 5.093483, "mean_abs": 0.110402, "relative": 0.827701 }, { "name": "video_stereo_right", "l2": 5.982869, "mean_abs": 0.131346, "relative": 0.843459 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.04296, "mean_abs": 0.080486, "relative": 0.731145 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 564, "start_frame": 2820, "end_frame": 2839, "center_frame": 2829, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.151135, "mean_abs": 0.203811, "relative": 0.856292 }, { "name": "hand_right_joints", "l2": 10.524775, "mean_abs": 0.247448, "relative": 0.829634 }, { "name": "body_joints", "l2": 5.532572, "mean_abs": 0.08529, "relative": 0.74728 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004414, "mean_abs": 0.000695, "relative": 0.010417 }, { "name": "camera_rotation_matrix", "l2": 2.999816, "mean_abs": 0.210397, "relative": 0.921195 }, { "name": "imu_accel_gyro", "l2": 17.081026, "mean_abs": 0.869793, "relative": 0.92158 }, { "name": "depth_confidence", "l2": 11.99423, "mean_abs": 0.217109, "relative": 0.000398 }, { "name": "video_fisheye_cam0", "l2": 5.037341, "mean_abs": 0.103749, "relative": 0.912155 }, { "name": "video_fisheye_cam1", "l2": 5.394812, "mean_abs": 0.113277, "relative": 0.891447 }, { "name": "video_fisheye_cam2", "l2": 5.486143, "mean_abs": 0.111797, "relative": 0.875781 }, { "name": "video_fisheye_cam3", "l2": 5.535955, "mean_abs": 0.11288, "relative": 0.926929 }, { "name": "video_stereo_left", "l2": 5.088891, "mean_abs": 0.110022, "relative": 0.826955 }, { "name": "video_stereo_right", "l2": 5.976295, "mean_abs": 0.130708, "relative": 0.842532 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.148602, "mean_abs": 0.090111, "relative": 0.768952 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 565, "start_frame": 2825, "end_frame": 2844, "center_frame": 2834, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.13852, "mean_abs": 0.204677, "relative": 0.855227 }, { "name": "hand_right_joints", "l2": 10.511817, "mean_abs": 0.247469, "relative": 0.828612 }, { "name": "body_joints", "l2": 5.5658, "mean_abs": 0.086324, "relative": 0.751768 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003845, "mean_abs": 0.000599, "relative": 0.009073 }, { "name": "camera_rotation_matrix", "l2": 2.99994, "mean_abs": 0.210284, "relative": 0.921233 }, { "name": "imu_accel_gyro", "l2": 17.07976, "mean_abs": 0.868447, "relative": 0.921512 }, { "name": "depth_confidence", "l2": 12.009793, "mean_abs": 0.217137, "relative": 0.000399 }, { "name": "video_fisheye_cam0", "l2": 5.034715, "mean_abs": 0.103653, "relative": 0.911679 }, { "name": "video_fisheye_cam1", "l2": 5.389323, "mean_abs": 0.11318, "relative": 0.89054 }, { "name": "video_fisheye_cam2", "l2": 5.482092, "mean_abs": 0.111673, "relative": 0.875134 }, { "name": "video_fisheye_cam3", "l2": 5.554853, "mean_abs": 0.112731, "relative": 0.930093 }, { "name": "video_stereo_left", "l2": 5.084811, "mean_abs": 0.10974, "relative": 0.826292 }, { "name": "video_stereo_right", "l2": 5.96997, "mean_abs": 0.130324, "relative": 0.841641 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.094447, "mean_abs": 0.085189, "relative": 0.749571 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 566, "start_frame": 2830, "end_frame": 2849, "center_frame": 2839, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.128055, "mean_abs": 0.20528, "relative": 0.854345 }, { "name": "hand_right_joints", "l2": 10.509518, "mean_abs": 0.248505, "relative": 0.828431 }, { "name": "body_joints", "l2": 5.599771, "mean_abs": 0.086687, "relative": 0.756356 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004717, "mean_abs": 0.000684, "relative": 0.011132 }, { "name": "camera_rotation_matrix", "l2": 2.999767, "mean_abs": 0.21038, "relative": 0.92118 }, { "name": "imu_accel_gyro", "l2": 17.078476, "mean_abs": 0.869594, "relative": 0.921443 }, { "name": "depth_confidence", "l2": 12.0178, "mean_abs": 0.217388, "relative": 0.000399 }, { "name": "video_fisheye_cam0", "l2": 5.032833, "mean_abs": 0.103927, "relative": 0.911338 }, { "name": "video_fisheye_cam1", "l2": 5.386956, "mean_abs": 0.113738, "relative": 0.890149 }, { "name": "video_fisheye_cam2", "l2": 5.479393, "mean_abs": 0.11178, "relative": 0.874704 }, { "name": "video_fisheye_cam3", "l2": 5.560732, "mean_abs": 0.114167, "relative": 0.931077 }, { "name": "video_stereo_left", "l2": 5.083372, "mean_abs": 0.110214, "relative": 0.826058 }, { "name": "video_stereo_right", "l2": 5.967715, "mean_abs": 0.131168, "relative": 0.841323 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.11379, "mean_abs": 0.084685, "relative": 0.756494 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 567, "start_frame": 2835, "end_frame": 2854, "center_frame": 2844, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.122799, "mean_abs": 0.206958, "relative": 0.853901 }, { "name": "hand_right_joints", "l2": 10.478474, "mean_abs": 0.249782, "relative": 0.825984 }, { "name": "body_joints", "l2": 5.628071, "mean_abs": 0.086826, "relative": 0.760179 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005159, "mean_abs": 0.000751, "relative": 0.012175 }, { "name": "camera_rotation_matrix", "l2": 2.999906, "mean_abs": 0.21045, "relative": 0.921223 }, { "name": "imu_accel_gyro", "l2": 17.073158, "mean_abs": 0.868363, "relative": 0.921156 }, { "name": "depth_confidence", "l2": 12.053716, "mean_abs": 0.217681, "relative": 0.0004 }, { "name": "video_fisheye_cam0", "l2": 5.024791, "mean_abs": 0.103327, "relative": 0.909882 }, { "name": "video_fisheye_cam1", "l2": 5.362229, "mean_abs": 0.112345, "relative": 0.886063 }, { "name": "video_fisheye_cam2", "l2": 5.471801, "mean_abs": 0.111369, "relative": 0.873492 }, { "name": "video_fisheye_cam3", "l2": 5.528902, "mean_abs": 0.111774, "relative": 0.925748 }, { "name": "video_stereo_left", "l2": 5.072104, "mean_abs": 0.109215, "relative": 0.824227 }, { "name": "video_stereo_right", "l2": 5.944269, "mean_abs": 0.129348, "relative": 0.838017 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.126932, "mean_abs": 0.085558, "relative": 0.761197 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 568, "start_frame": 2840, "end_frame": 2859, "center_frame": 2849, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.125948, "mean_abs": 0.207877, "relative": 0.854167 }, { "name": "hand_right_joints", "l2": 10.452581, "mean_abs": 0.250454, "relative": 0.823943 }, { "name": "body_joints", "l2": 5.670911, "mean_abs": 0.086925, "relative": 0.765965 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004686, "mean_abs": 0.000748, "relative": 0.011058 }, { "name": "camera_rotation_matrix", "l2": 3.000014, "mean_abs": 0.21047, "relative": 0.921255 }, { "name": "imu_accel_gyro", "l2": 17.088755, "mean_abs": 0.871412, "relative": 0.921997 }, { "name": "depth_confidence", "l2": 12.057149, "mean_abs": 0.217875, "relative": 0.0004 }, { "name": "video_fisheye_cam0", "l2": 5.016875, "mean_abs": 0.103197, "relative": 0.908449 }, { "name": "video_fisheye_cam1", "l2": 5.367803, "mean_abs": 0.112848, "relative": 0.886984 }, { "name": "video_fisheye_cam2", "l2": 5.470411, "mean_abs": 0.111548, "relative": 0.87327 }, { "name": "video_fisheye_cam3", "l2": 5.496913, "mean_abs": 0.111327, "relative": 0.920392 }, { "name": "video_stereo_left", "l2": 5.072591, "mean_abs": 0.109647, "relative": 0.824306 }, { "name": "video_stereo_right", "l2": 5.959484, "mean_abs": 0.130634, "relative": 0.840162 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.111642, "mean_abs": 0.085668, "relative": 0.755725 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 569, "start_frame": 2845, "end_frame": 2864, "center_frame": 2854, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.143989, "mean_abs": 0.209879, "relative": 0.855689 }, { "name": "hand_right_joints", "l2": 10.456052, "mean_abs": 0.250605, "relative": 0.824217 }, { "name": "body_joints", "l2": 5.678655, "mean_abs": 0.08717, "relative": 0.767011 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008211, "mean_abs": 0.001158, "relative": 0.019376 }, { "name": "camera_rotation_matrix", "l2": 3.000211, "mean_abs": 0.21197, "relative": 0.921316 }, { "name": "imu_accel_gyro", "l2": 17.122984, "mean_abs": 0.878737, "relative": 0.923844 }, { "name": "depth_confidence", "l2": 12.032562, "mean_abs": 0.218138, "relative": 0.000399 }, { "name": "video_fisheye_cam0", "l2": 5.029135, "mean_abs": 0.104114, "relative": 0.910669 }, { "name": "video_fisheye_cam1", "l2": 5.394278, "mean_abs": 0.114012, "relative": 0.891359 }, { "name": "video_fisheye_cam2", "l2": 5.474951, "mean_abs": 0.112086, "relative": 0.873995 }, { "name": "video_fisheye_cam3", "l2": 5.506329, "mean_abs": 0.111277, "relative": 0.921968 }, { "name": "video_stereo_left", "l2": 5.084604, "mean_abs": 0.110756, "relative": 0.826258 }, { "name": "video_stereo_right", "l2": 6.002128, "mean_abs": 0.132812, "relative": 0.846174 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.087397, "mean_abs": 0.082883, "relative": 0.747048 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 570, "start_frame": 2850, "end_frame": 2869, "center_frame": 2859, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.174331, "mean_abs": 0.213315, "relative": 0.858248 }, { "name": "hand_right_joints", "l2": 10.514378, "mean_abs": 0.252981, "relative": 0.828814 }, { "name": "body_joints", "l2": 5.689159, "mean_abs": 0.087979, "relative": 0.76843 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.025002, "mean_abs": 0.00373, "relative": 0.059 }, { "name": "camera_rotation_matrix", "l2": 3.001313, "mean_abs": 0.214285, "relative": 0.921655 }, { "name": "imu_accel_gyro", "l2": 17.130503, "mean_abs": 0.910497, "relative": 0.92425 }, { "name": "depth_confidence", "l2": 12.006116, "mean_abs": 0.219215, "relative": 0.000398 }, { "name": "video_fisheye_cam0", "l2": 5.037254, "mean_abs": 0.104871, "relative": 0.912139 }, { "name": "video_fisheye_cam1", "l2": 5.397856, "mean_abs": 0.114775, "relative": 0.89195 }, { "name": "video_fisheye_cam2", "l2": 5.477071, "mean_abs": 0.112924, "relative": 0.874333 }, { "name": "video_fisheye_cam3", "l2": 5.516882, "mean_abs": 0.112072, "relative": 0.923735 }, { "name": "video_stereo_left", "l2": 5.106273, "mean_abs": 0.112953, "relative": 0.829779 }, { "name": "video_stereo_right", "l2": 6.038186, "mean_abs": 0.135556, "relative": 0.851258 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.058341, "mean_abs": 0.084644, "relative": 0.736649 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 571, "start_frame": 2855, "end_frame": 2874, "center_frame": 2864, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.191216, "mean_abs": 0.214692, "relative": 0.859672 }, { "name": "hand_right_joints", "l2": 10.622488, "mean_abs": 0.256992, "relative": 0.837336 }, { "name": "body_joints", "l2": 5.725872, "mean_abs": 0.088909, "relative": 0.773389 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.045382, "mean_abs": 0.006543, "relative": 0.107096 }, { "name": "camera_rotation_matrix", "l2": 3.002667, "mean_abs": 0.215179, "relative": 0.92207 }, { "name": "imu_accel_gyro", "l2": 17.128376, "mean_abs": 0.937442, "relative": 0.924135 }, { "name": "depth_confidence", "l2": 11.903421, "mean_abs": 0.219139, "relative": 0.000395 }, { "name": "video_fisheye_cam0", "l2": 5.045626, "mean_abs": 0.105275, "relative": 0.913655 }, { "name": "video_fisheye_cam1", "l2": 5.407869, "mean_abs": 0.115091, "relative": 0.893605 }, { "name": "video_fisheye_cam2", "l2": 5.477488, "mean_abs": 0.113254, "relative": 0.874399 }, { "name": "video_fisheye_cam3", "l2": 5.526484, "mean_abs": 0.112727, "relative": 0.925343 }, { "name": "video_stereo_left", "l2": 5.116409, "mean_abs": 0.1144, "relative": 0.831427 }, { "name": "video_stereo_right", "l2": 6.07593, "mean_abs": 0.13728, "relative": 0.856579 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.099878, "mean_abs": 0.084903, "relative": 0.751515 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 572, "start_frame": 2860, "end_frame": 2879, "center_frame": 2869, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.174047, "mean_abs": 0.216562, "relative": 0.858224 }, { "name": "hand_right_joints", "l2": 10.721699, "mean_abs": 0.263167, "relative": 0.845157 }, { "name": "body_joints", "l2": 5.769908, "mean_abs": 0.089536, "relative": 0.779337 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.071969, "mean_abs": 0.009135, "relative": 0.169838 }, { "name": "camera_rotation_matrix", "l2": 3.003862, "mean_abs": 0.216206, "relative": 0.922437 }, { "name": "imu_accel_gyro", "l2": 17.166094, "mean_abs": 0.962493, "relative": 0.92617 }, { "name": "depth_confidence", "l2": 11.814172, "mean_abs": 0.218117, "relative": 0.000392 }, { "name": "video_fisheye_cam0", "l2": 5.053107, "mean_abs": 0.105569, "relative": 0.91501 }, { "name": "video_fisheye_cam1", "l2": 5.393133, "mean_abs": 0.115364, "relative": 0.89117 }, { "name": "video_fisheye_cam2", "l2": 5.489008, "mean_abs": 0.11366, "relative": 0.876238 }, { "name": "video_fisheye_cam3", "l2": 5.543383, "mean_abs": 0.113333, "relative": 0.928173 }, { "name": "video_stereo_left", "l2": 5.137933, "mean_abs": 0.115427, "relative": 0.834924 }, { "name": "video_stereo_right", "l2": 6.100012, "mean_abs": 0.1389, "relative": 0.859974 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.106599, "mean_abs": 0.085482, "relative": 0.75392 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 573, "start_frame": 2865, "end_frame": 2884, "center_frame": 2874, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.146688, "mean_abs": 0.219952, "relative": 0.855916 }, { "name": "hand_right_joints", "l2": 11.001089, "mean_abs": 0.275167, "relative": 0.86718 }, { "name": "body_joints", "l2": 5.812274, "mean_abs": 0.089285, "relative": 0.785059 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.101528, "mean_abs": 0.013206, "relative": 0.239591 }, { "name": "camera_rotation_matrix", "l2": 3.005366, "mean_abs": 0.217133, "relative": 0.922899 }, { "name": "imu_accel_gyro", "l2": 17.126722, "mean_abs": 0.982503, "relative": 0.924046 }, { "name": "depth_confidence", "l2": 11.718878, "mean_abs": 0.217096, "relative": 0.000389 }, { "name": "video_fisheye_cam0", "l2": 5.066696, "mean_abs": 0.105934, "relative": 0.91747 }, { "name": "video_fisheye_cam1", "l2": 5.379435, "mean_abs": 0.115402, "relative": 0.888906 }, { "name": "video_fisheye_cam2", "l2": 5.500122, "mean_abs": 0.113923, "relative": 0.878013 }, { "name": "video_fisheye_cam3", "l2": 5.555614, "mean_abs": 0.11427, "relative": 0.930221 }, { "name": "video_stereo_left", "l2": 5.179854, "mean_abs": 0.11691, "relative": 0.841737 }, { "name": "video_stereo_right", "l2": 6.145857, "mean_abs": 0.140473, "relative": 0.866437 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.049372, "mean_abs": 0.081639, "relative": 0.733439 }, { "name": "caption_objects_interaction_text", "l2": 1.954153, "mean_abs": 0.017797, "relative": 0.8973 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 574, "start_frame": 2870, "end_frame": 2889, "center_frame": 2879, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.094466, "mean_abs": 0.222264, "relative": 0.851511 }, { "name": "hand_right_joints", "l2": 11.315459, "mean_abs": 0.286147, "relative": 0.891961 }, { "name": "body_joints", "l2": 5.872455, "mean_abs": 0.09115, "relative": 0.793188 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.135053, "mean_abs": 0.018158, "relative": 0.318705 }, { "name": "camera_rotation_matrix", "l2": 3.009202, "mean_abs": 0.218294, "relative": 0.924077 }, { "name": "imu_accel_gyro", "l2": 17.136984, "mean_abs": 1.053841, "relative": 0.924599 }, { "name": "depth_confidence", "l2": 11.461466, "mean_abs": 0.214156, "relative": 0.00038 }, { "name": "video_fisheye_cam0", "l2": 5.074822, "mean_abs": 0.106703, "relative": 0.918942 }, { "name": "video_fisheye_cam1", "l2": 5.361502, "mean_abs": 0.115256, "relative": 0.885943 }, { "name": "video_fisheye_cam2", "l2": 5.508935, "mean_abs": 0.114288, "relative": 0.87942 }, { "name": "video_fisheye_cam3", "l2": 5.56354, "mean_abs": 0.115039, "relative": 0.931548 }, { "name": "video_stereo_left", "l2": 5.228404, "mean_abs": 0.118587, "relative": 0.849626 }, { "name": "video_stereo_right", "l2": 6.235792, "mean_abs": 0.143526, "relative": 0.879116 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.106953, "mean_abs": 0.083708, "relative": 0.754047 }, { "name": "caption_objects_interaction_text", "l2": 1.954153, "mean_abs": 0.018326, "relative": 0.8973 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 575, "start_frame": 2875, "end_frame": 2894, "center_frame": 2884, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.023284, "mean_abs": 0.221435, "relative": 0.845507 }, { "name": "hand_right_joints", "l2": 11.559294, "mean_abs": 0.291816, "relative": 0.911182 }, { "name": "body_joints", "l2": 5.907353, "mean_abs": 0.092607, "relative": 0.797901 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.155858, "mean_abs": 0.021786, "relative": 0.367804 }, { "name": "camera_rotation_matrix", "l2": 3.013588, "mean_abs": 0.21999, "relative": 0.925424 }, { "name": "imu_accel_gyro", "l2": 17.151495, "mean_abs": 1.087567, "relative": 0.925382 }, { "name": "depth_confidence", "l2": 11.306712, "mean_abs": 0.212935, "relative": 0.000375 }, { "name": "video_fisheye_cam0", "l2": 5.076684, "mean_abs": 0.107536, "relative": 0.919279 }, { "name": "video_fisheye_cam1", "l2": 5.334023, "mean_abs": 0.115615, "relative": 0.881402 }, { "name": "video_fisheye_cam2", "l2": 5.52067, "mean_abs": 0.115055, "relative": 0.881293 }, { "name": "video_fisheye_cam3", "l2": 5.545808, "mean_abs": 0.115342, "relative": 0.928579 }, { "name": "video_stereo_left", "l2": 5.279561, "mean_abs": 0.121399, "relative": 0.857939 }, { "name": "video_stereo_right", "l2": 6.307988, "mean_abs": 0.146502, "relative": 0.889294 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.095371, "mean_abs": 0.082601, "relative": 0.749902 }, { "name": "caption_objects_interaction_text", "l2": 1.954153, "mean_abs": 0.018355, "relative": 0.8973 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 576, "start_frame": 2880, "end_frame": 2899, "center_frame": 2889, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.955632, "mean_abs": 0.221238, "relative": 0.8398 }, { "name": "hand_right_joints", "l2": 11.749942, "mean_abs": 0.29577, "relative": 0.92621 }, { "name": "body_joints", "l2": 5.913837, "mean_abs": 0.092572, "relative": 0.798777 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.161448, "mean_abs": 0.022448, "relative": 0.380995 }, { "name": "camera_rotation_matrix", "l2": 3.013805, "mean_abs": 0.219857, "relative": 0.925491 }, { "name": "imu_accel_gyro", "l2": 17.147512, "mean_abs": 1.120928, "relative": 0.925167 }, { "name": "depth_confidence", "l2": 11.165165, "mean_abs": 0.209038, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.067856, "mean_abs": 0.1072, "relative": 0.91768 }, { "name": "video_fisheye_cam1", "l2": 5.296055, "mean_abs": 0.11475, "relative": 0.875128 }, { "name": "video_fisheye_cam2", "l2": 5.532298, "mean_abs": 0.114907, "relative": 0.883149 }, { "name": "video_fisheye_cam3", "l2": 5.526889, "mean_abs": 0.11486, "relative": 0.925411 }, { "name": "video_stereo_left", "l2": 5.266624, "mean_abs": 0.120612, "relative": 0.855837 }, { "name": "video_stereo_right", "l2": 6.31844, "mean_abs": 0.146116, "relative": 0.890767 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.089389, "mean_abs": 0.082628, "relative": 0.747761 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 577, "start_frame": 2885, "end_frame": 2904, "center_frame": 2894, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.924841, "mean_abs": 0.22129, "relative": 0.837203 }, { "name": "hand_right_joints", "l2": 11.948798, "mean_abs": 0.295987, "relative": 0.941885 }, { "name": "body_joints", "l2": 5.831911, "mean_abs": 0.091656, "relative": 0.787712 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.152368, "mean_abs": 0.02101, "relative": 0.359567 }, { "name": "camera_rotation_matrix", "l2": 3.010908, "mean_abs": 0.218162, "relative": 0.924601 }, { "name": "imu_accel_gyro", "l2": 17.108299, "mean_abs": 1.119475, "relative": 0.923052 }, { "name": "depth_confidence", "l2": 10.951529, "mean_abs": 0.204673, "relative": 0.000363 }, { "name": "video_fisheye_cam0", "l2": 5.070748, "mean_abs": 0.107091, "relative": 0.918204 }, { "name": "video_fisheye_cam1", "l2": 5.283212, "mean_abs": 0.114185, "relative": 0.873006 }, { "name": "video_fisheye_cam2", "l2": 5.554577, "mean_abs": 0.11475, "relative": 0.886706 }, { "name": "video_fisheye_cam3", "l2": 5.51559, "mean_abs": 0.114387, "relative": 0.923519 }, { "name": "video_stereo_left", "l2": 5.288895, "mean_abs": 0.119875, "relative": 0.859456 }, { "name": "video_stereo_right", "l2": 6.362752, "mean_abs": 0.145968, "relative": 0.897014 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.092551, "mean_abs": 0.082663, "relative": 0.748893 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 578, "start_frame": 2890, "end_frame": 2909, "center_frame": 2899, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.899107, "mean_abs": 0.221542, "relative": 0.835032 }, { "name": "hand_right_joints", "l2": 12.116664, "mean_abs": 0.297755, "relative": 0.955117 }, { "name": "body_joints", "l2": 5.76123, "mean_abs": 0.089944, "relative": 0.778165 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.127636, "mean_abs": 0.017117, "relative": 0.301204 }, { "name": "camera_rotation_matrix", "l2": 3.009476, "mean_abs": 0.215904, "relative": 0.924161 }, { "name": "imu_accel_gyro", "l2": 17.002535, "mean_abs": 1.12177, "relative": 0.917345 }, { "name": "depth_confidence", "l2": 10.814825, "mean_abs": 0.20159, "relative": 0.000359 }, { "name": "video_fisheye_cam0", "l2": 5.078431, "mean_abs": 0.107218, "relative": 0.919595 }, { "name": "video_fisheye_cam1", "l2": 5.241724, "mean_abs": 0.112926, "relative": 0.866151 }, { "name": "video_fisheye_cam2", "l2": 5.573841, "mean_abs": 0.114709, "relative": 0.889781 }, { "name": "video_fisheye_cam3", "l2": 5.459508, "mean_abs": 0.112676, "relative": 0.914129 }, { "name": "video_stereo_left", "l2": 5.314093, "mean_abs": 0.119861, "relative": 0.863551 }, { "name": "video_stereo_right", "l2": 6.388039, "mean_abs": 0.146325, "relative": 0.90058 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.100976, "mean_abs": 0.08249, "relative": 0.751908 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 579, "start_frame": 2895, "end_frame": 2914, "center_frame": 2904, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.900691, "mean_abs": 0.219935, "relative": 0.835166 }, { "name": "hand_right_joints", "l2": 12.113037, "mean_abs": 0.297146, "relative": 0.954831 }, { "name": "body_joints", "l2": 5.677301, "mean_abs": 0.088024, "relative": 0.766828 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.10204, "mean_abs": 0.013722, "relative": 0.240799 }, { "name": "camera_rotation_matrix", "l2": 3.001746, "mean_abs": 0.214429, "relative": 0.921788 }, { "name": "imu_accel_gyro", "l2": 17.12418, "mean_abs": 1.128415, "relative": 0.923908 }, { "name": "depth_confidence", "l2": 10.835094, "mean_abs": 0.201135, "relative": 0.00036 }, { "name": "video_fisheye_cam0", "l2": 5.082397, "mean_abs": 0.107425, "relative": 0.920313 }, { "name": "video_fisheye_cam1", "l2": 5.214921, "mean_abs": 0.112344, "relative": 0.861722 }, { "name": "video_fisheye_cam2", "l2": 5.589376, "mean_abs": 0.114687, "relative": 0.892261 }, { "name": "video_fisheye_cam3", "l2": 5.440823, "mean_abs": 0.111747, "relative": 0.911 }, { "name": "video_stereo_left", "l2": 5.327708, "mean_abs": 0.119769, "relative": 0.865763 }, { "name": "video_stereo_right", "l2": 6.384933, "mean_abs": 0.146716, "relative": 0.900142 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.143686, "mean_abs": 0.084568, "relative": 0.767193 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 580, "start_frame": 2900, "end_frame": 2919, "center_frame": 2909, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.918568, "mean_abs": 0.218203, "relative": 0.836673 }, { "name": "hand_right_joints", "l2": 11.993133, "mean_abs": 0.293431, "relative": 0.94538 }, { "name": "body_joints", "l2": 5.615803, "mean_abs": 0.086954, "relative": 0.758522 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.074711, "mean_abs": 0.010308, "relative": 0.176306 }, { "name": "camera_rotation_matrix", "l2": 3.002006, "mean_abs": 0.215134, "relative": 0.921867 }, { "name": "imu_accel_gyro", "l2": 17.113253, "mean_abs": 1.129987, "relative": 0.923319 }, { "name": "depth_confidence", "l2": 10.893829, "mean_abs": 0.200003, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.084884, "mean_abs": 0.107124, "relative": 0.920764 }, { "name": "video_fisheye_cam1", "l2": 5.195242, "mean_abs": 0.111193, "relative": 0.85847 }, { "name": "video_fisheye_cam2", "l2": 5.603056, "mean_abs": 0.114296, "relative": 0.894445 }, { "name": "video_fisheye_cam3", "l2": 5.450138, "mean_abs": 0.11124, "relative": 0.91256 }, { "name": "video_stereo_left", "l2": 5.301958, "mean_abs": 0.116356, "relative": 0.861579 }, { "name": "video_stereo_right", "l2": 6.318476, "mean_abs": 0.142034, "relative": 0.890773 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.167324, "mean_abs": 0.085803, "relative": 0.775653 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 581, "start_frame": 2905, "end_frame": 2924, "center_frame": 2914, "action": "Grasp gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.935592, "mean_abs": 0.218159, "relative": 0.83811 }, { "name": "hand_right_joints", "l2": 11.805133, "mean_abs": 0.290344, "relative": 0.93056 }, { "name": "body_joints", "l2": 5.580066, "mean_abs": 0.085865, "relative": 0.753695 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.0704, "mean_abs": 0.010123, "relative": 0.166135 }, { "name": "camera_rotation_matrix", "l2": 3.00604, "mean_abs": 0.218835, "relative": 0.923106 }, { "name": "imu_accel_gyro", "l2": 17.121965, "mean_abs": 1.142007, "relative": 0.923789 }, { "name": "depth_confidence", "l2": 10.973387, "mean_abs": 0.201958, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.075488, "mean_abs": 0.106947, "relative": 0.919062 }, { "name": "video_fisheye_cam1", "l2": 5.205159, "mean_abs": 0.112263, "relative": 0.860109 }, { "name": "video_fisheye_cam2", "l2": 5.613916, "mean_abs": 0.114955, "relative": 0.896178 }, { "name": "video_fisheye_cam3", "l2": 5.453198, "mean_abs": 0.111132, "relative": 0.913072 }, { "name": "video_stereo_left", "l2": 5.279009, "mean_abs": 0.115866, "relative": 0.857849 }, { "name": "video_stereo_right", "l2": 6.300679, "mean_abs": 0.141063, "relative": 0.888264 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.11053, "mean_abs": 0.084243, "relative": 0.755327 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 582, "start_frame": 2910, "end_frame": 2929, "center_frame": 2919, "action": "", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.942366, "mean_abs": 0.217726, "relative": 0.838681 }, { "name": "hand_right_joints", "l2": 11.649948, "mean_abs": 0.289518, "relative": 0.918328 }, { "name": "body_joints", "l2": 5.61658, "mean_abs": 0.087598, "relative": 0.758627 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.082049, "mean_abs": 0.012705, "relative": 0.193625 }, { "name": "camera_rotation_matrix", "l2": 3.00818, "mean_abs": 0.224016, "relative": 0.923763 }, { "name": "imu_accel_gyro", "l2": 17.266844, "mean_abs": 1.143773, "relative": 0.931606 }, { "name": "depth_confidence", "l2": 11.095869, "mean_abs": 0.205493, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.062318, "mean_abs": 0.106721, "relative": 0.916677 }, { "name": "video_fisheye_cam1", "l2": 5.240756, "mean_abs": 0.114496, "relative": 0.865991 }, { "name": "video_fisheye_cam2", "l2": 5.636144, "mean_abs": 0.116298, "relative": 0.899727 }, { "name": "video_fisheye_cam3", "l2": 5.504277, "mean_abs": 0.11319, "relative": 0.921625 }, { "name": "video_stereo_left", "l2": 5.280551, "mean_abs": 0.117861, "relative": 0.8581 }, { "name": "video_stereo_right", "l2": 6.348325, "mean_abs": 0.144868, "relative": 0.894981 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.194719, "mean_abs": 0.087647, "relative": 0.785457 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 583, "start_frame": 2915, "end_frame": 2934, "center_frame": 2924, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.979706, "mean_abs": 0.218526, "relative": 0.841831 }, { "name": "hand_right_joints", "l2": 11.459387, "mean_abs": 0.285151, "relative": 0.903306 }, { "name": "body_joints", "l2": 5.655914, "mean_abs": 0.088932, "relative": 0.76394 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.11379, "mean_abs": 0.017276, "relative": 0.268529 }, { "name": "camera_rotation_matrix", "l2": 3.011528, "mean_abs": 0.225767, "relative": 0.924791 }, { "name": "imu_accel_gyro", "l2": 17.460655, "mean_abs": 1.181195, "relative": 0.942062 }, { "name": "depth_confidence", "l2": 11.233564, "mean_abs": 0.209655, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.053166, "mean_abs": 0.106528, "relative": 0.91502 }, { "name": "video_fisheye_cam1", "l2": 5.304133, "mean_abs": 0.116034, "relative": 0.876463 }, { "name": "video_fisheye_cam2", "l2": 5.632207, "mean_abs": 0.11653, "relative": 0.899098 }, { "name": "video_fisheye_cam3", "l2": 5.645637, "mean_abs": 0.116967, "relative": 0.945294 }, { "name": "video_stereo_left", "l2": 5.256424, "mean_abs": 0.118921, "relative": 0.854179 }, { "name": "video_stereo_right", "l2": 6.370318, "mean_abs": 0.147412, "relative": 0.898081 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.192617, "mean_abs": 0.08549, "relative": 0.784705 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 584, "start_frame": 2920, "end_frame": 2939, "center_frame": 2929, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.98081, "mean_abs": 0.218379, "relative": 0.841924 }, { "name": "hand_right_joints", "l2": 11.24875, "mean_abs": 0.279247, "relative": 0.886702 }, { "name": "body_joints", "l2": 5.719718, "mean_abs": 0.091136, "relative": 0.772558 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.15972, "mean_abs": 0.02297, "relative": 0.376916 }, { "name": "camera_rotation_matrix", "l2": 3.009494, "mean_abs": 0.227356, "relative": 0.924167 }, { "name": "imu_accel_gyro", "l2": 17.518118, "mean_abs": 1.208217, "relative": 0.945163 }, { "name": "depth_confidence", "l2": 11.315121, "mean_abs": 0.212795, "relative": 0.000376 }, { "name": "video_fisheye_cam0", "l2": 5.07408, "mean_abs": 0.108318, "relative": 0.918807 }, { "name": "video_fisheye_cam1", "l2": 5.389495, "mean_abs": 0.11858, "relative": 0.890569 }, { "name": "video_fisheye_cam2", "l2": 5.617704, "mean_abs": 0.117103, "relative": 0.896783 }, { "name": "video_fisheye_cam3", "l2": 5.767519, "mean_abs": 0.120408, "relative": 0.965702 }, { "name": "video_stereo_left", "l2": 5.259548, "mean_abs": 0.120512, "relative": 0.854687 }, { "name": "video_stereo_right", "l2": 6.428042, "mean_abs": 0.150234, "relative": 0.906219 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.137932, "mean_abs": 0.085959, "relative": 0.765134 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 585, "start_frame": 2925, "end_frame": 2944, "center_frame": 2934, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.978771, "mean_abs": 0.219678, "relative": 0.841752 }, { "name": "hand_right_joints", "l2": 11.018482, "mean_abs": 0.275671, "relative": 0.868551 }, { "name": "body_joints", "l2": 5.784046, "mean_abs": 0.091364, "relative": 0.781246 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.180226, "mean_abs": 0.022264, "relative": 0.425307 }, { "name": "camera_rotation_matrix", "l2": 3.007054, "mean_abs": 0.22648, "relative": 0.923417 }, { "name": "imu_accel_gyro", "l2": 17.376013, "mean_abs": 1.147944, "relative": 0.937496 }, { "name": "depth_confidence", "l2": 11.357531, "mean_abs": 0.213574, "relative": 0.000377 }, { "name": "video_fisheye_cam0", "l2": 5.089083, "mean_abs": 0.109074, "relative": 0.921524 }, { "name": "video_fisheye_cam1", "l2": 5.444706, "mean_abs": 0.119151, "relative": 0.899692 }, { "name": "video_fisheye_cam2", "l2": 5.604431, "mean_abs": 0.116913, "relative": 0.894664 }, { "name": "video_fisheye_cam3", "l2": 5.827141, "mean_abs": 0.121187, "relative": 0.975685 }, { "name": "video_stereo_left", "l2": 5.243134, "mean_abs": 0.119999, "relative": 0.85202 }, { "name": "video_stereo_right", "l2": 6.437004, "mean_abs": 0.150072, "relative": 0.907483 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.079309, "mean_abs": 0.083682, "relative": 0.744153 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 586, "start_frame": 2930, "end_frame": 2949, "center_frame": 2939, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.991563, "mean_abs": 0.218024, "relative": 0.842831 }, { "name": "hand_right_joints", "l2": 10.942936, "mean_abs": 0.274553, "relative": 0.862596 }, { "name": "body_joints", "l2": 5.874323, "mean_abs": 0.09106, "relative": 0.79344 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.167617, "mean_abs": 0.019551, "relative": 0.395552 }, { "name": "camera_rotation_matrix", "l2": 3.004767, "mean_abs": 0.221735, "relative": 0.922715 }, { "name": "imu_accel_gyro", "l2": 17.28194, "mean_abs": 1.110747, "relative": 0.93242 }, { "name": "depth_confidence", "l2": 11.399239, "mean_abs": 0.212944, "relative": 0.000378 }, { "name": "video_fisheye_cam0", "l2": 5.088675, "mean_abs": 0.108385, "relative": 0.92145 }, { "name": "video_fisheye_cam1", "l2": 5.479104, "mean_abs": 0.118468, "relative": 0.905376 }, { "name": "video_fisheye_cam2", "l2": 5.584115, "mean_abs": 0.115801, "relative": 0.891421 }, { "name": "video_fisheye_cam3", "l2": 5.868826, "mean_abs": 0.120704, "relative": 0.982664 }, { "name": "video_stereo_left", "l2": 5.222771, "mean_abs": 0.118704, "relative": 0.848711 }, { "name": "video_stereo_right", "l2": 6.426334, "mean_abs": 0.148165, "relative": 0.905978 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.093979, "mean_abs": 0.081402, "relative": 0.749404 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 587, "start_frame": 2935, "end_frame": 2954, "center_frame": 2944, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.948577, "mean_abs": 0.213488, "relative": 0.839205 }, { "name": "hand_right_joints", "l2": 10.961928, "mean_abs": 0.277245, "relative": 0.864093 }, { "name": "body_joints", "l2": 5.954764, "mean_abs": 0.089871, "relative": 0.804305 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.119724, "mean_abs": 0.017615, "relative": 0.282533 }, { "name": "camera_rotation_matrix", "l2": 3.00418, "mean_abs": 0.216629, "relative": 0.922535 }, { "name": "imu_accel_gyro", "l2": 17.244247, "mean_abs": 1.083431, "relative": 0.930387 }, { "name": "depth_confidence", "l2": 11.344962, "mean_abs": 0.21106, "relative": 0.000377 }, { "name": "video_fisheye_cam0", "l2": 5.095844, "mean_abs": 0.107205, "relative": 0.922748 }, { "name": "video_fisheye_cam1", "l2": 5.501311, "mean_abs": 0.117435, "relative": 0.909045 }, { "name": "video_fisheye_cam2", "l2": 5.549468, "mean_abs": 0.114167, "relative": 0.88589 }, { "name": "video_fisheye_cam3", "l2": 5.877436, "mean_abs": 0.121468, "relative": 0.984106 }, { "name": "video_stereo_left", "l2": 5.210248, "mean_abs": 0.116791, "relative": 0.846676 }, { "name": "video_stereo_right", "l2": 6.405816, "mean_abs": 0.14616, "relative": 0.903086 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.064763, "mean_abs": 0.081729, "relative": 0.738948 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 588, "start_frame": 2940, "end_frame": 2959, "center_frame": 2949, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.923494, "mean_abs": 0.209844, "relative": 0.837089 }, { "name": "hand_right_joints", "l2": 11.019923, "mean_abs": 0.279843, "relative": 0.868665 }, { "name": "body_joints", "l2": 6.026238, "mean_abs": 0.089495, "relative": 0.813959 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.079432, "mean_abs": 0.011737, "relative": 0.187448 }, { "name": "camera_rotation_matrix", "l2": 3.003941, "mean_abs": 0.216114, "relative": 0.922461 }, { "name": "imu_accel_gyro", "l2": 17.168941, "mean_abs": 1.067442, "relative": 0.926324 }, { "name": "depth_confidence", "l2": 11.245295, "mean_abs": 0.209158, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.119732, "mean_abs": 0.107659, "relative": 0.927074 }, { "name": "video_fisheye_cam1", "l2": 5.500455, "mean_abs": 0.118176, "relative": 0.908904 }, { "name": "video_fisheye_cam2", "l2": 5.542724, "mean_abs": 0.114333, "relative": 0.884813 }, { "name": "video_fisheye_cam3", "l2": 5.859808, "mean_abs": 0.121958, "relative": 0.981154 }, { "name": "video_stereo_left", "l2": 5.227208, "mean_abs": 0.11789, "relative": 0.849432 }, { "name": "video_stereo_right", "l2": 6.441679, "mean_abs": 0.14793, "relative": 0.908141 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.216185, "mean_abs": 0.086806, "relative": 0.793139 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013393, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 589, "start_frame": 2945, "end_frame": 2964, "center_frame": 2954, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.906068, "mean_abs": 0.209646, "relative": 0.835619 }, { "name": "hand_right_joints", "l2": 11.096916, "mean_abs": 0.279754, "relative": 0.874734 }, { "name": "body_joints", "l2": 6.094361, "mean_abs": 0.088266, "relative": 0.82316 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.067605, "mean_abs": 0.00951, "relative": 0.159538 }, { "name": "camera_rotation_matrix", "l2": 3.002266, "mean_abs": 0.214241, "relative": 0.921947 }, { "name": "imu_accel_gyro", "l2": 17.092577, "mean_abs": 1.015213, "relative": 0.922203 }, { "name": "depth_confidence", "l2": 11.211993, "mean_abs": 0.207644, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.120317, "mean_abs": 0.107061, "relative": 0.92718 }, { "name": "video_fisheye_cam1", "l2": 5.475406, "mean_abs": 0.117104, "relative": 0.904765 }, { "name": "video_fisheye_cam2", "l2": 5.542763, "mean_abs": 0.113528, "relative": 0.88482 }, { "name": "video_fisheye_cam3", "l2": 5.773551, "mean_abs": 0.119528, "relative": 0.966711 }, { "name": "video_stereo_left", "l2": 5.217798, "mean_abs": 0.115993, "relative": 0.847903 }, { "name": "video_stereo_right", "l2": 6.440648, "mean_abs": 0.146134, "relative": 0.907996 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.260617, "mean_abs": 0.091508, "relative": 0.809041 }, { "name": "caption_objects_interaction_text", "l2": 1.879976, "mean_abs": 0.01779, "relative": 0.86324 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 590, "start_frame": 2950, "end_frame": 2969, "center_frame": 2959, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.910288, "mean_abs": 0.210044, "relative": 0.835975 }, { "name": "hand_right_joints", "l2": 11.123313, "mean_abs": 0.282175, "relative": 0.876815 }, { "name": "body_joints", "l2": 6.120937, "mean_abs": 0.087029, "relative": 0.82675 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.051453, "mean_abs": 0.007535, "relative": 0.121422 }, { "name": "camera_rotation_matrix", "l2": 3.00074, "mean_abs": 0.210928, "relative": 0.921478 }, { "name": "imu_accel_gyro", "l2": 17.113985, "mean_abs": 1.005051, "relative": 0.923358 }, { "name": "depth_confidence", "l2": 11.13396, "mean_abs": 0.205275, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.119347, "mean_abs": 0.106349, "relative": 0.927004 }, { "name": "video_fisheye_cam1", "l2": 5.436029, "mean_abs": 0.11519, "relative": 0.898258 }, { "name": "video_fisheye_cam2", "l2": 5.54822, "mean_abs": 0.112724, "relative": 0.885691 }, { "name": "video_fisheye_cam3", "l2": 5.695319, "mean_abs": 0.116435, "relative": 0.953612 }, { "name": "video_stereo_left", "l2": 5.223663, "mean_abs": 0.114308, "relative": 0.848856 }, { "name": "video_stereo_right", "l2": 6.440893, "mean_abs": 0.143559, "relative": 0.908031 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.252484, "mean_abs": 0.090039, "relative": 0.80613 }, { "name": "caption_objects_interaction_text", "l2": 1.879976, "mean_abs": 0.017768, "relative": 0.86324 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 591, "start_frame": 2955, "end_frame": 2974, "center_frame": 2964, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "water bottle", "white cup" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.918545, "mean_abs": 0.209836, "relative": 0.836672 }, { "name": "hand_right_joints", "l2": 11.100628, "mean_abs": 0.281309, "relative": 0.875026 }, { "name": "body_joints", "l2": 6.126183, "mean_abs": 0.086205, "relative": 0.827459 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.039465, "mean_abs": 0.006133, "relative": 0.093131 }, { "name": "camera_rotation_matrix", "l2": 3.000977, "mean_abs": 0.209339, "relative": 0.921551 }, { "name": "imu_accel_gyro", "l2": 17.061182, "mean_abs": 0.997507, "relative": 0.92051 }, { "name": "depth_confidence", "l2": 11.078757, "mean_abs": 0.20357, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.126303, "mean_abs": 0.106315, "relative": 0.928264 }, { "name": "video_fisheye_cam1", "l2": 5.419355, "mean_abs": 0.114549, "relative": 0.895503 }, { "name": "video_fisheye_cam2", "l2": 5.554071, "mean_abs": 0.112462, "relative": 0.886625 }, { "name": "video_fisheye_cam3", "l2": 5.665651, "mean_abs": 0.114661, "relative": 0.948645 }, { "name": "video_stereo_left", "l2": 5.238971, "mean_abs": 0.113959, "relative": 0.851343 }, { "name": "video_stereo_right", "l2": 6.458843, "mean_abs": 0.143245, "relative": 0.910561 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.187302, "mean_abs": 0.085118, "relative": 0.782803 }, { "name": "caption_objects_interaction_text", "l2": 1.879976, "mean_abs": 0.017488, "relative": 0.86324 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 592, "start_frame": 2960, "end_frame": 2979, "center_frame": 2969, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.930024, "mean_abs": 0.20888, "relative": 0.83764 }, { "name": "hand_right_joints", "l2": 11.11805, "mean_abs": 0.280681, "relative": 0.8764 }, { "name": "body_joints", "l2": 6.107988, "mean_abs": 0.085336, "relative": 0.825001 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.032314, "mean_abs": 0.004627, "relative": 0.076256 }, { "name": "camera_rotation_matrix", "l2": 3.00145, "mean_abs": 0.20848, "relative": 0.921697 }, { "name": "imu_accel_gyro", "l2": 17.068331, "mean_abs": 0.987134, "relative": 0.920895 }, { "name": "depth_confidence", "l2": 11.05711, "mean_abs": 0.202821, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.133774, "mean_abs": 0.10634, "relative": 0.929617 }, { "name": "video_fisheye_cam1", "l2": 5.406536, "mean_abs": 0.114199, "relative": 0.893384 }, { "name": "video_fisheye_cam2", "l2": 5.556618, "mean_abs": 0.112397, "relative": 0.887031 }, { "name": "video_fisheye_cam3", "l2": 5.626797, "mean_abs": 0.113871, "relative": 0.942139 }, { "name": "video_stereo_left", "l2": 5.24988, "mean_abs": 0.11353, "relative": 0.853116 }, { "name": "video_stereo_right", "l2": 6.471568, "mean_abs": 0.142783, "relative": 0.912355 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.091655, "mean_abs": 0.082095, "relative": 0.748572 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011421, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 593, "start_frame": 2965, "end_frame": 2984, "center_frame": 2974, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.951982, "mean_abs": 0.205999, "relative": 0.839492 }, { "name": "hand_right_joints", "l2": 11.094027, "mean_abs": 0.278001, "relative": 0.874506 }, { "name": "body_joints", "l2": 6.065163, "mean_abs": 0.083957, "relative": 0.819217 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028814, "mean_abs": 0.004076, "relative": 0.067998 }, { "name": "camera_rotation_matrix", "l2": 3.000175, "mean_abs": 0.207578, "relative": 0.921305 }, { "name": "imu_accel_gyro", "l2": 17.091431, "mean_abs": 0.982735, "relative": 0.922142 }, { "name": "depth_confidence", "l2": 11.078154, "mean_abs": 0.202525, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.137793, "mean_abs": 0.106337, "relative": 0.930344 }, { "name": "video_fisheye_cam1", "l2": 5.404896, "mean_abs": 0.114075, "relative": 0.893113 }, { "name": "video_fisheye_cam2", "l2": 5.556736, "mean_abs": 0.112375, "relative": 0.88705 }, { "name": "video_fisheye_cam3", "l2": 5.631413, "mean_abs": 0.113809, "relative": 0.942912 }, { "name": "video_stereo_left", "l2": 5.258276, "mean_abs": 0.11311, "relative": 0.85448 }, { "name": "video_stereo_right", "l2": 6.488375, "mean_abs": 0.142567, "relative": 0.914725 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.135736, "mean_abs": 0.083348, "relative": 0.764348 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011421, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 594, "start_frame": 2970, "end_frame": 2989, "center_frame": 2979, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.960517, "mean_abs": 0.20339, "relative": 0.840212 }, { "name": "hand_right_joints", "l2": 11.073523, "mean_abs": 0.276243, "relative": 0.87289 }, { "name": "body_joints", "l2": 6.029208, "mean_abs": 0.082865, "relative": 0.81436 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.025446, "mean_abs": 0.003801, "relative": 0.06005 }, { "name": "camera_rotation_matrix", "l2": 2.999506, "mean_abs": 0.206461, "relative": 0.9211 }, { "name": "imu_accel_gyro", "l2": 17.062647, "mean_abs": 0.985947, "relative": 0.920589 }, { "name": "depth_confidence", "l2": 11.058916, "mean_abs": 0.20177, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.142758, "mean_abs": 0.106407, "relative": 0.931244 }, { "name": "video_fisheye_cam1", "l2": 5.408549, "mean_abs": 0.11386, "relative": 0.893717 }, { "name": "video_fisheye_cam2", "l2": 5.556358, "mean_abs": 0.112149, "relative": 0.88699 }, { "name": "video_fisheye_cam3", "l2": 5.620244, "mean_abs": 0.113138, "relative": 0.941042 }, { "name": "video_stereo_left", "l2": 5.267127, "mean_abs": 0.112798, "relative": 0.855919 }, { "name": "video_stereo_right", "l2": 6.50422, "mean_abs": 0.142324, "relative": 0.916959 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.027997, "mean_abs": 0.078442, "relative": 0.72579 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011421, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 595, "start_frame": 2975, "end_frame": 2994, "center_frame": 2984, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.971259, "mean_abs": 0.202089, "relative": 0.841118 }, { "name": "hand_right_joints", "l2": 11.115383, "mean_abs": 0.275708, "relative": 0.87619 }, { "name": "body_joints", "l2": 6.005125, "mean_abs": 0.08212, "relative": 0.811107 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01577, "mean_abs": 0.002227, "relative": 0.037214 }, { "name": "camera_rotation_matrix", "l2": 3.000115, "mean_abs": 0.206577, "relative": 0.921287 }, { "name": "imu_accel_gyro", "l2": 17.072248, "mean_abs": 0.979686, "relative": 0.921107 }, { "name": "depth_confidence", "l2": 11.061787, "mean_abs": 0.20147, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.142752, "mean_abs": 0.106411, "relative": 0.931242 }, { "name": "video_fisheye_cam1", "l2": 5.409514, "mean_abs": 0.114098, "relative": 0.893876 }, { "name": "video_fisheye_cam2", "l2": 5.550306, "mean_abs": 0.112128, "relative": 0.886024 }, { "name": "video_fisheye_cam3", "l2": 5.604897, "mean_abs": 0.112905, "relative": 0.938472 }, { "name": "video_stereo_left", "l2": 5.268493, "mean_abs": 0.11269, "relative": 0.856141 }, { "name": "video_stereo_right", "l2": 6.508666, "mean_abs": 0.142491, "relative": 0.917585 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.102131, "mean_abs": 0.079905, "relative": 0.752321 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011421, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 596, "start_frame": 2980, "end_frame": 2999, "center_frame": 2989, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.974473, "mean_abs": 0.202335, "relative": 0.841389 }, { "name": "hand_right_joints", "l2": 11.089103, "mean_abs": 0.275252, "relative": 0.874118 }, { "name": "body_joints", "l2": 5.986389, "mean_abs": 0.081825, "relative": 0.808577 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011969, "mean_abs": 0.001689, "relative": 0.028245 }, { "name": "camera_rotation_matrix", "l2": 3.000867, "mean_abs": 0.208048, "relative": 0.921518 }, { "name": "imu_accel_gyro", "l2": 17.088161, "mean_abs": 0.976951, "relative": 0.921965 }, { "name": "depth_confidence", "l2": 11.019107, "mean_abs": 0.202478, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.141526, "mean_abs": 0.106739, "relative": 0.93102 }, { "name": "video_fisheye_cam1", "l2": 5.418155, "mean_abs": 0.114995, "relative": 0.895304 }, { "name": "video_fisheye_cam2", "l2": 5.541138, "mean_abs": 0.112803, "relative": 0.88456 }, { "name": "video_fisheye_cam3", "l2": 5.602352, "mean_abs": 0.113371, "relative": 0.938046 }, { "name": "video_stereo_left", "l2": 5.257885, "mean_abs": 0.113855, "relative": 0.854417 }, { "name": "video_stereo_right", "l2": 6.507215, "mean_abs": 0.144108, "relative": 0.917381 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.085405, "mean_abs": 0.080226, "relative": 0.746335 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011421, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 597, "start_frame": 2985, "end_frame": 3004, "center_frame": 2994, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.970477, "mean_abs": 0.201648, "relative": 0.841052 }, { "name": "hand_right_joints", "l2": 11.078213, "mean_abs": 0.273599, "relative": 0.87326 }, { "name": "body_joints", "l2": 5.970657, "mean_abs": 0.082043, "relative": 0.806452 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020877, "mean_abs": 0.002279, "relative": 0.049266 }, { "name": "camera_rotation_matrix", "l2": 3.001545, "mean_abs": 0.208259, "relative": 0.921726 }, { "name": "imu_accel_gyro", "l2": 17.076359, "mean_abs": 0.975269, "relative": 0.921328 }, { "name": "depth_confidence", "l2": 10.993114, "mean_abs": 0.201946, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.140216, "mean_abs": 0.107016, "relative": 0.930783 }, { "name": "video_fisheye_cam1", "l2": 5.425459, "mean_abs": 0.115389, "relative": 0.896511 }, { "name": "video_fisheye_cam2", "l2": 5.530916, "mean_abs": 0.112966, "relative": 0.882928 }, { "name": "video_fisheye_cam3", "l2": 5.5877, "mean_abs": 0.113827, "relative": 0.935593 }, { "name": "video_stereo_left", "l2": 5.247627, "mean_abs": 0.114338, "relative": 0.85275 }, { "name": "video_stereo_right", "l2": 6.502568, "mean_abs": 0.144878, "relative": 0.916726 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.075847, "mean_abs": 0.081077, "relative": 0.742915 }, { "name": "caption_objects_interaction_text", "l2": 1.923904, "mean_abs": 0.016903, "relative": 0.88341 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 598, "start_frame": 2990, "end_frame": 3009, "center_frame": 2999, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.964108, "mean_abs": 0.201193, "relative": 0.840515 }, { "name": "hand_right_joints", "l2": 11.079162, "mean_abs": 0.273199, "relative": 0.873334 }, { "name": "body_joints", "l2": 5.972972, "mean_abs": 0.082261, "relative": 0.806764 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028377, "mean_abs": 0.003179, "relative": 0.066965 }, { "name": "camera_rotation_matrix", "l2": 3.001682, "mean_abs": 0.207991, "relative": 0.921768 }, { "name": "imu_accel_gyro", "l2": 17.083565, "mean_abs": 0.977992, "relative": 0.921717 }, { "name": "depth_confidence", "l2": 10.961614, "mean_abs": 0.202066, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.137717, "mean_abs": 0.107012, "relative": 0.930331 }, { "name": "video_fisheye_cam1", "l2": 5.432135, "mean_abs": 0.115311, "relative": 0.897614 }, { "name": "video_fisheye_cam2", "l2": 5.51537, "mean_abs": 0.112603, "relative": 0.880447 }, { "name": "video_fisheye_cam3", "l2": 5.573489, "mean_abs": 0.113672, "relative": 0.933213 }, { "name": "video_stereo_left", "l2": 5.223874, "mean_abs": 0.11432, "relative": 0.84889 }, { "name": "video_stereo_right", "l2": 6.481711, "mean_abs": 0.144554, "relative": 0.913785 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.051028, "mean_abs": 0.078119, "relative": 0.734032 }, { "name": "caption_objects_interaction_text", "l2": 1.923904, "mean_abs": 0.017148, "relative": 0.88341 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 599, "start_frame": 2995, "end_frame": 3014, "center_frame": 3004, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.930676, "mean_abs": 0.201327, "relative": 0.837695 }, { "name": "hand_right_joints", "l2": 11.064374, "mean_abs": 0.271883, "relative": 0.872169 }, { "name": "body_joints", "l2": 5.97997, "mean_abs": 0.08268, "relative": 0.80771 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.024561, "mean_abs": 0.002979, "relative": 0.057961 }, { "name": "camera_rotation_matrix", "l2": 3.001477, "mean_abs": 0.207117, "relative": 0.921705 }, { "name": "imu_accel_gyro", "l2": 17.063911, "mean_abs": 0.974109, "relative": 0.920657 }, { "name": "depth_confidence", "l2": 10.886938, "mean_abs": 0.201039, "relative": 0.000361 }, { "name": "video_fisheye_cam0", "l2": 5.13305, "mean_abs": 0.106763, "relative": 0.929486 }, { "name": "video_fisheye_cam1", "l2": 5.436522, "mean_abs": 0.115181, "relative": 0.898339 }, { "name": "video_fisheye_cam2", "l2": 5.498626, "mean_abs": 0.111948, "relative": 0.877774 }, { "name": "video_fisheye_cam3", "l2": 5.547747, "mean_abs": 0.113043, "relative": 0.928903 }, { "name": "video_stereo_left", "l2": 5.196854, "mean_abs": 0.113789, "relative": 0.844499 }, { "name": "video_stereo_right", "l2": 6.454734, "mean_abs": 0.143654, "relative": 0.909982 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.97751, "mean_abs": 0.075587, "relative": 0.707721 }, { "name": "caption_objects_interaction_text", "l2": 1.923904, "mean_abs": 0.016978, "relative": 0.88341 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 600, "start_frame": 3000, "end_frame": 3019, "center_frame": 3009, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.9092, "mean_abs": 0.20119, "relative": 0.835883 }, { "name": "hand_right_joints", "l2": 11.076548, "mean_abs": 0.273014, "relative": 0.873128 }, { "name": "body_joints", "l2": 6.009944, "mean_abs": 0.083128, "relative": 0.811758 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020069, "mean_abs": 0.002331, "relative": 0.04736 }, { "name": "camera_rotation_matrix", "l2": 3.001151, "mean_abs": 0.205991, "relative": 0.921605 }, { "name": "imu_accel_gyro", "l2": 17.065096, "mean_abs": 0.977437, "relative": 0.920721 }, { "name": "depth_confidence", "l2": 10.899689, "mean_abs": 0.200322, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.13194, "mean_abs": 0.106545, "relative": 0.929285 }, { "name": "video_fisheye_cam1", "l2": 5.439841, "mean_abs": 0.114889, "relative": 0.898888 }, { "name": "video_fisheye_cam2", "l2": 5.47703, "mean_abs": 0.111271, "relative": 0.874326 }, { "name": "video_fisheye_cam3", "l2": 5.537668, "mean_abs": 0.112546, "relative": 0.927216 }, { "name": "video_stereo_left", "l2": 5.169689, "mean_abs": 0.112864, "relative": 0.840085 }, { "name": "video_stereo_right", "l2": 6.426826, "mean_abs": 0.142294, "relative": 0.906048 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.028873, "mean_abs": 0.0765, "relative": 0.726103 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 601, "start_frame": 3005, "end_frame": 3024, "center_frame": 3014, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.894696, "mean_abs": 0.2016, "relative": 0.83466 }, { "name": "hand_right_joints", "l2": 11.0623, "mean_abs": 0.27547, "relative": 0.872005 }, { "name": "body_joints", "l2": 6.042836, "mean_abs": 0.083494, "relative": 0.816201 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013633, "mean_abs": 0.001718, "relative": 0.032172 }, { "name": "camera_rotation_matrix", "l2": 3.00065, "mean_abs": 0.204533, "relative": 0.921451 }, { "name": "imu_accel_gyro", "l2": 17.078928, "mean_abs": 0.977631, "relative": 0.921467 }, { "name": "depth_confidence", "l2": 10.889791, "mean_abs": 0.199792, "relative": 0.000361 }, { "name": "video_fisheye_cam0", "l2": 5.130543, "mean_abs": 0.106392, "relative": 0.929032 }, { "name": "video_fisheye_cam1", "l2": 5.434694, "mean_abs": 0.114518, "relative": 0.898037 }, { "name": "video_fisheye_cam2", "l2": 5.460255, "mean_abs": 0.110701, "relative": 0.871648 }, { "name": "video_fisheye_cam3", "l2": 5.526175, "mean_abs": 0.112596, "relative": 0.925291 }, { "name": "video_stereo_left", "l2": 5.146862, "mean_abs": 0.112121, "relative": 0.836375 }, { "name": "video_stereo_right", "l2": 6.405809, "mean_abs": 0.141241, "relative": 0.903085 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.085907, "mean_abs": 0.079939, "relative": 0.746515 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 602, "start_frame": 3010, "end_frame": 3029, "center_frame": 3019, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.863124, "mean_abs": 0.201957, "relative": 0.831997 }, { "name": "hand_right_joints", "l2": 11.03424, "mean_abs": 0.276809, "relative": 0.869793 }, { "name": "body_joints", "l2": 6.079621, "mean_abs": 0.084276, "relative": 0.821169 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005413, "mean_abs": 0.000798, "relative": 0.012774 }, { "name": "camera_rotation_matrix", "l2": 3.000172, "mean_abs": 0.202766, "relative": 0.921304 }, { "name": "imu_accel_gyro", "l2": 17.059858, "mean_abs": 0.974916, "relative": 0.920438 }, { "name": "depth_confidence", "l2": 10.903135, "mean_abs": 0.199753, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.124366, "mean_abs": 0.105866, "relative": 0.927913 }, { "name": "video_fisheye_cam1", "l2": 5.430432, "mean_abs": 0.113937, "relative": 0.897333 }, { "name": "video_fisheye_cam2", "l2": 5.452216, "mean_abs": 0.110097, "relative": 0.870365 }, { "name": "video_fisheye_cam3", "l2": 5.511972, "mean_abs": 0.111824, "relative": 0.922913 }, { "name": "video_stereo_left", "l2": 5.128661, "mean_abs": 0.11054, "relative": 0.833418 }, { "name": "video_stereo_right", "l2": 6.380684, "mean_abs": 0.139489, "relative": 0.899543 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.100291, "mean_abs": 0.080396, "relative": 0.751663 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 603, "start_frame": 3015, "end_frame": 3034, "center_frame": 3024, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.852378, "mean_abs": 0.201755, "relative": 0.83109 }, { "name": "hand_right_joints", "l2": 10.978513, "mean_abs": 0.278926, "relative": 0.8654 }, { "name": "body_joints", "l2": 6.130692, "mean_abs": 0.085317, "relative": 0.828068 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007764, "mean_abs": 0.001114, "relative": 0.018321 }, { "name": "camera_rotation_matrix", "l2": 2.999961, "mean_abs": 0.201179, "relative": 0.921239 }, { "name": "imu_accel_gyro", "l2": 17.054478, "mean_abs": 0.964121, "relative": 0.920148 }, { "name": "depth_confidence", "l2": 10.899073, "mean_abs": 0.199094, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.121599, "mean_abs": 0.105484, "relative": 0.927412 }, { "name": "video_fisheye_cam1", "l2": 5.425202, "mean_abs": 0.11336, "relative": 0.896469 }, { "name": "video_fisheye_cam2", "l2": 5.464244, "mean_abs": 0.11033, "relative": 0.872285 }, { "name": "video_fisheye_cam3", "l2": 5.505257, "mean_abs": 0.111474, "relative": 0.921789 }, { "name": "video_stereo_left", "l2": 5.134915, "mean_abs": 0.110283, "relative": 0.834434 }, { "name": "video_stereo_right", "l2": 6.364495, "mean_abs": 0.138319, "relative": 0.89726 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.165146, "mean_abs": 0.080813, "relative": 0.774873 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 604, "start_frame": 3020, "end_frame": 3039, "center_frame": 3029, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.809222, "mean_abs": 0.200852, "relative": 0.82745 }, { "name": "hand_right_joints", "l2": 10.975862, "mean_abs": 0.279381, "relative": 0.865192 }, { "name": "body_joints", "l2": 6.172046, "mean_abs": 0.085551, "relative": 0.833653 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011739, "mean_abs": 0.001577, "relative": 0.027702 }, { "name": "camera_rotation_matrix", "l2": 3.000052, "mean_abs": 0.201847, "relative": 0.921267 }, { "name": "imu_accel_gyro", "l2": 17.05208, "mean_abs": 0.963663, "relative": 0.920018 }, { "name": "depth_confidence", "l2": 10.905912, "mean_abs": 0.198966, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.118917, "mean_abs": 0.105617, "relative": 0.926926 }, { "name": "video_fisheye_cam1", "l2": 5.413348, "mean_abs": 0.113455, "relative": 0.89451 }, { "name": "video_fisheye_cam2", "l2": 5.475461, "mean_abs": 0.110819, "relative": 0.874076 }, { "name": "video_fisheye_cam3", "l2": 5.500248, "mean_abs": 0.111905, "relative": 0.92095 }, { "name": "video_stereo_left", "l2": 5.148883, "mean_abs": 0.111118, "relative": 0.836704 }, { "name": "video_stereo_right", "l2": 6.355624, "mean_abs": 0.138545, "relative": 0.89601 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.018861, "mean_abs": 0.07433, "relative": 0.72252 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 605, "start_frame": 3025, "end_frame": 3044, "center_frame": 3034, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.784744, "mean_abs": 0.200712, "relative": 0.825385 }, { "name": "hand_right_joints", "l2": 10.955296, "mean_abs": 0.278807, "relative": 0.86357 }, { "name": "body_joints", "l2": 6.197361, "mean_abs": 0.085157, "relative": 0.837073 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017023, "mean_abs": 0.002395, "relative": 0.040171 }, { "name": "camera_rotation_matrix", "l2": 3.000201, "mean_abs": 0.202869, "relative": 0.921313 }, { "name": "imu_accel_gyro", "l2": 17.068922, "mean_abs": 0.973386, "relative": 0.920927 }, { "name": "depth_confidence", "l2": 10.931417, "mean_abs": 0.199898, "relative": 0.000363 }, { "name": "video_fisheye_cam0", "l2": 5.123184, "mean_abs": 0.105944, "relative": 0.927699 }, { "name": "video_fisheye_cam1", "l2": 5.417365, "mean_abs": 0.113933, "relative": 0.895174 }, { "name": "video_fisheye_cam2", "l2": 5.496697, "mean_abs": 0.111454, "relative": 0.877466 }, { "name": "video_fisheye_cam3", "l2": 5.525286, "mean_abs": 0.112871, "relative": 0.925143 }, { "name": "video_stereo_left", "l2": 5.176718, "mean_abs": 0.112475, "relative": 0.841227 }, { "name": "video_stereo_right", "l2": 6.378668, "mean_abs": 0.140019, "relative": 0.899258 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.020332, "mean_abs": 0.075625, "relative": 0.723047 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 606, "start_frame": 3030, "end_frame": 3049, "center_frame": 3039, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.765294, "mean_abs": 0.199777, "relative": 0.823744 }, { "name": "hand_right_joints", "l2": 10.944619, "mean_abs": 0.278486, "relative": 0.862729 }, { "name": "body_joints", "l2": 6.224742, "mean_abs": 0.085471, "relative": 0.840771 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016233, "mean_abs": 0.002359, "relative": 0.038307 }, { "name": "camera_rotation_matrix", "l2": 3.000356, "mean_abs": 0.203579, "relative": 0.921361 }, { "name": "imu_accel_gyro", "l2": 17.06204, "mean_abs": 0.966707, "relative": 0.920556 }, { "name": "depth_confidence", "l2": 10.959421, "mean_abs": 0.200079, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.12094, "mean_abs": 0.106288, "relative": 0.927293 }, { "name": "video_fisheye_cam1", "l2": 5.40942, "mean_abs": 0.114097, "relative": 0.893861 }, { "name": "video_fisheye_cam2", "l2": 5.519981, "mean_abs": 0.111822, "relative": 0.881183 }, { "name": "video_fisheye_cam3", "l2": 5.52326, "mean_abs": 0.113684, "relative": 0.924803 }, { "name": "video_stereo_left", "l2": 5.197419, "mean_abs": 0.113058, "relative": 0.844591 }, { "name": "video_stereo_right", "l2": 6.380352, "mean_abs": 0.140715, "relative": 0.899496 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.087635, "mean_abs": 0.079713, "relative": 0.747133 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 607, "start_frame": 3035, "end_frame": 3054, "center_frame": 3044, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.759134, "mean_abs": 0.199813, "relative": 0.823225 }, { "name": "hand_right_joints", "l2": 10.957458, "mean_abs": 0.278596, "relative": 0.863741 }, { "name": "body_joints", "l2": 6.225652, "mean_abs": 0.084843, "relative": 0.840894 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016863, "mean_abs": 0.002163, "relative": 0.039794 }, { "name": "camera_rotation_matrix", "l2": 3.000519, "mean_abs": 0.203867, "relative": 0.921411 }, { "name": "imu_accel_gyro", "l2": 17.075382, "mean_abs": 0.965835, "relative": 0.921276 }, { "name": "depth_confidence", "l2": 10.970398, "mean_abs": 0.200259, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.120787, "mean_abs": 0.106176, "relative": 0.927265 }, { "name": "video_fisheye_cam1", "l2": 5.399518, "mean_abs": 0.113864, "relative": 0.892225 }, { "name": "video_fisheye_cam2", "l2": 5.535137, "mean_abs": 0.111999, "relative": 0.883602 }, { "name": "video_fisheye_cam3", "l2": 5.521746, "mean_abs": 0.112992, "relative": 0.92455 }, { "name": "video_stereo_left", "l2": 5.21605, "mean_abs": 0.113374, "relative": 0.847618 }, { "name": "video_stereo_right", "l2": 6.385488, "mean_abs": 0.140743, "relative": 0.90022 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.05155, "mean_abs": 0.078074, "relative": 0.734219 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 608, "start_frame": 3040, "end_frame": 3059, "center_frame": 3049, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.762963, "mean_abs": 0.199275, "relative": 0.823548 }, { "name": "hand_right_joints", "l2": 10.935733, "mean_abs": 0.277709, "relative": 0.862028 }, { "name": "body_joints", "l2": 6.197271, "mean_abs": 0.084527, "relative": 0.83706 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017891, "mean_abs": 0.002394, "relative": 0.042219 }, { "name": "camera_rotation_matrix", "l2": 3.000432, "mean_abs": 0.203993, "relative": 0.921384 }, { "name": "imu_accel_gyro", "l2": 17.059248, "mean_abs": 0.964975, "relative": 0.920405 }, { "name": "depth_confidence", "l2": 10.985212, "mean_abs": 0.200076, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.122666, "mean_abs": 0.106503, "relative": 0.927605 }, { "name": "video_fisheye_cam1", "l2": 5.403389, "mean_abs": 0.114213, "relative": 0.892864 }, { "name": "video_fisheye_cam2", "l2": 5.546495, "mean_abs": 0.112231, "relative": 0.885415 }, { "name": "video_fisheye_cam3", "l2": 5.529412, "mean_abs": 0.114015, "relative": 0.925833 }, { "name": "video_stereo_left", "l2": 5.232488, "mean_abs": 0.113669, "relative": 0.85029 }, { "name": "video_stereo_right", "l2": 6.404498, "mean_abs": 0.141402, "relative": 0.9029 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.154031, "mean_abs": 0.084097, "relative": 0.770895 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 609, "start_frame": 3045, "end_frame": 3064, "center_frame": 3054, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.771731, "mean_abs": 0.198703, "relative": 0.824287 }, { "name": "hand_right_joints", "l2": 10.910932, "mean_abs": 0.276121, "relative": 0.860073 }, { "name": "body_joints", "l2": 6.177734, "mean_abs": 0.084367, "relative": 0.834422 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018393, "mean_abs": 0.002476, "relative": 0.043405 }, { "name": "camera_rotation_matrix", "l2": 3.000268, "mean_abs": 0.203607, "relative": 0.921333 }, { "name": "imu_accel_gyro", "l2": 17.067577, "mean_abs": 0.963339, "relative": 0.920855 }, { "name": "depth_confidence", "l2": 11.009089, "mean_abs": 0.199673, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.118747, "mean_abs": 0.106219, "relative": 0.926896 }, { "name": "video_fisheye_cam1", "l2": 5.400204, "mean_abs": 0.113925, "relative": 0.892338 }, { "name": "video_fisheye_cam2", "l2": 5.553526, "mean_abs": 0.112285, "relative": 0.886538 }, { "name": "video_fisheye_cam3", "l2": 5.51316, "mean_abs": 0.112535, "relative": 0.923112 }, { "name": "video_stereo_left", "l2": 5.244098, "mean_abs": 0.113348, "relative": 0.852176 }, { "name": "video_stereo_right", "l2": 6.416276, "mean_abs": 0.141069, "relative": 0.90456 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.111583, "mean_abs": 0.081383, "relative": 0.755704 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 610, "start_frame": 3050, "end_frame": 3069, "center_frame": 3059, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.784895, "mean_abs": 0.198172, "relative": 0.825398 }, { "name": "hand_right_joints", "l2": 10.884422, "mean_abs": 0.274834, "relative": 0.857984 }, { "name": "body_joints", "l2": 6.151702, "mean_abs": 0.083771, "relative": 0.830905 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016782, "mean_abs": 0.002411, "relative": 0.039602 }, { "name": "camera_rotation_matrix", "l2": 2.999971, "mean_abs": 0.202969, "relative": 0.921242 }, { "name": "imu_accel_gyro", "l2": 17.069565, "mean_abs": 0.970962, "relative": 0.920962 }, { "name": "depth_confidence", "l2": 10.972797, "mean_abs": 0.199457, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.116755, "mean_abs": 0.105732, "relative": 0.926535 }, { "name": "video_fisheye_cam1", "l2": 5.394883, "mean_abs": 0.113164, "relative": 0.891459 }, { "name": "video_fisheye_cam2", "l2": 5.557414, "mean_abs": 0.111992, "relative": 0.887158 }, { "name": "video_fisheye_cam3", "l2": 5.499404, "mean_abs": 0.111349, "relative": 0.920809 }, { "name": "video_stereo_left", "l2": 5.25129, "mean_abs": 0.112752, "relative": 0.853345 }, { "name": "video_stereo_right", "l2": 6.421054, "mean_abs": 0.140207, "relative": 0.905234 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.025574, "mean_abs": 0.079254, "relative": 0.724923 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 611, "start_frame": 3055, "end_frame": 3074, "center_frame": 3064, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.783487, "mean_abs": 0.196767, "relative": 0.825279 }, { "name": "hand_right_joints", "l2": 10.875596, "mean_abs": 0.273625, "relative": 0.857288 }, { "name": "body_joints", "l2": 6.118688, "mean_abs": 0.082966, "relative": 0.826446 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012427, "mean_abs": 0.001621, "relative": 0.029327 }, { "name": "camera_rotation_matrix", "l2": 2.99991, "mean_abs": 0.202072, "relative": 0.921224 }, { "name": "imu_accel_gyro", "l2": 17.056957, "mean_abs": 0.967008, "relative": 0.920282 }, { "name": "depth_confidence", "l2": 10.961098, "mean_abs": 0.199203, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.11571, "mean_abs": 0.105452, "relative": 0.926346 }, { "name": "video_fisheye_cam1", "l2": 5.392504, "mean_abs": 0.112617, "relative": 0.891066 }, { "name": "video_fisheye_cam2", "l2": 5.560421, "mean_abs": 0.111696, "relative": 0.887638 }, { "name": "video_fisheye_cam3", "l2": 5.468265, "mean_abs": 0.111198, "relative": 0.915595 }, { "name": "video_stereo_left", "l2": 5.260221, "mean_abs": 0.11215, "relative": 0.854796 }, { "name": "video_stereo_right", "l2": 6.428641, "mean_abs": 0.139483, "relative": 0.906303 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.085486, "mean_abs": 0.083027, "relative": 0.746364 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 612, "start_frame": 3060, "end_frame": 3079, "center_frame": 3069, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.781943, "mean_abs": 0.197679, "relative": 0.825149 }, { "name": "hand_right_joints", "l2": 10.858107, "mean_abs": 0.272981, "relative": 0.855909 }, { "name": "body_joints", "l2": 6.111205, "mean_abs": 0.082731, "relative": 0.825436 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011763, "mean_abs": 0.001681, "relative": 0.027759 }, { "name": "camera_rotation_matrix", "l2": 3.000052, "mean_abs": 0.203036, "relative": 0.921267 }, { "name": "imu_accel_gyro", "l2": 17.052427, "mean_abs": 0.96586, "relative": 0.920037 }, { "name": "depth_confidence", "l2": 10.983378, "mean_abs": 0.199446, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.116573, "mean_abs": 0.1056, "relative": 0.926502 }, { "name": "video_fisheye_cam1", "l2": 5.397694, "mean_abs": 0.113205, "relative": 0.891923 }, { "name": "video_fisheye_cam2", "l2": 5.559024, "mean_abs": 0.11204, "relative": 0.887416 }, { "name": "video_fisheye_cam3", "l2": 5.456451, "mean_abs": 0.110531, "relative": 0.913617 }, { "name": "video_stereo_left", "l2": 5.257394, "mean_abs": 0.112873, "relative": 0.854337 }, { "name": "video_stereo_right", "l2": 6.432116, "mean_abs": 0.140239, "relative": 0.906793 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.050343, "mean_abs": 0.080535, "relative": 0.733787 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 613, "start_frame": 3065, "end_frame": 3084, "center_frame": 3074, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.799525, "mean_abs": 0.198475, "relative": 0.826632 }, { "name": "hand_right_joints", "l2": 10.861176, "mean_abs": 0.271738, "relative": 0.856151 }, { "name": "body_joints", "l2": 6.09831, "mean_abs": 0.083268, "relative": 0.823694 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019431, "mean_abs": 0.002267, "relative": 0.045854 }, { "name": "camera_rotation_matrix", "l2": 3.000539, "mean_abs": 0.204112, "relative": 0.921417 }, { "name": "imu_accel_gyro", "l2": 17.046881, "mean_abs": 0.967157, "relative": 0.919738 }, { "name": "depth_confidence", "l2": 10.971399, "mean_abs": 0.200157, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.108797, "mean_abs": 0.105719, "relative": 0.925094 }, { "name": "video_fisheye_cam1", "l2": 5.396808, "mean_abs": 0.113658, "relative": 0.891777 }, { "name": "video_fisheye_cam2", "l2": 5.549771, "mean_abs": 0.112356, "relative": 0.885938 }, { "name": "video_fisheye_cam3", "l2": 5.418393, "mean_abs": 0.110307, "relative": 0.907245 }, { "name": "video_stereo_left", "l2": 5.247257, "mean_abs": 0.113619, "relative": 0.85269 }, { "name": "video_stereo_right", "l2": 6.423345, "mean_abs": 0.140976, "relative": 0.905557 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.069009, "mean_abs": 0.082501, "relative": 0.740467 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 614, "start_frame": 3070, "end_frame": 3089, "center_frame": 3079, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.8047, "mean_abs": 0.199469, "relative": 0.827068 }, { "name": "hand_right_joints", "l2": 10.876042, "mean_abs": 0.272001, "relative": 0.857323 }, { "name": "body_joints", "l2": 6.09409, "mean_abs": 0.083674, "relative": 0.823124 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030432, "mean_abs": 0.003476, "relative": 0.071815 }, { "name": "camera_rotation_matrix", "l2": 3.001045, "mean_abs": 0.204996, "relative": 0.921572 }, { "name": "imu_accel_gyro", "l2": 17.04587, "mean_abs": 0.971827, "relative": 0.919683 }, { "name": "depth_confidence", "l2": 10.973646, "mean_abs": 0.200489, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.109967, "mean_abs": 0.105988, "relative": 0.925306 }, { "name": "video_fisheye_cam1", "l2": 5.40625, "mean_abs": 0.114157, "relative": 0.893337 }, { "name": "video_fisheye_cam2", "l2": 5.541281, "mean_abs": 0.112489, "relative": 0.884583 }, { "name": "video_fisheye_cam3", "l2": 5.446662, "mean_abs": 0.111176, "relative": 0.911978 }, { "name": "video_stereo_left", "l2": 5.239996, "mean_abs": 0.114402, "relative": 0.85151 }, { "name": "video_stereo_right", "l2": 6.420115, "mean_abs": 0.141666, "relative": 0.905102 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.047253, "mean_abs": 0.081183, "relative": 0.732681 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 615, "start_frame": 3075, "end_frame": 3094, "center_frame": 3084, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.811644, "mean_abs": 0.199661, "relative": 0.827654 }, { "name": "hand_right_joints", "l2": 10.901308, "mean_abs": 0.273654, "relative": 0.859315 }, { "name": "body_joints", "l2": 6.098789, "mean_abs": 0.084158, "relative": 0.823758 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.033751, "mean_abs": 0.004199, "relative": 0.079648 }, { "name": "camera_rotation_matrix", "l2": 3.001137, "mean_abs": 0.205069, "relative": 0.921601 }, { "name": "imu_accel_gyro", "l2": 17.064655, "mean_abs": 0.975059, "relative": 0.920697 }, { "name": "depth_confidence", "l2": 10.983347, "mean_abs": 0.200235, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.111519, "mean_abs": 0.105967, "relative": 0.925587 }, { "name": "video_fisheye_cam1", "l2": 5.417799, "mean_abs": 0.114399, "relative": 0.895246 }, { "name": "video_fisheye_cam2", "l2": 5.532553, "mean_abs": 0.112319, "relative": 0.88319 }, { "name": "video_fisheye_cam3", "l2": 5.454831, "mean_abs": 0.111709, "relative": 0.913346 }, { "name": "video_stereo_left", "l2": 5.227855, "mean_abs": 0.114393, "relative": 0.849537 }, { "name": "video_stereo_right", "l2": 6.418552, "mean_abs": 0.14182, "relative": 0.904881 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.060794, "mean_abs": 0.079468, "relative": 0.737527 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 616, "start_frame": 3080, "end_frame": 3099, "center_frame": 3089, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.827292, "mean_abs": 0.198836, "relative": 0.828974 }, { "name": "hand_right_joints", "l2": 10.92287, "mean_abs": 0.273914, "relative": 0.861014 }, { "name": "body_joints", "l2": 6.103982, "mean_abs": 0.08417, "relative": 0.82446 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.031938, "mean_abs": 0.004227, "relative": 0.075369 }, { "name": "camera_rotation_matrix", "l2": 3.000764, "mean_abs": 0.204166, "relative": 0.921486 }, { "name": "imu_accel_gyro", "l2": 17.080111, "mean_abs": 0.971375, "relative": 0.921531 }, { "name": "depth_confidence", "l2": 10.989521, "mean_abs": 0.20069, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.109376, "mean_abs": 0.105773, "relative": 0.925199 }, { "name": "video_fisheye_cam1", "l2": 5.419255, "mean_abs": 0.114234, "relative": 0.895486 }, { "name": "video_fisheye_cam2", "l2": 5.51901, "mean_abs": 0.11187, "relative": 0.881028 }, { "name": "video_fisheye_cam3", "l2": 5.450427, "mean_abs": 0.11159, "relative": 0.912608 }, { "name": "video_stereo_left", "l2": 5.204776, "mean_abs": 0.113383, "relative": 0.845786 }, { "name": "video_stereo_right", "l2": 6.398847, "mean_abs": 0.141004, "relative": 0.902103 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.027387, "mean_abs": 0.078983, "relative": 0.725571 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 617, "start_frame": 3085, "end_frame": 3104, "center_frame": 3094, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.831397, "mean_abs": 0.199622, "relative": 0.82932 }, { "name": "hand_right_joints", "l2": 10.943671, "mean_abs": 0.276114, "relative": 0.862654 }, { "name": "body_joints", "l2": 6.118621, "mean_abs": 0.084141, "relative": 0.826437 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.023307, "mean_abs": 0.003471, "relative": 0.055002 }, { "name": "camera_rotation_matrix", "l2": 3.000609, "mean_abs": 0.202481, "relative": 0.921438 }, { "name": "imu_accel_gyro", "l2": 17.065538, "mean_abs": 0.979433, "relative": 0.920745 }, { "name": "depth_confidence", "l2": 10.997591, "mean_abs": 0.200137, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.112029, "mean_abs": 0.105505, "relative": 0.925679 }, { "name": "video_fisheye_cam1", "l2": 5.428749, "mean_abs": 0.113809, "relative": 0.897055 }, { "name": "video_fisheye_cam2", "l2": 5.511084, "mean_abs": 0.111262, "relative": 0.879763 }, { "name": "video_fisheye_cam3", "l2": 5.464562, "mean_abs": 0.111161, "relative": 0.914975 }, { "name": "video_stereo_left", "l2": 5.191373, "mean_abs": 0.112251, "relative": 0.843608 }, { "name": "video_stereo_right", "l2": 6.388058, "mean_abs": 0.139766, "relative": 0.900582 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.041899, "mean_abs": 0.080092, "relative": 0.730765 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 618, "start_frame": 3090, "end_frame": 3109, "center_frame": 3099, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.841976, "mean_abs": 0.2003, "relative": 0.830213 }, { "name": "hand_right_joints", "l2": 10.978034, "mean_abs": 0.278032, "relative": 0.865363 }, { "name": "body_joints", "l2": 6.152429, "mean_abs": 0.084282, "relative": 0.831004 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011964, "mean_abs": 0.001861, "relative": 0.028233 }, { "name": "camera_rotation_matrix", "l2": 2.999567, "mean_abs": 0.201945, "relative": 0.921118 }, { "name": "imu_accel_gyro", "l2": 17.063908, "mean_abs": 0.971266, "relative": 0.920657 }, { "name": "depth_confidence", "l2": 10.999884, "mean_abs": 0.200077, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.110215, "mean_abs": 0.105139, "relative": 0.925351 }, { "name": "video_fisheye_cam1", "l2": 5.434772, "mean_abs": 0.113423, "relative": 0.89805 }, { "name": "video_fisheye_cam2", "l2": 5.508758, "mean_abs": 0.110971, "relative": 0.879391 }, { "name": "video_fisheye_cam3", "l2": 5.458979, "mean_abs": 0.110238, "relative": 0.91404 }, { "name": "video_stereo_left", "l2": 5.182969, "mean_abs": 0.111334, "relative": 0.842243 }, { "name": "video_stereo_right", "l2": 6.383089, "mean_abs": 0.13848, "relative": 0.899882 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.995077, "mean_abs": 0.076494, "relative": 0.714008 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 619, "start_frame": 3095, "end_frame": 3114, "center_frame": 3104, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.847318, "mean_abs": 0.20034, "relative": 0.830663 }, { "name": "hand_right_joints", "l2": 11.005317, "mean_abs": 0.279912, "relative": 0.867513 }, { "name": "body_joints", "l2": 6.18342, "mean_abs": 0.084577, "relative": 0.835189 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010222, "mean_abs": 0.001393, "relative": 0.024123 }, { "name": "camera_rotation_matrix", "l2": 3.000391, "mean_abs": 0.202718, "relative": 0.921371 }, { "name": "imu_accel_gyro", "l2": 17.06344, "mean_abs": 0.96129, "relative": 0.920631 }, { "name": "depth_confidence", "l2": 11.055391, "mean_abs": 0.200719, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.113705, "mean_abs": 0.105492, "relative": 0.925983 }, { "name": "video_fisheye_cam1", "l2": 5.44178, "mean_abs": 0.113968, "relative": 0.899208 }, { "name": "video_fisheye_cam2", "l2": 5.515886, "mean_abs": 0.111423, "relative": 0.880529 }, { "name": "video_fisheye_cam3", "l2": 5.477936, "mean_abs": 0.111146, "relative": 0.917214 }, { "name": "video_stereo_left", "l2": 5.188096, "mean_abs": 0.112051, "relative": 0.843076 }, { "name": "video_stereo_right", "l2": 6.397253, "mean_abs": 0.139455, "relative": 0.901878 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.068643, "mean_abs": 0.083607, "relative": 0.740336 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 620, "start_frame": 3100, "end_frame": 3119, "center_frame": 3109, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.853373, "mean_abs": 0.200557, "relative": 0.831174 }, { "name": "hand_right_joints", "l2": 11.014186, "mean_abs": 0.280225, "relative": 0.868213 }, { "name": "body_joints", "l2": 6.208992, "mean_abs": 0.085188, "relative": 0.838643 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019184, "mean_abs": 0.002576, "relative": 0.045271 }, { "name": "camera_rotation_matrix", "l2": 3.000506, "mean_abs": 0.203738, "relative": 0.921407 }, { "name": "imu_accel_gyro", "l2": 17.083994, "mean_abs": 0.962442, "relative": 0.92174 }, { "name": "depth_confidence", "l2": 11.059134, "mean_abs": 0.201056, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.116592, "mean_abs": 0.105845, "relative": 0.926505 }, { "name": "video_fisheye_cam1", "l2": 5.440937, "mean_abs": 0.114369, "relative": 0.899069 }, { "name": "video_fisheye_cam2", "l2": 5.521566, "mean_abs": 0.111773, "relative": 0.881436 }, { "name": "video_fisheye_cam3", "l2": 5.490218, "mean_abs": 0.111778, "relative": 0.919271 }, { "name": "video_stereo_left", "l2": 5.196803, "mean_abs": 0.11281, "relative": 0.844491 }, { "name": "video_stereo_right", "l2": 6.40964, "mean_abs": 0.140574, "relative": 0.903625 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.11341, "mean_abs": 0.085419, "relative": 0.756358 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 621, "start_frame": 3105, "end_frame": 3124, "center_frame": 3114, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.865049, "mean_abs": 0.200814, "relative": 0.832159 }, { "name": "hand_right_joints", "l2": 11.028566, "mean_abs": 0.280761, "relative": 0.869346 }, { "name": "body_joints", "l2": 6.228521, "mean_abs": 0.085146, "relative": 0.841281 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.022968, "mean_abs": 0.003192, "relative": 0.054202 }, { "name": "camera_rotation_matrix", "l2": 3.000534, "mean_abs": 0.204293, "relative": 0.921415 }, { "name": "imu_accel_gyro", "l2": 17.091759, "mean_abs": 0.964566, "relative": 0.922159 }, { "name": "depth_confidence", "l2": 11.070308, "mean_abs": 0.200879, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.118376, "mean_abs": 0.105931, "relative": 0.926829 }, { "name": "video_fisheye_cam1", "l2": 5.443024, "mean_abs": 0.114464, "relative": 0.899414 }, { "name": "video_fisheye_cam2", "l2": 5.530608, "mean_abs": 0.111979, "relative": 0.882879 }, { "name": "video_fisheye_cam3", "l2": 5.503525, "mean_abs": 0.112025, "relative": 0.921499 }, { "name": "video_stereo_left", "l2": 5.20969, "mean_abs": 0.113221, "relative": 0.846585 }, { "name": "video_stereo_right", "l2": 6.432078, "mean_abs": 0.141092, "relative": 0.906788 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.122666, "mean_abs": 0.084456, "relative": 0.75967 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 622, "start_frame": 3110, "end_frame": 3129, "center_frame": 3119, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.864899, "mean_abs": 0.200174, "relative": 0.832146 }, { "name": "hand_right_joints", "l2": 11.017323, "mean_abs": 0.280908, "relative": 0.86846 }, { "name": "body_joints", "l2": 6.24781, "mean_abs": 0.085332, "relative": 0.843887 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019952, "mean_abs": 0.002733, "relative": 0.047084 }, { "name": "camera_rotation_matrix", "l2": 3.000453, "mean_abs": 0.204364, "relative": 0.92139 }, { "name": "imu_accel_gyro", "l2": 17.048151, "mean_abs": 0.961979, "relative": 0.919806 }, { "name": "depth_confidence", "l2": 11.069493, "mean_abs": 0.201121, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.118544, "mean_abs": 0.106145, "relative": 0.926859 }, { "name": "video_fisheye_cam1", "l2": 5.442299, "mean_abs": 0.114477, "relative": 0.899294 }, { "name": "video_fisheye_cam2", "l2": 5.540179, "mean_abs": 0.112117, "relative": 0.884407 }, { "name": "video_fisheye_cam3", "l2": 5.508944, "mean_abs": 0.112521, "relative": 0.922406 }, { "name": "video_stereo_left", "l2": 5.227247, "mean_abs": 0.11346, "relative": 0.849438 }, { "name": "video_stereo_right", "l2": 6.456735, "mean_abs": 0.141807, "relative": 0.910264 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.181662, "mean_abs": 0.084635, "relative": 0.780784 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 623, "start_frame": 3115, "end_frame": 3134, "center_frame": 3124, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.872115, "mean_abs": 0.199678, "relative": 0.832755 }, { "name": "hand_right_joints", "l2": 11.01384, "mean_abs": 0.281476, "relative": 0.868185 }, { "name": "body_joints", "l2": 6.25051, "mean_abs": 0.08503, "relative": 0.844251 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018424, "mean_abs": 0.002554, "relative": 0.043479 }, { "name": "camera_rotation_matrix", "l2": 3.00049, "mean_abs": 0.204749, "relative": 0.921402 }, { "name": "imu_accel_gyro", "l2": 17.054634, "mean_abs": 0.961354, "relative": 0.920156 }, { "name": "depth_confidence", "l2": 11.05235, "mean_abs": 0.200259, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.12004, "mean_abs": 0.106088, "relative": 0.92713 }, { "name": "video_fisheye_cam1", "l2": 5.442501, "mean_abs": 0.114455, "relative": 0.899327 }, { "name": "video_fisheye_cam2", "l2": 5.549956, "mean_abs": 0.112354, "relative": 0.885968 }, { "name": "video_fisheye_cam3", "l2": 5.520738, "mean_abs": 0.112247, "relative": 0.924381 }, { "name": "video_stereo_left", "l2": 5.244956, "mean_abs": 0.114017, "relative": 0.852316 }, { "name": "video_stereo_right", "l2": 6.481631, "mean_abs": 0.142491, "relative": 0.913774 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.218929, "mean_abs": 0.08539, "relative": 0.794122 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 624, "start_frame": 3120, "end_frame": 3139, "center_frame": 3129, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.872519, "mean_abs": 0.198654, "relative": 0.832789 }, { "name": "hand_right_joints", "l2": 11.004384, "mean_abs": 0.28187, "relative": 0.86744 }, { "name": "body_joints", "l2": 6.247711, "mean_abs": 0.084582, "relative": 0.843873 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020175, "mean_abs": 0.002971, "relative": 0.047609 }, { "name": "camera_rotation_matrix", "l2": 3.000446, "mean_abs": 0.204676, "relative": 0.921388 }, { "name": "imu_accel_gyro", "l2": 17.061956, "mean_abs": 0.962691, "relative": 0.920551 }, { "name": "depth_confidence", "l2": 11.071919, "mean_abs": 0.200231, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.120892, "mean_abs": 0.106094, "relative": 0.927284 }, { "name": "video_fisheye_cam1", "l2": 5.441973, "mean_abs": 0.114465, "relative": 0.89924 }, { "name": "video_fisheye_cam2", "l2": 5.554204, "mean_abs": 0.112394, "relative": 0.886646 }, { "name": "video_fisheye_cam3", "l2": 5.529454, "mean_abs": 0.112236, "relative": 0.92584 }, { "name": "video_stereo_left", "l2": 5.258558, "mean_abs": 0.114021, "relative": 0.854526 }, { "name": "video_stereo_right", "l2": 6.503557, "mean_abs": 0.142811, "relative": 0.916865 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.231395, "mean_abs": 0.086967, "relative": 0.798583 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 625, "start_frame": 3125, "end_frame": 3144, "center_frame": 3134, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.882539, "mean_abs": 0.1988, "relative": 0.833634 }, { "name": "hand_right_joints", "l2": 10.984913, "mean_abs": 0.281147, "relative": 0.865905 }, { "name": "body_joints", "l2": 6.24091, "mean_abs": 0.084246, "relative": 0.842955 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.02209, "mean_abs": 0.003083, "relative": 0.05213 }, { "name": "camera_rotation_matrix", "l2": 3.000252, "mean_abs": 0.204269, "relative": 0.921329 }, { "name": "imu_accel_gyro", "l2": 17.067972, "mean_abs": 0.969259, "relative": 0.920876 }, { "name": "depth_confidence", "l2": 11.060225, "mean_abs": 0.199237, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.123535, "mean_abs": 0.105959, "relative": 0.927763 }, { "name": "video_fisheye_cam1", "l2": 5.448083, "mean_abs": 0.114362, "relative": 0.90025 }, { "name": "video_fisheye_cam2", "l2": 5.561053, "mean_abs": 0.112426, "relative": 0.887739 }, { "name": "video_fisheye_cam3", "l2": 5.544358, "mean_abs": 0.112214, "relative": 0.928336 }, { "name": "video_stereo_left", "l2": 5.273198, "mean_abs": 0.113773, "relative": 0.856905 }, { "name": "video_stereo_right", "l2": 6.529938, "mean_abs": 0.142826, "relative": 0.920584 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.130975, "mean_abs": 0.084334, "relative": 0.762644 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 626, "start_frame": 3130, "end_frame": 3149, "center_frame": 3139, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.878271, "mean_abs": 0.199415, "relative": 0.833274 }, { "name": "hand_right_joints", "l2": 10.964658, "mean_abs": 0.281185, "relative": 0.864308 }, { "name": "body_joints", "l2": 6.223868, "mean_abs": 0.084018, "relative": 0.840653 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.022046, "mean_abs": 0.002866, "relative": 0.052026 }, { "name": "camera_rotation_matrix", "l2": 3.000224, "mean_abs": 0.203786, "relative": 0.92132 }, { "name": "imu_accel_gyro", "l2": 17.065985, "mean_abs": 0.971841, "relative": 0.920769 }, { "name": "depth_confidence", "l2": 11.056756, "mean_abs": 0.198509, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.128629, "mean_abs": 0.105871, "relative": 0.928685 }, { "name": "video_fisheye_cam1", "l2": 5.460463, "mean_abs": 0.114126, "relative": 0.902295 }, { "name": "video_fisheye_cam2", "l2": 5.567309, "mean_abs": 0.112286, "relative": 0.888738 }, { "name": "video_fisheye_cam3", "l2": 5.571281, "mean_abs": 0.112435, "relative": 0.932844 }, { "name": "video_stereo_left", "l2": 5.289605, "mean_abs": 0.113512, "relative": 0.859571 }, { "name": "video_stereo_right", "l2": 6.557578, "mean_abs": 0.142506, "relative": 0.924481 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.090042, "mean_abs": 0.080101, "relative": 0.747995 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 627, "start_frame": 3135, "end_frame": 3154, "center_frame": 3144, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.860723, "mean_abs": 0.198932, "relative": 0.831794 }, { "name": "hand_right_joints", "l2": 10.951238, "mean_abs": 0.279665, "relative": 0.86325 }, { "name": "body_joints", "l2": 6.221446, "mean_abs": 0.08372, "relative": 0.840326 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015615, "mean_abs": 0.001859, "relative": 0.036849 }, { "name": "camera_rotation_matrix", "l2": 3.00017, "mean_abs": 0.20396, "relative": 0.921304 }, { "name": "imu_accel_gyro", "l2": 17.08131, "mean_abs": 0.972189, "relative": 0.921596 }, { "name": "depth_confidence", "l2": 11.069202, "mean_abs": 0.1988, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.130855, "mean_abs": 0.10587, "relative": 0.929088 }, { "name": "video_fisheye_cam1", "l2": 5.468345, "mean_abs": 0.114223, "relative": 0.903598 }, { "name": "video_fisheye_cam2", "l2": 5.569607, "mean_abs": 0.112205, "relative": 0.889105 }, { "name": "video_fisheye_cam3", "l2": 5.572292, "mean_abs": 0.112692, "relative": 0.933013 }, { "name": "video_stereo_left", "l2": 5.295462, "mean_abs": 0.11362, "relative": 0.860523 }, { "name": "video_stereo_right", "l2": 6.570079, "mean_abs": 0.142883, "relative": 0.926243 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.093995, "mean_abs": 0.081637, "relative": 0.749409 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 628, "start_frame": 3140, "end_frame": 3159, "center_frame": 3149, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.85978, "mean_abs": 0.199086, "relative": 0.831715 }, { "name": "hand_right_joints", "l2": 10.934983, "mean_abs": 0.278835, "relative": 0.861969 }, { "name": "body_joints", "l2": 6.227649, "mean_abs": 0.08411, "relative": 0.841163 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011535, "mean_abs": 0.001529, "relative": 0.027221 }, { "name": "camera_rotation_matrix", "l2": 3.000191, "mean_abs": 0.204621, "relative": 0.92131 }, { "name": "imu_accel_gyro", "l2": 17.080147, "mean_abs": 0.97407, "relative": 0.921533 }, { "name": "depth_confidence", "l2": 11.005553, "mean_abs": 0.198696, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.132091, "mean_abs": 0.10625, "relative": 0.929312 }, { "name": "video_fisheye_cam1", "l2": 5.477753, "mean_abs": 0.114927, "relative": 0.905152 }, { "name": "video_fisheye_cam2", "l2": 5.566771, "mean_abs": 0.112437, "relative": 0.888652 }, { "name": "video_fisheye_cam3", "l2": 5.574733, "mean_abs": 0.11349, "relative": 0.933422 }, { "name": "video_stereo_left", "l2": 5.294957, "mean_abs": 0.114139, "relative": 0.860441 }, { "name": "video_stereo_right", "l2": 6.579493, "mean_abs": 0.143739, "relative": 0.92757 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.057047, "mean_abs": 0.078827, "relative": 0.736186 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 629, "start_frame": 3145, "end_frame": 3164, "center_frame": 3154, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.858573, "mean_abs": 0.19973, "relative": 0.831613 }, { "name": "hand_right_joints", "l2": 10.885848, "mean_abs": 0.277251, "relative": 0.858096 }, { "name": "body_joints", "l2": 6.228639, "mean_abs": 0.084554, "relative": 0.841297 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013159, "mean_abs": 0.001754, "relative": 0.031053 }, { "name": "camera_rotation_matrix", "l2": 3.000261, "mean_abs": 0.204812, "relative": 0.921332 }, { "name": "imu_accel_gyro", "l2": 17.068977, "mean_abs": 0.970962, "relative": 0.92093 }, { "name": "depth_confidence", "l2": 11.024928, "mean_abs": 0.199362, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.136685, "mean_abs": 0.106115, "relative": 0.930144 }, { "name": "video_fisheye_cam1", "l2": 5.495374, "mean_abs": 0.115053, "relative": 0.908064 }, { "name": "video_fisheye_cam2", "l2": 5.565452, "mean_abs": 0.112507, "relative": 0.888442 }, { "name": "video_fisheye_cam3", "l2": 5.585233, "mean_abs": 0.112989, "relative": 0.93518 }, { "name": "video_stereo_left", "l2": 5.29719, "mean_abs": 0.114316, "relative": 0.860804 }, { "name": "video_stereo_right", "l2": 6.598298, "mean_abs": 0.143825, "relative": 0.930222 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.038424, "mean_abs": 0.080523, "relative": 0.729521 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 630, "start_frame": 3150, "end_frame": 3169, "center_frame": 3159, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.859209, "mean_abs": 0.200068, "relative": 0.831666 }, { "name": "hand_right_joints", "l2": 10.896318, "mean_abs": 0.278897, "relative": 0.858921 }, { "name": "body_joints", "l2": 6.2455, "mean_abs": 0.084951, "relative": 0.843575 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015446, "mean_abs": 0.001693, "relative": 0.036451 }, { "name": "camera_rotation_matrix", "l2": 3.000324, "mean_abs": 0.204785, "relative": 0.921351 }, { "name": "imu_accel_gyro", "l2": 17.060452, "mean_abs": 0.974391, "relative": 0.92047 }, { "name": "depth_confidence", "l2": 10.996735, "mean_abs": 0.199467, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.141179, "mean_abs": 0.106279, "relative": 0.930958 }, { "name": "video_fisheye_cam1", "l2": 5.503475, "mean_abs": 0.115232, "relative": 0.909403 }, { "name": "video_fisheye_cam2", "l2": 5.561895, "mean_abs": 0.112615, "relative": 0.887874 }, { "name": "video_fisheye_cam3", "l2": 5.583761, "mean_abs": 0.113002, "relative": 0.934933 }, { "name": "video_stereo_left", "l2": 5.289349, "mean_abs": 0.114427, "relative": 0.85953 }, { "name": "video_stereo_right", "l2": 6.591643, "mean_abs": 0.143685, "relative": 0.929283 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.019367, "mean_abs": 0.079773, "relative": 0.722701 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 631, "start_frame": 3155, "end_frame": 3174, "center_frame": 3164, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.86347, "mean_abs": 0.200461, "relative": 0.832026 }, { "name": "hand_right_joints", "l2": 10.884115, "mean_abs": 0.27881, "relative": 0.857959 }, { "name": "body_joints", "l2": 6.264054, "mean_abs": 0.08526, "relative": 0.846081 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01991, "mean_abs": 0.002322, "relative": 0.046985 }, { "name": "camera_rotation_matrix", "l2": 3.000451, "mean_abs": 0.204827, "relative": 0.92139 }, { "name": "imu_accel_gyro", "l2": 17.085211, "mean_abs": 0.9747, "relative": 0.921806 }, { "name": "depth_confidence", "l2": 10.981943, "mean_abs": 0.200055, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.146427, "mean_abs": 0.106471, "relative": 0.931908 }, { "name": "video_fisheye_cam1", "l2": 5.509326, "mean_abs": 0.115488, "relative": 0.91037 }, { "name": "video_fisheye_cam2", "l2": 5.55584, "mean_abs": 0.11257, "relative": 0.886907 }, { "name": "video_fisheye_cam3", "l2": 5.586315, "mean_abs": 0.113148, "relative": 0.935361 }, { "name": "video_stereo_left", "l2": 5.279081, "mean_abs": 0.114679, "relative": 0.857861 }, { "name": "video_stereo_right", "l2": 6.582924, "mean_abs": 0.143715, "relative": 0.928054 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.993636, "mean_abs": 0.078368, "relative": 0.713492 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 632, "start_frame": 3160, "end_frame": 3179, "center_frame": 3169, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.84805, "mean_abs": 0.200421, "relative": 0.830725 }, { "name": "hand_right_joints", "l2": 10.89997, "mean_abs": 0.279814, "relative": 0.859209 }, { "name": "body_joints", "l2": 6.285058, "mean_abs": 0.085421, "relative": 0.848918 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019018, "mean_abs": 0.002447, "relative": 0.044879 }, { "name": "camera_rotation_matrix", "l2": 3.000462, "mean_abs": 0.204502, "relative": 0.921393 }, { "name": "imu_accel_gyro", "l2": 17.087616, "mean_abs": 0.972561, "relative": 0.921936 }, { "name": "depth_confidence", "l2": 11.002136, "mean_abs": 0.200872, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.153262, "mean_abs": 0.106537, "relative": 0.933146 }, { "name": "video_fisheye_cam1", "l2": 5.522616, "mean_abs": 0.115709, "relative": 0.912566 }, { "name": "video_fisheye_cam2", "l2": 5.5521, "mean_abs": 0.112438, "relative": 0.88631 }, { "name": "video_fisheye_cam3", "l2": 5.600323, "mean_abs": 0.113398, "relative": 0.937707 }, { "name": "video_stereo_left", "l2": 5.273111, "mean_abs": 0.11449, "relative": 0.856891 }, { "name": "video_stereo_right", "l2": 6.586626, "mean_abs": 0.143591, "relative": 0.928576 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.030395, "mean_abs": 0.076364, "relative": 0.726648 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 633, "start_frame": 3165, "end_frame": 3184, "center_frame": 3174, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.848776, "mean_abs": 0.200378, "relative": 0.830786 }, { "name": "hand_right_joints", "l2": 10.925346, "mean_abs": 0.281081, "relative": 0.86121 }, { "name": "body_joints", "l2": 6.307703, "mean_abs": 0.085602, "relative": 0.851976 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01428, "mean_abs": 0.00197, "relative": 0.033699 }, { "name": "camera_rotation_matrix", "l2": 3.000322, "mean_abs": 0.203506, "relative": 0.92135 }, { "name": "imu_accel_gyro", "l2": 17.065964, "mean_abs": 0.973304, "relative": 0.920767 }, { "name": "depth_confidence", "l2": 11.025356, "mean_abs": 0.201046, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.157938, "mean_abs": 0.106419, "relative": 0.933992 }, { "name": "video_fisheye_cam1", "l2": 5.528951, "mean_abs": 0.115481, "relative": 0.913612 }, { "name": "video_fisheye_cam2", "l2": 5.551638, "mean_abs": 0.112212, "relative": 0.886236 }, { "name": "video_fisheye_cam3", "l2": 5.605755, "mean_abs": 0.113242, "relative": 0.938616 }, { "name": "video_stereo_left", "l2": 5.264242, "mean_abs": 0.113733, "relative": 0.85545 }, { "name": "video_stereo_right", "l2": 6.584141, "mean_abs": 0.143077, "relative": 0.928226 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.037096, "mean_abs": 0.078043, "relative": 0.729046 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 634, "start_frame": 3170, "end_frame": 3189, "center_frame": 3179, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.84181, "mean_abs": 0.200303, "relative": 0.830199 }, { "name": "hand_right_joints", "l2": 10.935905, "mean_abs": 0.281656, "relative": 0.862042 }, { "name": "body_joints", "l2": 6.330682, "mean_abs": 0.085789, "relative": 0.85508 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007353, "mean_abs": 0.001031, "relative": 0.017352 }, { "name": "camera_rotation_matrix", "l2": 2.999656, "mean_abs": 0.202742, "relative": 0.921146 }, { "name": "imu_accel_gyro", "l2": 17.079617, "mean_abs": 0.977417, "relative": 0.921504 }, { "name": "depth_confidence", "l2": 11.092041, "mean_abs": 0.200491, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.164055, "mean_abs": 0.106218, "relative": 0.9351 }, { "name": "video_fisheye_cam1", "l2": 5.537928, "mean_abs": 0.11535, "relative": 0.915096 }, { "name": "video_fisheye_cam2", "l2": 5.555782, "mean_abs": 0.112079, "relative": 0.886898 }, { "name": "video_fisheye_cam3", "l2": 5.611341, "mean_abs": 0.113019, "relative": 0.939551 }, { "name": "video_stereo_left", "l2": 5.261295, "mean_abs": 0.113032, "relative": 0.854971 }, { "name": "video_stereo_right", "l2": 6.588267, "mean_abs": 0.142643, "relative": 0.928807 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.083308, "mean_abs": 0.08198, "relative": 0.745585 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 635, "start_frame": 3175, "end_frame": 3194, "center_frame": 3184, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.841842, "mean_abs": 0.200114, "relative": 0.830201 }, { "name": "hand_right_joints", "l2": 10.949683, "mean_abs": 0.282826, "relative": 0.863128 }, { "name": "body_joints", "l2": 6.356424, "mean_abs": 0.086401, "relative": 0.858557 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00725, "mean_abs": 0.001126, "relative": 0.017109 }, { "name": "camera_rotation_matrix", "l2": 3.000235, "mean_abs": 0.204186, "relative": 0.921323 }, { "name": "imu_accel_gyro", "l2": 17.077723, "mean_abs": 0.971495, "relative": 0.921402 }, { "name": "depth_confidence", "l2": 11.115539, "mean_abs": 0.201388, "relative": 0.000369 }, { "name": "video_fisheye_cam0", "l2": 5.168004, "mean_abs": 0.10658, "relative": 0.935815 }, { "name": "video_fisheye_cam1", "l2": 5.546599, "mean_abs": 0.115854, "relative": 0.916529 }, { "name": "video_fisheye_cam2", "l2": 5.563885, "mean_abs": 0.112579, "relative": 0.888191 }, { "name": "video_fisheye_cam3", "l2": 5.614989, "mean_abs": 0.113205, "relative": 0.940162 }, { "name": "video_stereo_left", "l2": 5.269623, "mean_abs": 0.113984, "relative": 0.856324 }, { "name": "video_stereo_right", "l2": 6.605911, "mean_abs": 0.14362, "relative": 0.931295 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.101868, "mean_abs": 0.083592, "relative": 0.752227 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 636, "start_frame": 3180, "end_frame": 3199, "center_frame": 3189, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.843791, "mean_abs": 0.200453, "relative": 0.830366 }, { "name": "hand_right_joints", "l2": 10.965607, "mean_abs": 0.284745, "relative": 0.864383 }, { "name": "body_joints", "l2": 6.39677, "mean_abs": 0.087533, "relative": 0.864006 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013759, "mean_abs": 0.001732, "relative": 0.032469 }, { "name": "camera_rotation_matrix", "l2": 3.000624, "mean_abs": 0.205209, "relative": 0.921443 }, { "name": "imu_accel_gyro", "l2": 17.063717, "mean_abs": 0.972833, "relative": 0.920646 }, { "name": "depth_confidence", "l2": 11.108537, "mean_abs": 0.20112, "relative": 0.000369 }, { "name": "video_fisheye_cam0", "l2": 5.173516, "mean_abs": 0.106955, "relative": 0.936813 }, { "name": "video_fisheye_cam1", "l2": 5.556668, "mean_abs": 0.116385, "relative": 0.918193 }, { "name": "video_fisheye_cam2", "l2": 5.576371, "mean_abs": 0.113229, "relative": 0.890185 }, { "name": "video_fisheye_cam3", "l2": 5.62745, "mean_abs": 0.113682, "relative": 0.942249 }, { "name": "video_stereo_left", "l2": 5.282484, "mean_abs": 0.114912, "relative": 0.858414 }, { "name": "video_stereo_right", "l2": 6.64344, "mean_abs": 0.145181, "relative": 0.936586 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.067186, "mean_abs": 0.08168, "relative": 0.739815 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 637, "start_frame": 3185, "end_frame": 3204, "center_frame": 3194, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.842139, "mean_abs": 0.200308, "relative": 0.830226 }, { "name": "hand_right_joints", "l2": 10.983222, "mean_abs": 0.285889, "relative": 0.865772 }, { "name": "body_joints", "l2": 6.417771, "mean_abs": 0.088048, "relative": 0.866843 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019137, "mean_abs": 0.00236, "relative": 0.04516 }, { "name": "camera_rotation_matrix", "l2": 3.000821, "mean_abs": 0.206019, "relative": 0.921503 }, { "name": "imu_accel_gyro", "l2": 17.067684, "mean_abs": 0.967334, "relative": 0.92086 }, { "name": "depth_confidence", "l2": 11.089769, "mean_abs": 0.200552, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.178613, "mean_abs": 0.107211, "relative": 0.937736 }, { "name": "video_fisheye_cam1", "l2": 5.560098, "mean_abs": 0.116585, "relative": 0.918759 }, { "name": "video_fisheye_cam2", "l2": 5.592035, "mean_abs": 0.113658, "relative": 0.892685 }, { "name": "video_fisheye_cam3", "l2": 5.632394, "mean_abs": 0.113779, "relative": 0.943076 }, { "name": "video_stereo_left", "l2": 5.299975, "mean_abs": 0.115677, "relative": 0.861256 }, { "name": "video_stereo_right", "l2": 6.672779, "mean_abs": 0.146127, "relative": 0.940722 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.150826, "mean_abs": 0.086994, "relative": 0.769748 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 638, "start_frame": 3190, "end_frame": 3209, "center_frame": 3199, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.851817, "mean_abs": 0.200034, "relative": 0.831043 }, { "name": "hand_right_joints", "l2": 10.978973, "mean_abs": 0.286662, "relative": 0.865437 }, { "name": "body_joints", "l2": 6.43737, "mean_abs": 0.088334, "relative": 0.86949 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.02254, "mean_abs": 0.002837, "relative": 0.05319 }, { "name": "camera_rotation_matrix", "l2": 3.000598, "mean_abs": 0.206254, "relative": 0.921435 }, { "name": "imu_accel_gyro", "l2": 17.0669, "mean_abs": 0.965614, "relative": 0.920818 }, { "name": "depth_confidence", "l2": 11.09335, "mean_abs": 0.200757, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.183354, "mean_abs": 0.107198, "relative": 0.938595 }, { "name": "video_fisheye_cam1", "l2": 5.564646, "mean_abs": 0.116839, "relative": 0.919511 }, { "name": "video_fisheye_cam2", "l2": 5.608324, "mean_abs": 0.113921, "relative": 0.895285 }, { "name": "video_fisheye_cam3", "l2": 5.639246, "mean_abs": 0.113782, "relative": 0.944224 }, { "name": "video_stereo_left", "l2": 5.321434, "mean_abs": 0.115792, "relative": 0.864744 }, { "name": "video_stereo_right", "l2": 6.699458, "mean_abs": 0.146773, "relative": 0.944483 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.138065, "mean_abs": 0.085689, "relative": 0.765181 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 639, "start_frame": 3195, "end_frame": 3214, "center_frame": 3204, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.848571, "mean_abs": 0.200576, "relative": 0.830769 }, { "name": "hand_right_joints", "l2": 10.95996, "mean_abs": 0.286856, "relative": 0.863938 }, { "name": "body_joints", "l2": 6.449122, "mean_abs": 0.088174, "relative": 0.871078 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021985, "mean_abs": 0.002986, "relative": 0.051882 }, { "name": "camera_rotation_matrix", "l2": 3.00032, "mean_abs": 0.20592, "relative": 0.921349 }, { "name": "imu_accel_gyro", "l2": 17.091391, "mean_abs": 0.977441, "relative": 0.922139 }, { "name": "depth_confidence", "l2": 11.065748, "mean_abs": 0.199972, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.184081, "mean_abs": 0.107154, "relative": 0.938726 }, { "name": "video_fisheye_cam1", "l2": 5.562498, "mean_abs": 0.116632, "relative": 0.919156 }, { "name": "video_fisheye_cam2", "l2": 5.616932, "mean_abs": 0.113868, "relative": 0.89666 }, { "name": "video_fisheye_cam3", "l2": 5.634318, "mean_abs": 0.113551, "relative": 0.943399 }, { "name": "video_stereo_left", "l2": 5.334788, "mean_abs": 0.11571, "relative": 0.866914 }, { "name": "video_stereo_right", "l2": 6.71639, "mean_abs": 0.146752, "relative": 0.94687 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.087794, "mean_abs": 0.081831, "relative": 0.74719 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 640, "start_frame": 3200, "end_frame": 3219, "center_frame": 3209, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.848408, "mean_abs": 0.200838, "relative": 0.830755 }, { "name": "hand_right_joints", "l2": 10.947935, "mean_abs": 0.285597, "relative": 0.86299 }, { "name": "body_joints", "l2": 6.427849, "mean_abs": 0.087784, "relative": 0.868204 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015583, "mean_abs": 0.002234, "relative": 0.036773 }, { "name": "camera_rotation_matrix", "l2": 2.999971, "mean_abs": 0.20459, "relative": 0.921243 }, { "name": "imu_accel_gyro", "l2": 17.070059, "mean_abs": 0.974509, "relative": 0.920988 }, { "name": "depth_confidence", "l2": 11.050828, "mean_abs": 0.198904, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.189335, "mean_abs": 0.107059, "relative": 0.939678 }, { "name": "video_fisheye_cam1", "l2": 5.565131, "mean_abs": 0.116144, "relative": 0.919591 }, { "name": "video_fisheye_cam2", "l2": 5.625392, "mean_abs": 0.113649, "relative": 0.89801 }, { "name": "video_fisheye_cam3", "l2": 5.638673, "mean_abs": 0.11332, "relative": 0.944128 }, { "name": "video_stereo_left", "l2": 5.345389, "mean_abs": 0.114988, "relative": 0.868636 }, { "name": "video_stereo_right", "l2": 6.737096, "mean_abs": 0.145821, "relative": 0.949789 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.107794, "mean_abs": 0.082762, "relative": 0.754348 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 641, "start_frame": 3205, "end_frame": 3224, "center_frame": 3214, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.842595, "mean_abs": 0.201249, "relative": 0.830265 }, { "name": "hand_right_joints", "l2": 10.913876, "mean_abs": 0.284537, "relative": 0.860305 }, { "name": "body_joints", "l2": 6.403478, "mean_abs": 0.087601, "relative": 0.864913 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00918, "mean_abs": 0.001502, "relative": 0.021663 }, { "name": "camera_rotation_matrix", "l2": 3.000152, "mean_abs": 0.204983, "relative": 0.921298 }, { "name": "imu_accel_gyro", "l2": 17.061546, "mean_abs": 0.974509, "relative": 0.920529 }, { "name": "depth_confidence", "l2": 11.047801, "mean_abs": 0.199232, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.199897, "mean_abs": 0.107309, "relative": 0.94159 }, { "name": "video_fisheye_cam1", "l2": 5.574286, "mean_abs": 0.116477, "relative": 0.921104 }, { "name": "video_fisheye_cam2", "l2": 5.633207, "mean_abs": 0.113834, "relative": 0.899258 }, { "name": "video_fisheye_cam3", "l2": 5.641364, "mean_abs": 0.113653, "relative": 0.944578 }, { "name": "video_stereo_left", "l2": 5.355565, "mean_abs": 0.115321, "relative": 0.87029 }, { "name": "video_stereo_right", "l2": 6.751658, "mean_abs": 0.1464, "relative": 0.951842 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.17311, "mean_abs": 0.083848, "relative": 0.777724 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 642, "start_frame": 3210, "end_frame": 3229, "center_frame": 3219, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.826626, "mean_abs": 0.201318, "relative": 0.828918 }, { "name": "hand_right_joints", "l2": 10.877234, "mean_abs": 0.283104, "relative": 0.857417 }, { "name": "body_joints", "l2": 6.366887, "mean_abs": 0.087347, "relative": 0.85997 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015792, "mean_abs": 0.001973, "relative": 0.037267 }, { "name": "camera_rotation_matrix", "l2": 3.000624, "mean_abs": 0.20595, "relative": 0.921443 }, { "name": "imu_accel_gyro", "l2": 17.098942, "mean_abs": 0.977267, "relative": 0.922547 }, { "name": "depth_confidence", "l2": 11.042928, "mean_abs": 0.199576, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.206353, "mean_abs": 0.107823, "relative": 0.942759 }, { "name": "video_fisheye_cam1", "l2": 5.562414, "mean_abs": 0.116944, "relative": 0.919142 }, { "name": "video_fisheye_cam2", "l2": 5.632474, "mean_abs": 0.114217, "relative": 0.899141 }, { "name": "video_fisheye_cam3", "l2": 5.63747, "mean_abs": 0.114658, "relative": 0.943926 }, { "name": "video_stereo_left", "l2": 5.357553, "mean_abs": 0.116333, "relative": 0.870613 }, { "name": "video_stereo_right", "l2": 6.73102, "mean_abs": 0.14718, "relative": 0.948933 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.117929, "mean_abs": 0.083405, "relative": 0.757975 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 643, "start_frame": 3215, "end_frame": 3234, "center_frame": 3224, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.823691, "mean_abs": 0.201704, "relative": 0.82867 }, { "name": "hand_right_joints", "l2": 10.863349, "mean_abs": 0.280842, "relative": 0.856323 }, { "name": "body_joints", "l2": 6.348511, "mean_abs": 0.087326, "relative": 0.857488 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021567, "mean_abs": 0.002429, "relative": 0.050896 }, { "name": "camera_rotation_matrix", "l2": 3.000774, "mean_abs": 0.206292, "relative": 0.921489 }, { "name": "imu_accel_gyro", "l2": 17.077635, "mean_abs": 0.977368, "relative": 0.921397 }, { "name": "depth_confidence", "l2": 11.045638, "mean_abs": 0.200238, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.213765, "mean_abs": 0.108325, "relative": 0.944101 }, { "name": "video_fisheye_cam1", "l2": 5.562396, "mean_abs": 0.117265, "relative": 0.919139 }, { "name": "video_fisheye_cam2", "l2": 5.62979, "mean_abs": 0.114484, "relative": 0.898712 }, { "name": "video_fisheye_cam3", "l2": 5.615975, "mean_abs": 0.114722, "relative": 0.940327 }, { "name": "video_stereo_left", "l2": 5.353256, "mean_abs": 0.116892, "relative": 0.869915 }, { "name": "video_stereo_right", "l2": 6.715254, "mean_abs": 0.14735, "relative": 0.94671 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.078375, "mean_abs": 0.078197, "relative": 0.743819 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 644, "start_frame": 3220, "end_frame": 3239, "center_frame": 3229, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.822495, "mean_abs": 0.201753, "relative": 0.828569 }, { "name": "hand_right_joints", "l2": 10.856791, "mean_abs": 0.279897, "relative": 0.855806 }, { "name": "body_joints", "l2": 6.326943, "mean_abs": 0.087078, "relative": 0.854575 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.023087, "mean_abs": 0.0024, "relative": 0.054483 }, { "name": "camera_rotation_matrix", "l2": 3.000843, "mean_abs": 0.206089, "relative": 0.92151 }, { "name": "imu_accel_gyro", "l2": 17.066313, "mean_abs": 0.974489, "relative": 0.920786 }, { "name": "depth_confidence", "l2": 11.060555, "mean_abs": 0.2004, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.217662, "mean_abs": 0.108393, "relative": 0.944807 }, { "name": "video_fisheye_cam1", "l2": 5.553615, "mean_abs": 0.117019, "relative": 0.917688 }, { "name": "video_fisheye_cam2", "l2": 5.620598, "mean_abs": 0.114299, "relative": 0.897245 }, { "name": "video_fisheye_cam3", "l2": 5.595248, "mean_abs": 0.114561, "relative": 0.936857 }, { "name": "video_stereo_left", "l2": 5.331336, "mean_abs": 0.116668, "relative": 0.866353 }, { "name": "video_stereo_right", "l2": 6.684117, "mean_abs": 0.146846, "relative": 0.94232 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.125169, "mean_abs": 0.081143, "relative": 0.760566 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 645, "start_frame": 3225, "end_frame": 3244, "center_frame": 3234, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.807808, "mean_abs": 0.201327, "relative": 0.82733 }, { "name": "hand_right_joints", "l2": 10.849998, "mean_abs": 0.280358, "relative": 0.85527 }, { "name": "body_joints", "l2": 6.318976, "mean_abs": 0.086995, "relative": 0.853499 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.022275, "mean_abs": 0.002536, "relative": 0.052566 }, { "name": "camera_rotation_matrix", "l2": 3.000734, "mean_abs": 0.205151, "relative": 0.921477 }, { "name": "imu_accel_gyro", "l2": 17.06649, "mean_abs": 0.97894, "relative": 0.920796 }, { "name": "depth_confidence", "l2": 11.049189, "mean_abs": 0.199848, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.223259, "mean_abs": 0.108361, "relative": 0.945821 }, { "name": "video_fisheye_cam1", "l2": 5.544015, "mean_abs": 0.116793, "relative": 0.916102 }, { "name": "video_fisheye_cam2", "l2": 5.610453, "mean_abs": 0.113941, "relative": 0.895625 }, { "name": "video_fisheye_cam3", "l2": 5.57856, "mean_abs": 0.114731, "relative": 0.934063 }, { "name": "video_stereo_left", "l2": 5.311985, "mean_abs": 0.115931, "relative": 0.863208 }, { "name": "video_stereo_right", "l2": 6.647692, "mean_abs": 0.146151, "relative": 0.937185 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.092618, "mean_abs": 0.082692, "relative": 0.748917 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 646, "start_frame": 3230, "end_frame": 3249, "center_frame": 3239, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.808062, "mean_abs": 0.202233, "relative": 0.827352 }, { "name": "hand_right_joints", "l2": 10.851892, "mean_abs": 0.281841, "relative": 0.855419 }, { "name": "body_joints", "l2": 6.335598, "mean_abs": 0.087017, "relative": 0.855744 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020072, "mean_abs": 0.002473, "relative": 0.047368 }, { "name": "camera_rotation_matrix", "l2": 3.000248, "mean_abs": 0.203401, "relative": 0.921327 }, { "name": "imu_accel_gyro", "l2": 17.067064, "mean_abs": 0.97955, "relative": 0.920827 }, { "name": "depth_confidence", "l2": 11.050441, "mean_abs": 0.199382, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.225603, "mean_abs": 0.108121, "relative": 0.946245 }, { "name": "video_fisheye_cam1", "l2": 5.526666, "mean_abs": 0.115969, "relative": 0.913235 }, { "name": "video_fisheye_cam2", "l2": 5.606106, "mean_abs": 0.113559, "relative": 0.894931 }, { "name": "video_fisheye_cam3", "l2": 5.56408, "mean_abs": 0.114605, "relative": 0.931638 }, { "name": "video_stereo_left", "l2": 5.295653, "mean_abs": 0.114807, "relative": 0.860554 }, { "name": "video_stereo_right", "l2": 6.608994, "mean_abs": 0.144225, "relative": 0.93173 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.098721, "mean_abs": 0.082287, "relative": 0.751101 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 647, "start_frame": 3235, "end_frame": 3254, "center_frame": 3244, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.808763, "mean_abs": 0.202488, "relative": 0.827411 }, { "name": "hand_right_joints", "l2": 10.864113, "mean_abs": 0.282849, "relative": 0.856383 }, { "name": "body_joints", "l2": 6.354694, "mean_abs": 0.086674, "relative": 0.858323 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012656, "mean_abs": 0.001637, "relative": 0.029867 }, { "name": "camera_rotation_matrix", "l2": 2.999938, "mean_abs": 0.2021, "relative": 0.921232 }, { "name": "imu_accel_gyro", "l2": 17.059822, "mean_abs": 0.978142, "relative": 0.920436 }, { "name": "depth_confidence", "l2": 11.048926, "mean_abs": 0.19906, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.233332, "mean_abs": 0.107832, "relative": 0.947645 }, { "name": "video_fisheye_cam1", "l2": 5.521126, "mean_abs": 0.115033, "relative": 0.912319 }, { "name": "video_fisheye_cam2", "l2": 5.605423, "mean_abs": 0.113065, "relative": 0.894822 }, { "name": "video_fisheye_cam3", "l2": 5.538759, "mean_abs": 0.112596, "relative": 0.927398 }, { "name": "video_stereo_left", "l2": 5.284281, "mean_abs": 0.113404, "relative": 0.858706 }, { "name": "video_stereo_right", "l2": 6.594706, "mean_abs": 0.142409, "relative": 0.929715 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.066972, "mean_abs": 0.080547, "relative": 0.739738 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 648, "start_frame": 3240, "end_frame": 3259, "center_frame": 3249, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.7958, "mean_abs": 0.202079, "relative": 0.826318 }, { "name": "hand_right_joints", "l2": 10.854191, "mean_abs": 0.283371, "relative": 0.855601 }, { "name": "body_joints", "l2": 6.378236, "mean_abs": 0.086948, "relative": 0.861503 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006067, "mean_abs": 0.0009, "relative": 0.014316 }, { "name": "camera_rotation_matrix", "l2": 3.000198, "mean_abs": 0.203131, "relative": 0.921312 }, { "name": "imu_accel_gyro", "l2": 17.071705, "mean_abs": 0.975948, "relative": 0.921077 }, { "name": "depth_confidence", "l2": 11.047355, "mean_abs": 0.199308, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.242157, "mean_abs": 0.108363, "relative": 0.949243 }, { "name": "video_fisheye_cam1", "l2": 5.522544, "mean_abs": 0.115407, "relative": 0.912554 }, { "name": "video_fisheye_cam2", "l2": 5.614668, "mean_abs": 0.113565, "relative": 0.896298 }, { "name": "video_fisheye_cam3", "l2": 5.542371, "mean_abs": 0.113011, "relative": 0.928003 }, { "name": "video_stereo_left", "l2": 5.289952, "mean_abs": 0.114289, "relative": 0.859628 }, { "name": "video_stereo_right", "l2": 6.600895, "mean_abs": 0.143122, "relative": 0.930588 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.010705, "mean_abs": 0.076398, "relative": 0.719601 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 649, "start_frame": 3245, "end_frame": 3264, "center_frame": 3254, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.807157, "mean_abs": 0.20243, "relative": 0.827276 }, { "name": "hand_right_joints", "l2": 10.892896, "mean_abs": 0.286188, "relative": 0.858652 }, { "name": "body_joints", "l2": 6.411011, "mean_abs": 0.088075, "relative": 0.86593 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00885, "mean_abs": 0.001229, "relative": 0.020885 }, { "name": "camera_rotation_matrix", "l2": 3.000378, "mean_abs": 0.204102, "relative": 0.921367 }, { "name": "imu_accel_gyro", "l2": 17.061176, "mean_abs": 0.97164, "relative": 0.920509 }, { "name": "depth_confidence", "l2": 11.040788, "mean_abs": 0.199653, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.253019, "mean_abs": 0.108613, "relative": 0.951209 }, { "name": "video_fisheye_cam1", "l2": 5.526951, "mean_abs": 0.115821, "relative": 0.913282 }, { "name": "video_fisheye_cam2", "l2": 5.62834, "mean_abs": 0.114099, "relative": 0.898481 }, { "name": "video_fisheye_cam3", "l2": 5.553598, "mean_abs": 0.113394, "relative": 0.929883 }, { "name": "video_stereo_left", "l2": 5.306376, "mean_abs": 0.115332, "relative": 0.862297 }, { "name": "video_stereo_right", "l2": 6.622786, "mean_abs": 0.144244, "relative": 0.933674 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.026055, "mean_abs": 0.078637, "relative": 0.725095 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 650, "start_frame": 3250, "end_frame": 3269, "center_frame": 3259, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.807314, "mean_abs": 0.202191, "relative": 0.827289 }, { "name": "hand_right_joints", "l2": 10.925008, "mean_abs": 0.287753, "relative": 0.861183 }, { "name": "body_joints", "l2": 6.445243, "mean_abs": 0.088562, "relative": 0.870554 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015081, "mean_abs": 0.001934, "relative": 0.035588 }, { "name": "camera_rotation_matrix", "l2": 3.000366, "mean_abs": 0.204271, "relative": 0.921364 }, { "name": "imu_accel_gyro", "l2": 17.058666, "mean_abs": 0.967521, "relative": 0.920374 }, { "name": "depth_confidence", "l2": 11.018284, "mean_abs": 0.199155, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.262413, "mean_abs": 0.109078, "relative": 0.95291 }, { "name": "video_fisheye_cam1", "l2": 5.539561, "mean_abs": 0.116498, "relative": 0.915366 }, { "name": "video_fisheye_cam2", "l2": 5.647663, "mean_abs": 0.11459, "relative": 0.901565 }, { "name": "video_fisheye_cam3", "l2": 5.57126, "mean_abs": 0.114696, "relative": 0.93284 }, { "name": "video_stereo_left", "l2": 5.330283, "mean_abs": 0.116246, "relative": 0.866182 }, { "name": "video_stereo_right", "l2": 6.656319, "mean_abs": 0.145462, "relative": 0.938401 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.958161, "mean_abs": 0.074346, "relative": 0.700796 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 651, "start_frame": 3255, "end_frame": 3274, "center_frame": 3264, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.809522, "mean_abs": 0.202273, "relative": 0.827475 }, { "name": "hand_right_joints", "l2": 10.965047, "mean_abs": 0.290443, "relative": 0.864339 }, { "name": "body_joints", "l2": 6.472142, "mean_abs": 0.088861, "relative": 0.874187 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019661, "mean_abs": 0.002535, "relative": 0.046397 }, { "name": "camera_rotation_matrix", "l2": 3.000387, "mean_abs": 0.204616, "relative": 0.92137 }, { "name": "imu_accel_gyro", "l2": 17.052454, "mean_abs": 0.971346, "relative": 0.920039 }, { "name": "depth_confidence", "l2": 10.985959, "mean_abs": 0.198258, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.267935, "mean_abs": 0.10898, "relative": 0.95391 }, { "name": "video_fisheye_cam1", "l2": 5.540349, "mean_abs": 0.116226, "relative": 0.915496 }, { "name": "video_fisheye_cam2", "l2": 5.662991, "mean_abs": 0.114771, "relative": 0.904012 }, { "name": "video_fisheye_cam3", "l2": 5.576596, "mean_abs": 0.114147, "relative": 0.933734 }, { "name": "video_stereo_left", "l2": 5.344912, "mean_abs": 0.116371, "relative": 0.868559 }, { "name": "video_stereo_right", "l2": 6.674954, "mean_abs": 0.145828, "relative": 0.941028 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.104712, "mean_abs": 0.086376, "relative": 0.753245 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 652, "start_frame": 3260, "end_frame": 3279, "center_frame": 3269, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.819139, "mean_abs": 0.202235, "relative": 0.828286 }, { "name": "hand_right_joints", "l2": 10.980197, "mean_abs": 0.289893, "relative": 0.865533 }, { "name": "body_joints", "l2": 6.494057, "mean_abs": 0.088847, "relative": 0.877147 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.023109, "mean_abs": 0.002879, "relative": 0.054534 }, { "name": "camera_rotation_matrix", "l2": 3.000331, "mean_abs": 0.204419, "relative": 0.921353 }, { "name": "imu_accel_gyro", "l2": 17.063267, "mean_abs": 0.970918, "relative": 0.920622 }, { "name": "depth_confidence", "l2": 10.986213, "mean_abs": 0.198409, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.274017, "mean_abs": 0.109133, "relative": 0.955012 }, { "name": "video_fisheye_cam1", "l2": 5.54223, "mean_abs": 0.116315, "relative": 0.915807 }, { "name": "video_fisheye_cam2", "l2": 5.676106, "mean_abs": 0.114972, "relative": 0.906106 }, { "name": "video_fisheye_cam3", "l2": 5.582526, "mean_abs": 0.114379, "relative": 0.934727 }, { "name": "video_stereo_left", "l2": 5.360975, "mean_abs": 0.116494, "relative": 0.871169 }, { "name": "video_stereo_right", "l2": 6.690393, "mean_abs": 0.145903, "relative": 0.943205 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.022452, "mean_abs": 0.078376, "relative": 0.723805 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 653, "start_frame": 3265, "end_frame": 3284, "center_frame": 3274, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.817096, "mean_abs": 0.202103, "relative": 0.828114 }, { "name": "hand_right_joints", "l2": 10.982157, "mean_abs": 0.289754, "relative": 0.865688 }, { "name": "body_joints", "l2": 6.51635, "mean_abs": 0.088688, "relative": 0.880158 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021224, "mean_abs": 0.002636, "relative": 0.050085 }, { "name": "camera_rotation_matrix", "l2": 3.000214, "mean_abs": 0.204247, "relative": 0.921317 }, { "name": "imu_accel_gyro", "l2": 17.079071, "mean_abs": 0.971599, "relative": 0.921475 }, { "name": "depth_confidence", "l2": 10.987007, "mean_abs": 0.197869, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.288258, "mean_abs": 0.109408, "relative": 0.95759 }, { "name": "video_fisheye_cam1", "l2": 5.547304, "mean_abs": 0.116332, "relative": 0.916645 }, { "name": "video_fisheye_cam2", "l2": 5.692839, "mean_abs": 0.115193, "relative": 0.908777 }, { "name": "video_fisheye_cam3", "l2": 5.603323, "mean_abs": 0.114688, "relative": 0.938209 }, { "name": "video_stereo_left", "l2": 5.379068, "mean_abs": 0.116621, "relative": 0.874109 }, { "name": "video_stereo_right", "l2": 6.720911, "mean_abs": 0.146255, "relative": 0.947507 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.083404, "mean_abs": 0.082367, "relative": 0.745619 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 654, "start_frame": 3270, "end_frame": 3289, "center_frame": 3279, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.81598, "mean_abs": 0.20197, "relative": 0.82802 }, { "name": "hand_right_joints", "l2": 10.963338, "mean_abs": 0.29096, "relative": 0.864204 }, { "name": "body_joints", "l2": 6.513326, "mean_abs": 0.088678, "relative": 0.87975 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014439, "mean_abs": 0.001814, "relative": 0.034073 }, { "name": "camera_rotation_matrix", "l2": 2.99999, "mean_abs": 0.204131, "relative": 0.921248 }, { "name": "imu_accel_gyro", "l2": 17.05654, "mean_abs": 0.971794, "relative": 0.920259 }, { "name": "depth_confidence", "l2": 10.983912, "mean_abs": 0.197698, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.29486, "mean_abs": 0.109494, "relative": 0.958786 }, { "name": "video_fisheye_cam1", "l2": 5.551073, "mean_abs": 0.116244, "relative": 0.917268 }, { "name": "video_fisheye_cam2", "l2": 5.708462, "mean_abs": 0.115414, "relative": 0.911271 }, { "name": "video_fisheye_cam3", "l2": 5.610216, "mean_abs": 0.114431, "relative": 0.939363 }, { "name": "video_stereo_left", "l2": 5.394791, "mean_abs": 0.116554, "relative": 0.876664 }, { "name": "video_stereo_right", "l2": 6.740945, "mean_abs": 0.146453, "relative": 0.950332 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.147652, "mean_abs": 0.087729, "relative": 0.768612 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 655, "start_frame": 3275, "end_frame": 3294, "center_frame": 3284, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.819914, "mean_abs": 0.201908, "relative": 0.828352 }, { "name": "hand_right_joints", "l2": 10.934377, "mean_abs": 0.289334, "relative": 0.861921 }, { "name": "body_joints", "l2": 6.494863, "mean_abs": 0.088692, "relative": 0.877256 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011124, "mean_abs": 0.001573, "relative": 0.026251 }, { "name": "camera_rotation_matrix", "l2": 2.999946, "mean_abs": 0.204006, "relative": 0.921235 }, { "name": "imu_accel_gyro", "l2": 17.063734, "mean_abs": 0.970095, "relative": 0.920647 }, { "name": "depth_confidence", "l2": 10.980763, "mean_abs": 0.197299, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.30385, "mean_abs": 0.10951, "relative": 0.960414 }, { "name": "video_fisheye_cam1", "l2": 5.555573, "mean_abs": 0.116074, "relative": 0.918011 }, { "name": "video_fisheye_cam2", "l2": 5.719943, "mean_abs": 0.11553, "relative": 0.913104 }, { "name": "video_fisheye_cam3", "l2": 5.615322, "mean_abs": 0.113791, "relative": 0.940218 }, { "name": "video_stereo_left", "l2": 5.408029, "mean_abs": 0.116534, "relative": 0.878815 }, { "name": "video_stereo_right", "l2": 6.760137, "mean_abs": 0.146655, "relative": 0.953037 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.065866, "mean_abs": 0.07982, "relative": 0.739342 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 656, "start_frame": 3280, "end_frame": 3299, "center_frame": 3289, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.815859, "mean_abs": 0.201275, "relative": 0.82801 }, { "name": "hand_right_joints", "l2": 10.915198, "mean_abs": 0.28816, "relative": 0.86041 }, { "name": "body_joints", "l2": 6.476373, "mean_abs": 0.088655, "relative": 0.874758 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012248, "mean_abs": 0.001671, "relative": 0.028904 }, { "name": "camera_rotation_matrix", "l2": 3.000419, "mean_abs": 0.204045, "relative": 0.92138 }, { "name": "imu_accel_gyro", "l2": 17.061106, "mean_abs": 0.972331, "relative": 0.920505 }, { "name": "depth_confidence", "l2": 10.996907, "mean_abs": 0.19749, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.308583, "mean_abs": 0.109426, "relative": 0.961271 }, { "name": "video_fisheye_cam1", "l2": 5.562021, "mean_abs": 0.116107, "relative": 0.919077 }, { "name": "video_fisheye_cam2", "l2": 5.727307, "mean_abs": 0.115572, "relative": 0.914279 }, { "name": "video_fisheye_cam3", "l2": 5.618711, "mean_abs": 0.11395, "relative": 0.940785 }, { "name": "video_stereo_left", "l2": 5.414396, "mean_abs": 0.116433, "relative": 0.87985 }, { "name": "video_stereo_right", "l2": 6.771602, "mean_abs": 0.146686, "relative": 0.954654 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.013793, "mean_abs": 0.076672, "relative": 0.720706 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 657, "start_frame": 3285, "end_frame": 3304, "center_frame": 3294, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.820652, "mean_abs": 0.200202, "relative": 0.828414 }, { "name": "hand_right_joints", "l2": 10.889841, "mean_abs": 0.287714, "relative": 0.858411 }, { "name": "body_joints", "l2": 6.450105, "mean_abs": 0.088223, "relative": 0.87121 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013118, "mean_abs": 0.001837, "relative": 0.030958 }, { "name": "camera_rotation_matrix", "l2": 2.999852, "mean_abs": 0.204522, "relative": 0.921206 }, { "name": "imu_accel_gyro", "l2": 17.096029, "mean_abs": 0.978025, "relative": 0.92239 }, { "name": "depth_confidence", "l2": 11.004173, "mean_abs": 0.197629, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.308595, "mean_abs": 0.109395, "relative": 0.961273 }, { "name": "video_fisheye_cam1", "l2": 5.567461, "mean_abs": 0.116226, "relative": 0.919976 }, { "name": "video_fisheye_cam2", "l2": 5.730481, "mean_abs": 0.115538, "relative": 0.914786 }, { "name": "video_fisheye_cam3", "l2": 5.609653, "mean_abs": 0.113835, "relative": 0.939269 }, { "name": "video_stereo_left", "l2": 5.411483, "mean_abs": 0.116543, "relative": 0.879377 }, { "name": "video_stereo_right", "l2": 6.767222, "mean_abs": 0.14658, "relative": 0.954036 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.021947, "mean_abs": 0.076919, "relative": 0.723625 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 658, "start_frame": 3290, "end_frame": 3309, "center_frame": 3299, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.818506, "mean_abs": 0.199744, "relative": 0.828233 }, { "name": "hand_right_joints", "l2": 10.861587, "mean_abs": 0.283646, "relative": 0.856184 }, { "name": "body_joints", "l2": 6.405184, "mean_abs": 0.087539, "relative": 0.865143 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014874, "mean_abs": 0.002158, "relative": 0.035101 }, { "name": "camera_rotation_matrix", "l2": 3.000159, "mean_abs": 0.20538, "relative": 0.9213 }, { "name": "imu_accel_gyro", "l2": 17.102158, "mean_abs": 0.981037, "relative": 0.92272 }, { "name": "depth_confidence", "l2": 11.016793, "mean_abs": 0.198358, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.310481, "mean_abs": 0.109617, "relative": 0.961614 }, { "name": "video_fisheye_cam1", "l2": 5.575432, "mean_abs": 0.116584, "relative": 0.921293 }, { "name": "video_fisheye_cam2", "l2": 5.728771, "mean_abs": 0.11576, "relative": 0.914513 }, { "name": "video_fisheye_cam3", "l2": 5.609288, "mean_abs": 0.113976, "relative": 0.939208 }, { "name": "video_stereo_left", "l2": 5.40536, "mean_abs": 0.117001, "relative": 0.878382 }, { "name": "video_stereo_right", "l2": 6.767509, "mean_abs": 0.147052, "relative": 0.954077 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.982559, "mean_abs": 0.077172, "relative": 0.709528 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 659, "start_frame": 3295, "end_frame": 3314, "center_frame": 3304, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.817236, "mean_abs": 0.199531, "relative": 0.828126 }, { "name": "hand_right_joints", "l2": 10.86125, "mean_abs": 0.283078, "relative": 0.856157 }, { "name": "body_joints", "l2": 6.382871, "mean_abs": 0.08747, "relative": 0.862129 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016534, "mean_abs": 0.002372, "relative": 0.039019 }, { "name": "camera_rotation_matrix", "l2": 3.000458, "mean_abs": 0.205916, "relative": 0.921392 }, { "name": "imu_accel_gyro", "l2": 17.083061, "mean_abs": 0.976624, "relative": 0.92169 }, { "name": "depth_confidence", "l2": 11.040033, "mean_abs": 0.199361, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.302959, "mean_abs": 0.109498, "relative": 0.960252 }, { "name": "video_fisheye_cam1", "l2": 5.575667, "mean_abs": 0.1166, "relative": 0.921332 }, { "name": "video_fisheye_cam2", "l2": 5.718699, "mean_abs": 0.115811, "relative": 0.912905 }, { "name": "video_fisheye_cam3", "l2": 5.582241, "mean_abs": 0.113224, "relative": 0.934679 }, { "name": "video_stereo_left", "l2": 5.394613, "mean_abs": 0.117224, "relative": 0.876635 }, { "name": "video_stereo_right", "l2": 6.750526, "mean_abs": 0.146736, "relative": 0.951683 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.982178, "mean_abs": 0.076662, "relative": 0.709392 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 660, "start_frame": 3300, "end_frame": 3319, "center_frame": 3309, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.817823, "mean_abs": 0.198931, "relative": 0.828175 }, { "name": "hand_right_joints", "l2": 10.842307, "mean_abs": 0.280996, "relative": 0.854664 }, { "name": "body_joints", "l2": 6.369351, "mean_abs": 0.087523, "relative": 0.860303 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018948, "mean_abs": 0.002344, "relative": 0.044715 }, { "name": "camera_rotation_matrix", "l2": 3.000594, "mean_abs": 0.206051, "relative": 0.921434 }, { "name": "imu_accel_gyro", "l2": 17.061361, "mean_abs": 0.97551, "relative": 0.920519 }, { "name": "depth_confidence", "l2": 11.025423, "mean_abs": 0.19901, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.294297, "mean_abs": 0.109492, "relative": 0.958684 }, { "name": "video_fisheye_cam1", "l2": 5.576359, "mean_abs": 0.116711, "relative": 0.921446 }, { "name": "video_fisheye_cam2", "l2": 5.704245, "mean_abs": 0.115756, "relative": 0.910598 }, { "name": "video_fisheye_cam3", "l2": 5.563268, "mean_abs": 0.113076, "relative": 0.931502 }, { "name": "video_stereo_left", "l2": 5.373206, "mean_abs": 0.117077, "relative": 0.873157 }, { "name": "video_stereo_right", "l2": 6.727896, "mean_abs": 0.146474, "relative": 0.948492 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.896019, "mean_abs": 0.072882, "relative": 0.678557 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 661, "start_frame": 3305, "end_frame": 3324, "center_frame": 3314, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.819764, "mean_abs": 0.199035, "relative": 0.828339 }, { "name": "hand_right_joints", "l2": 10.839803, "mean_abs": 0.280979, "relative": 0.854466 }, { "name": "body_joints", "l2": 6.356502, "mean_abs": 0.086914, "relative": 0.858568 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018406, "mean_abs": 0.002198, "relative": 0.043436 }, { "name": "camera_rotation_matrix", "l2": 3.000529, "mean_abs": 0.205669, "relative": 0.921414 }, { "name": "imu_accel_gyro", "l2": 17.03734, "mean_abs": 0.974556, "relative": 0.919223 }, { "name": "depth_confidence", "l2": 11.047985, "mean_abs": 0.199474, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.287804, "mean_abs": 0.109352, "relative": 0.957508 }, { "name": "video_fisheye_cam1", "l2": 5.577112, "mean_abs": 0.116753, "relative": 0.921571 }, { "name": "video_fisheye_cam2", "l2": 5.69023, "mean_abs": 0.115494, "relative": 0.908361 }, { "name": "video_fisheye_cam3", "l2": 5.55747, "mean_abs": 0.112958, "relative": 0.930531 }, { "name": "video_stereo_left", "l2": 5.352698, "mean_abs": 0.116608, "relative": 0.869824 }, { "name": "video_stereo_right", "l2": 6.708796, "mean_abs": 0.145949, "relative": 0.945799 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.007004, "mean_abs": 0.078187, "relative": 0.718277 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 662, "start_frame": 3310, "end_frame": 3329, "center_frame": 3319, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.81417, "mean_abs": 0.199603, "relative": 0.827867 }, { "name": "hand_right_joints", "l2": 10.850679, "mean_abs": 0.281716, "relative": 0.855324 }, { "name": "body_joints", "l2": 6.365513, "mean_abs": 0.086826, "relative": 0.859785 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013948, "mean_abs": 0.00182, "relative": 0.032914 }, { "name": "camera_rotation_matrix", "l2": 3.000242, "mean_abs": 0.204582, "relative": 0.921326 }, { "name": "imu_accel_gyro", "l2": 17.044987, "mean_abs": 0.968714, "relative": 0.919636 }, { "name": "depth_confidence", "l2": 11.062522, "mean_abs": 0.199566, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.278217, "mean_abs": 0.109119, "relative": 0.955772 }, { "name": "video_fisheye_cam1", "l2": 5.567756, "mean_abs": 0.116223, "relative": 0.920025 }, { "name": "video_fisheye_cam2", "l2": 5.675498, "mean_abs": 0.114921, "relative": 0.906009 }, { "name": "video_fisheye_cam3", "l2": 5.545998, "mean_abs": 0.11261, "relative": 0.92861 }, { "name": "video_stereo_left", "l2": 5.329895, "mean_abs": 0.115624, "relative": 0.866119 }, { "name": "video_stereo_right", "l2": 6.685869, "mean_abs": 0.145025, "relative": 0.942567 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.014853, "mean_abs": 0.07384, "relative": 0.721086 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 663, "start_frame": 3315, "end_frame": 3334, "center_frame": 3324, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.808973, "mean_abs": 0.20047, "relative": 0.827429 }, { "name": "hand_right_joints", "l2": 10.872455, "mean_abs": 0.283701, "relative": 0.85704 }, { "name": "body_joints", "l2": 6.394422, "mean_abs": 0.087243, "relative": 0.863689 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008474, "mean_abs": 0.001227, "relative": 0.019997 }, { "name": "camera_rotation_matrix", "l2": 2.999717, "mean_abs": 0.203303, "relative": 0.921164 }, { "name": "imu_accel_gyro", "l2": 17.075552, "mean_abs": 0.972361, "relative": 0.921285 }, { "name": "depth_confidence", "l2": 11.095334, "mean_abs": 0.199616, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.265225, "mean_abs": 0.108593, "relative": 0.95342 }, { "name": "video_fisheye_cam1", "l2": 5.562497, "mean_abs": 0.11568, "relative": 0.919156 }, { "name": "video_fisheye_cam2", "l2": 5.660549, "mean_abs": 0.114377, "relative": 0.903622 }, { "name": "video_fisheye_cam3", "l2": 5.542768, "mean_abs": 0.112249, "relative": 0.92807 }, { "name": "video_stereo_left", "l2": 5.310773, "mean_abs": 0.114355, "relative": 0.863011 }, { "name": "video_stereo_right", "l2": 6.665392, "mean_abs": 0.143592, "relative": 0.93968 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.081789, "mean_abs": 0.079967, "relative": 0.745041 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 664, "start_frame": 3320, "end_frame": 3339, "center_frame": 3329, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.809503, "mean_abs": 0.200782, "relative": 0.827473 }, { "name": "hand_right_joints", "l2": 10.889934, "mean_abs": 0.285163, "relative": 0.858418 }, { "name": "body_joints", "l2": 6.428267, "mean_abs": 0.087921, "relative": 0.868261 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009061, "mean_abs": 0.001315, "relative": 0.021382 }, { "name": "camera_rotation_matrix", "l2": 3.000331, "mean_abs": 0.20482, "relative": 0.921353 }, { "name": "imu_accel_gyro", "l2": 17.076288, "mean_abs": 0.971321, "relative": 0.921325 }, { "name": "depth_confidence", "l2": 11.181368, "mean_abs": 0.20042, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.251091, "mean_abs": 0.108599, "relative": 0.95086 }, { "name": "video_fisheye_cam1", "l2": 5.552537, "mean_abs": 0.116048, "relative": 0.91751 }, { "name": "video_fisheye_cam2", "l2": 5.651132, "mean_abs": 0.114523, "relative": 0.902119 }, { "name": "video_fisheye_cam3", "l2": 5.536891, "mean_abs": 0.112446, "relative": 0.927086 }, { "name": "video_stereo_left", "l2": 5.300121, "mean_abs": 0.114776, "relative": 0.86128 }, { "name": "video_stereo_right", "l2": 6.653715, "mean_abs": 0.143973, "relative": 0.938034 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.14558, "mean_abs": 0.084862, "relative": 0.767871 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 665, "start_frame": 3325, "end_frame": 3344, "center_frame": 3334, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.806755, "mean_abs": 0.201116, "relative": 0.827242 }, { "name": "hand_right_joints", "l2": 10.910147, "mean_abs": 0.28649, "relative": 0.860011 }, { "name": "body_joints", "l2": 6.462605, "mean_abs": 0.088532, "relative": 0.872899 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013831, "mean_abs": 0.001723, "relative": 0.032638 }, { "name": "camera_rotation_matrix", "l2": 3.000577, "mean_abs": 0.205857, "relative": 0.921429 }, { "name": "imu_accel_gyro", "l2": 17.072948, "mean_abs": 0.966976, "relative": 0.921144 }, { "name": "depth_confidence", "l2": 11.227692, "mean_abs": 0.201008, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.242323, "mean_abs": 0.108495, "relative": 0.949273 }, { "name": "video_fisheye_cam1", "l2": 5.549976, "mean_abs": 0.116206, "relative": 0.917087 }, { "name": "video_fisheye_cam2", "l2": 5.648007, "mean_abs": 0.114516, "relative": 0.90162 }, { "name": "video_fisheye_cam3", "l2": 5.543158, "mean_abs": 0.112711, "relative": 0.928135 }, { "name": "video_stereo_left", "l2": 5.305379, "mean_abs": 0.115314, "relative": 0.862135 }, { "name": "video_stereo_right", "l2": 6.659169, "mean_abs": 0.144543, "relative": 0.938803 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.196309, "mean_abs": 0.086633, "relative": 0.786026 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 666, "start_frame": 3330, "end_frame": 3349, "center_frame": 3339, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.808959, "mean_abs": 0.201191, "relative": 0.827428 }, { "name": "hand_right_joints", "l2": 10.913441, "mean_abs": 0.286901, "relative": 0.860271 }, { "name": "body_joints", "l2": 6.488338, "mean_abs": 0.088878, "relative": 0.876375 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017554, "mean_abs": 0.002179, "relative": 0.041424 }, { "name": "camera_rotation_matrix", "l2": 3.000648, "mean_abs": 0.206275, "relative": 0.92145 }, { "name": "imu_accel_gyro", "l2": 17.083178, "mean_abs": 0.975289, "relative": 0.921696 }, { "name": "depth_confidence", "l2": 11.224815, "mean_abs": 0.200954, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.230821, "mean_abs": 0.108408, "relative": 0.94719 }, { "name": "video_fisheye_cam1", "l2": 5.539061, "mean_abs": 0.116082, "relative": 0.915283 }, { "name": "video_fisheye_cam2", "l2": 5.644156, "mean_abs": 0.11445, "relative": 0.901005 }, { "name": "video_fisheye_cam3", "l2": 5.5407, "mean_abs": 0.113179, "relative": 0.927723 }, { "name": "video_stereo_left", "l2": 5.306409, "mean_abs": 0.115469, "relative": 0.862302 }, { "name": "video_stereo_right", "l2": 6.655988, "mean_abs": 0.144945, "relative": 0.938355 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.105283, "mean_abs": 0.085527, "relative": 0.753449 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 667, "start_frame": 3335, "end_frame": 3354, "center_frame": 3344, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.823477, "mean_abs": 0.201316, "relative": 0.828652 }, { "name": "hand_right_joints", "l2": 10.916388, "mean_abs": 0.287418, "relative": 0.860503 }, { "name": "body_joints", "l2": 6.516349, "mean_abs": 0.089075, "relative": 0.880158 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017022, "mean_abs": 0.002067, "relative": 0.040171 }, { "name": "camera_rotation_matrix", "l2": 3.000464, "mean_abs": 0.206476, "relative": 0.921394 }, { "name": "imu_accel_gyro", "l2": 17.014326, "mean_abs": 0.968898, "relative": 0.917982 }, { "name": "depth_confidence", "l2": 11.22475, "mean_abs": 0.201123, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.22182, "mean_abs": 0.108202, "relative": 0.94556 }, { "name": "video_fisheye_cam1", "l2": 5.53725, "mean_abs": 0.115951, "relative": 0.914984 }, { "name": "video_fisheye_cam2", "l2": 5.645328, "mean_abs": 0.114421, "relative": 0.901193 }, { "name": "video_fisheye_cam3", "l2": 5.544511, "mean_abs": 0.113067, "relative": 0.928362 }, { "name": "video_stereo_left", "l2": 5.318334, "mean_abs": 0.115544, "relative": 0.86424 }, { "name": "video_stereo_right", "l2": 6.669331, "mean_abs": 0.145328, "relative": 0.940236 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.155275, "mean_abs": 0.085727, "relative": 0.771341 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 668, "start_frame": 3340, "end_frame": 3359, "center_frame": 3349, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.825323, "mean_abs": 0.201066, "relative": 0.828808 }, { "name": "hand_right_joints", "l2": 10.912333, "mean_abs": 0.288366, "relative": 0.860184 }, { "name": "body_joints", "l2": 6.531889, "mean_abs": 0.088977, "relative": 0.882257 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016425, "mean_abs": 0.002079, "relative": 0.038761 }, { "name": "camera_rotation_matrix", "l2": 3.000324, "mean_abs": 0.206333, "relative": 0.921351 }, { "name": "imu_accel_gyro", "l2": 17.014929, "mean_abs": 0.969084, "relative": 0.918014 }, { "name": "depth_confidence", "l2": 11.213283, "mean_abs": 0.20156, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.215001, "mean_abs": 0.107922, "relative": 0.944325 }, { "name": "video_fisheye_cam1", "l2": 5.535348, "mean_abs": 0.115839, "relative": 0.91467 }, { "name": "video_fisheye_cam2", "l2": 5.646614, "mean_abs": 0.114364, "relative": 0.901398 }, { "name": "video_fisheye_cam3", "l2": 5.55514, "mean_abs": 0.113034, "relative": 0.930141 }, { "name": "video_stereo_left", "l2": 5.332402, "mean_abs": 0.115415, "relative": 0.866526 }, { "name": "video_stereo_right", "l2": 6.692125, "mean_abs": 0.145678, "relative": 0.943449 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.188026, "mean_abs": 0.084752, "relative": 0.783062 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 669, "start_frame": 3345, "end_frame": 3364, "center_frame": 3354, "action": "Lift gooseneck kettle", "subtask": "Lift gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.823544, "mean_abs": 0.200985, "relative": 0.828658 }, { "name": "hand_right_joints", "l2": 10.903365, "mean_abs": 0.288903, "relative": 0.859477 }, { "name": "body_joints", "l2": 6.531968, "mean_abs": 0.089349, "relative": 0.882268 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015552, "mean_abs": 0.002127, "relative": 0.036699 }, { "name": "camera_rotation_matrix", "l2": 3.000223, "mean_abs": 0.205911, "relative": 0.92132 }, { "name": "imu_accel_gyro", "l2": 17.03602, "mean_abs": 0.97621, "relative": 0.919152 }, { "name": "depth_confidence", "l2": 11.109515, "mean_abs": 0.199678, "relative": 0.000369 }, { "name": "video_fisheye_cam0", "l2": 5.205334, "mean_abs": 0.107837, "relative": 0.942575 }, { "name": "video_fisheye_cam1", "l2": 5.524384, "mean_abs": 0.115678, "relative": 0.912858 }, { "name": "video_fisheye_cam2", "l2": 5.641724, "mean_abs": 0.114139, "relative": 0.900617 }, { "name": "video_fisheye_cam3", "l2": 5.552523, "mean_abs": 0.113105, "relative": 0.929703 }, { "name": "video_stereo_left", "l2": 5.335911, "mean_abs": 0.115251, "relative": 0.867096 }, { "name": "video_stereo_right", "l2": 6.692741, "mean_abs": 0.145465, "relative": 0.943536 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.059124, "mean_abs": 0.080561, "relative": 0.73693 }, { "name": "caption_objects_interaction_text", "l2": 1.811123, "mean_abs": 0.014676, "relative": 0.831624 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 670, "start_frame": 3350, "end_frame": 3369, "center_frame": 3359, "action": "Lift gooseneck kettle", "subtask": "", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.829467, "mean_abs": 0.200734, "relative": 0.829157 }, { "name": "hand_right_joints", "l2": 10.887904, "mean_abs": 0.287497, "relative": 0.858258 }, { "name": "body_joints", "l2": 6.509136, "mean_abs": 0.089374, "relative": 0.879184 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013265, "mean_abs": 0.001875, "relative": 0.031303 }, { "name": "camera_rotation_matrix", "l2": 2.999997, "mean_abs": 0.205278, "relative": 0.92125 }, { "name": "imu_accel_gyro", "l2": 17.063078, "mean_abs": 0.977265, "relative": 0.920612 }, { "name": "depth_confidence", "l2": 11.079096, "mean_abs": 0.199526, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.196656, "mean_abs": 0.107769, "relative": 0.941003 }, { "name": "video_fisheye_cam1", "l2": 5.520781, "mean_abs": 0.115583, "relative": 0.912262 }, { "name": "video_fisheye_cam2", "l2": 5.638243, "mean_abs": 0.113981, "relative": 0.900062 }, { "name": "video_fisheye_cam3", "l2": 5.554876, "mean_abs": 0.113589, "relative": 0.930097 }, { "name": "video_stereo_left", "l2": 5.342153, "mean_abs": 0.115017, "relative": 0.86811 }, { "name": "video_stereo_right", "l2": 6.699148, "mean_abs": 0.14537, "relative": 0.944439 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.095721, "mean_abs": 0.081167, "relative": 0.750027 }, { "name": "caption_objects_interaction_text", "l2": 1.811123, "mean_abs": 0.014754, "relative": 0.831624 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 671, "start_frame": 3355, "end_frame": 3374, "center_frame": 3364, "action": "Lift gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.832068, "mean_abs": 0.200487, "relative": 0.829377 }, { "name": "hand_right_joints", "l2": 10.873433, "mean_abs": 0.28671, "relative": 0.857117 }, { "name": "body_joints", "l2": 6.485256, "mean_abs": 0.089039, "relative": 0.875958 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009464, "mean_abs": 0.001403, "relative": 0.022334 }, { "name": "camera_rotation_matrix", "l2": 2.999929, "mean_abs": 0.204795, "relative": 0.92123 }, { "name": "imu_accel_gyro", "l2": 17.084167, "mean_abs": 0.975466, "relative": 0.92175 }, { "name": "depth_confidence", "l2": 11.032343, "mean_abs": 0.199084, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.191332, "mean_abs": 0.107332, "relative": 0.940039 }, { "name": "video_fisheye_cam1", "l2": 5.523702, "mean_abs": 0.115198, "relative": 0.912745 }, { "name": "video_fisheye_cam2", "l2": 5.637288, "mean_abs": 0.113785, "relative": 0.899909 }, { "name": "video_fisheye_cam3", "l2": 5.558483, "mean_abs": 0.112866, "relative": 0.930701 }, { "name": "video_stereo_left", "l2": 5.3504, "mean_abs": 0.114972, "relative": 0.869451 }, { "name": "video_stereo_right", "l2": 6.710292, "mean_abs": 0.145087, "relative": 0.94601 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.998747, "mean_abs": 0.076489, "relative": 0.715322 }, { "name": "caption_objects_interaction_text", "l2": 1.811123, "mean_abs": 0.014597, "relative": 0.831624 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 672, "start_frame": 3360, "end_frame": 3379, "center_frame": 3369, "action": "Lift gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.830605, "mean_abs": 0.200018, "relative": 0.829253 }, { "name": "hand_right_joints", "l2": 10.841221, "mean_abs": 0.285411, "relative": 0.854578 }, { "name": "body_joints", "l2": 6.448533, "mean_abs": 0.088716, "relative": 0.870998 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009186, "mean_abs": 0.001297, "relative": 0.021678 }, { "name": "camera_rotation_matrix", "l2": 2.999915, "mean_abs": 0.205561, "relative": 0.921225 }, { "name": "imu_accel_gyro", "l2": 17.064184, "mean_abs": 0.980395, "relative": 0.920672 }, { "name": "depth_confidence", "l2": 11.029177, "mean_abs": 0.199065, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.182391, "mean_abs": 0.107293, "relative": 0.93842 }, { "name": "video_fisheye_cam1", "l2": 5.519504, "mean_abs": 0.115574, "relative": 0.912051 }, { "name": "video_fisheye_cam2", "l2": 5.626029, "mean_abs": 0.113797, "relative": 0.898112 }, { "name": "video_fisheye_cam3", "l2": 5.549385, "mean_abs": 0.113652, "relative": 0.929178 }, { "name": "video_stereo_left", "l2": 5.341795, "mean_abs": 0.115377, "relative": 0.868052 }, { "name": "video_stereo_right", "l2": 6.696332, "mean_abs": 0.145569, "relative": 0.944042 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.014875, "mean_abs": 0.078554, "relative": 0.721093 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01139, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 673, "start_frame": 3365, "end_frame": 3384, "center_frame": 3374, "action": "Lift gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.833362, "mean_abs": 0.19966, "relative": 0.829486 }, { "name": "hand_right_joints", "l2": 10.825429, "mean_abs": 0.283125, "relative": 0.853333 }, { "name": "body_joints", "l2": 6.427964, "mean_abs": 0.088406, "relative": 0.86822 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013096, "mean_abs": 0.00175, "relative": 0.030904 }, { "name": "camera_rotation_matrix", "l2": 3.000118, "mean_abs": 0.206365, "relative": 0.921288 }, { "name": "imu_accel_gyro", "l2": 17.046453, "mean_abs": 0.975918, "relative": 0.919715 }, { "name": "depth_confidence", "l2": 11.032741, "mean_abs": 0.199101, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.177166, "mean_abs": 0.107151, "relative": 0.937474 }, { "name": "video_fisheye_cam1", "l2": 5.523132, "mean_abs": 0.115618, "relative": 0.912651 }, { "name": "video_fisheye_cam2", "l2": 5.618805, "mean_abs": 0.113819, "relative": 0.896959 }, { "name": "video_fisheye_cam3", "l2": 5.545033, "mean_abs": 0.112958, "relative": 0.928449 }, { "name": "video_stereo_left", "l2": 5.336513, "mean_abs": 0.115492, "relative": 0.867194 }, { "name": "video_stereo_right", "l2": 6.692765, "mean_abs": 0.145676, "relative": 0.943539 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.064615, "mean_abs": 0.080272, "relative": 0.738895 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01139, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 674, "start_frame": 3370, "end_frame": 3389, "center_frame": 3379, "action": "Lift gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.832448, "mean_abs": 0.200252, "relative": 0.829409 }, { "name": "hand_right_joints", "l2": 10.818862, "mean_abs": 0.281473, "relative": 0.852816 }, { "name": "body_joints", "l2": 6.405514, "mean_abs": 0.088207, "relative": 0.865188 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017173, "mean_abs": 0.002087, "relative": 0.040525 }, { "name": "camera_rotation_matrix", "l2": 3.000382, "mean_abs": 0.207029, "relative": 0.921369 }, { "name": "imu_accel_gyro", "l2": 17.033268, "mean_abs": 0.974996, "relative": 0.919003 }, { "name": "depth_confidence", "l2": 11.090087, "mean_abs": 0.200227, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.172508, "mean_abs": 0.107124, "relative": 0.936631 }, { "name": "video_fisheye_cam1", "l2": 5.53073, "mean_abs": 0.115914, "relative": 0.913906 }, { "name": "video_fisheye_cam2", "l2": 5.610095, "mean_abs": 0.113849, "relative": 0.895568 }, { "name": "video_fisheye_cam3", "l2": 5.544268, "mean_abs": 0.112994, "relative": 0.928321 }, { "name": "video_stereo_left", "l2": 5.333712, "mean_abs": 0.115864, "relative": 0.866739 }, { "name": "video_stereo_right", "l2": 6.687011, "mean_abs": 0.145876, "relative": 0.942728 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.048716, "mean_abs": 0.078393, "relative": 0.733205 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01139, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 675, "start_frame": 3375, "end_frame": 3394, "center_frame": 3384, "action": "Lift gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.828074, "mean_abs": 0.199796, "relative": 0.82904 }, { "name": "hand_right_joints", "l2": 10.822074, "mean_abs": 0.279818, "relative": 0.853069 }, { "name": "body_joints", "l2": 6.392364, "mean_abs": 0.087695, "relative": 0.863411 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019238, "mean_abs": 0.002222, "relative": 0.045398 }, { "name": "camera_rotation_matrix", "l2": 3.000638, "mean_abs": 0.207264, "relative": 0.921447 }, { "name": "imu_accel_gyro", "l2": 17.067789, "mean_abs": 0.981791, "relative": 0.920866 }, { "name": "depth_confidence", "l2": 11.122359, "mean_abs": 0.200858, "relative": 0.000369 }, { "name": "video_fisheye_cam0", "l2": 5.168608, "mean_abs": 0.107134, "relative": 0.935924 }, { "name": "video_fisheye_cam1", "l2": 5.529158, "mean_abs": 0.116037, "relative": 0.913647 }, { "name": "video_fisheye_cam2", "l2": 5.595027, "mean_abs": 0.11379, "relative": 0.893163 }, { "name": "video_fisheye_cam3", "l2": 5.535932, "mean_abs": 0.112784, "relative": 0.926925 }, { "name": "video_stereo_left", "l2": 5.31614, "mean_abs": 0.115942, "relative": 0.863883 }, { "name": "video_stereo_right", "l2": 6.670081, "mean_abs": 0.145774, "relative": 0.940341 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.1104, "mean_abs": 0.08329, "relative": 0.75528 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01139, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 676, "start_frame": 3380, "end_frame": 3399, "center_frame": 3389, "action": "Lift gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.81284, "mean_abs": 0.200448, "relative": 0.827755 }, { "name": "hand_right_joints", "l2": 10.834079, "mean_abs": 0.281384, "relative": 0.854015 }, { "name": "body_joints", "l2": 6.410258, "mean_abs": 0.087901, "relative": 0.865828 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020367, "mean_abs": 0.002457, "relative": 0.048064 }, { "name": "camera_rotation_matrix", "l2": 3.000589, "mean_abs": 0.20695, "relative": 0.921432 }, { "name": "imu_accel_gyro", "l2": 17.071514, "mean_abs": 0.980689, "relative": 0.921067 }, { "name": "depth_confidence", "l2": 11.143774, "mean_abs": 0.201329, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.16442, "mean_abs": 0.106934, "relative": 0.935166 }, { "name": "video_fisheye_cam1", "l2": 5.531484, "mean_abs": 0.115943, "relative": 0.914031 }, { "name": "video_fisheye_cam2", "l2": 5.582266, "mean_abs": 0.113508, "relative": 0.891126 }, { "name": "video_fisheye_cam3", "l2": 5.535629, "mean_abs": 0.112531, "relative": 0.926874 }, { "name": "video_stereo_left", "l2": 5.301062, "mean_abs": 0.115506, "relative": 0.861433 }, { "name": "video_stereo_right", "l2": 6.655672, "mean_abs": 0.145196, "relative": 0.93831 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.109469, "mean_abs": 0.084344, "relative": 0.754947 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01139, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 677, "start_frame": 3385, "end_frame": 3404, "center_frame": 3394, "action": "Lift gooseneck kettle", "subtask": "Handle gooseneck kettle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.811404, "mean_abs": 0.201101, "relative": 0.827634 }, { "name": "hand_right_joints", "l2": 10.843924, "mean_abs": 0.282624, "relative": 0.854791 }, { "name": "body_joints", "l2": 6.4337, "mean_abs": 0.087992, "relative": 0.868995 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015902, "mean_abs": 0.002132, "relative": 0.037526 }, { "name": "camera_rotation_matrix", "l2": 3.000472, "mean_abs": 0.206171, "relative": 0.921396 }, { "name": "imu_accel_gyro", "l2": 17.056278, "mean_abs": 0.981315, "relative": 0.920245 }, { "name": "depth_confidence", "l2": 11.163875, "mean_abs": 0.201059, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.162185, "mean_abs": 0.106681, "relative": 0.934761 }, { "name": "video_fisheye_cam1", "l2": 5.534351, "mean_abs": 0.115827, "relative": 0.914505 }, { "name": "video_fisheye_cam2", "l2": 5.574841, "mean_abs": 0.11317, "relative": 0.88994 }, { "name": "video_fisheye_cam3", "l2": 5.533369, "mean_abs": 0.112196, "relative": 0.926496 }, { "name": "video_stereo_left", "l2": 5.289292, "mean_abs": 0.114913, "relative": 0.85952 }, { "name": "video_stereo_right", "l2": 6.649656, "mean_abs": 0.144532, "relative": 0.937462 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.134616, "mean_abs": 0.084751, "relative": 0.763947 }, { "name": "caption_objects_interaction_text", "l2": 1.840838, "mean_abs": 0.015391, "relative": 0.845268 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 678, "start_frame": 3390, "end_frame": 3409, "center_frame": 3399, "action": "", "subtask": "", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.792637, "mean_abs": 0.201572, "relative": 0.826051 }, { "name": "hand_right_joints", "l2": 10.850554, "mean_abs": 0.285047, "relative": 0.855314 }, { "name": "body_joints", "l2": 6.466867, "mean_abs": 0.088316, "relative": 0.873475 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009909, "mean_abs": 0.001372, "relative": 0.023384 }, { "name": "camera_rotation_matrix", "l2": 3.00024, "mean_abs": 0.204761, "relative": 0.921325 }, { "name": "imu_accel_gyro", "l2": 17.053501, "mean_abs": 0.983367, "relative": 0.920095 }, { "name": "depth_confidence", "l2": 11.170556, "mean_abs": 0.201131, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.15872, "mean_abs": 0.106346, "relative": 0.934134 }, { "name": "video_fisheye_cam1", "l2": 5.539795, "mean_abs": 0.115388, "relative": 0.915404 }, { "name": "video_fisheye_cam2", "l2": 5.565965, "mean_abs": 0.11255, "relative": 0.888523 }, { "name": "video_fisheye_cam3", "l2": 5.532415, "mean_abs": 0.112114, "relative": 0.926336 }, { "name": "video_stereo_left", "l2": 5.27523, "mean_abs": 0.113757, "relative": 0.857235 }, { "name": "video_stereo_right", "l2": 6.647513, "mean_abs": 0.143399, "relative": 0.93716 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.094882, "mean_abs": 0.084081, "relative": 0.749727 }, { "name": "caption_objects_interaction_text", "l2": 1.840838, "mean_abs": 0.015577, "relative": 0.845268 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 679, "start_frame": 3395, "end_frame": 3414, "center_frame": 3404, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "digital scale", "white mug", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.783391, "mean_abs": 0.201414, "relative": 0.825271 }, { "name": "hand_right_joints", "l2": 10.857458, "mean_abs": 0.287093, "relative": 0.855858 }, { "name": "body_joints", "l2": 6.496423, "mean_abs": 0.088497, "relative": 0.877467 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003425, "mean_abs": 0.000547, "relative": 0.008082 }, { "name": "camera_rotation_matrix", "l2": 3.000189, "mean_abs": 0.204112, "relative": 0.921309 }, { "name": "imu_accel_gyro", "l2": 17.043951, "mean_abs": 0.980354, "relative": 0.91958 }, { "name": "depth_confidence", "l2": 11.184196, "mean_abs": 0.201113, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.159473, "mean_abs": 0.10615, "relative": 0.93427 }, { "name": "video_fisheye_cam1", "l2": 5.540293, "mean_abs": 0.115094, "relative": 0.915487 }, { "name": "video_fisheye_cam2", "l2": 5.560656, "mean_abs": 0.112224, "relative": 0.887676 }, { "name": "video_fisheye_cam3", "l2": 5.535397, "mean_abs": 0.111815, "relative": 0.926835 }, { "name": "video_stereo_left", "l2": 5.268158, "mean_abs": 0.113021, "relative": 0.856086 }, { "name": "video_stereo_right", "l2": 6.645312, "mean_abs": 0.142687, "relative": 0.93685 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.071242, "mean_abs": 0.083079, "relative": 0.741266 }, { "name": "caption_objects_interaction_text", "l2": 1.840838, "mean_abs": 0.015463, "relative": 0.845268 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 680, "start_frame": 3400, "end_frame": 3419, "center_frame": 3409, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.783979, "mean_abs": 0.202026, "relative": 0.82532 }, { "name": "hand_right_joints", "l2": 10.880851, "mean_abs": 0.287611, "relative": 0.857702 }, { "name": "body_joints", "l2": 6.523566, "mean_abs": 0.088713, "relative": 0.881133 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00821, "mean_abs": 0.001184, "relative": 0.019375 }, { "name": "camera_rotation_matrix", "l2": 3.000322, "mean_abs": 0.205203, "relative": 0.92135 }, { "name": "imu_accel_gyro", "l2": 17.06428, "mean_abs": 0.981634, "relative": 0.920677 }, { "name": "depth_confidence", "l2": 11.190773, "mean_abs": 0.200692, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.167021, "mean_abs": 0.106552, "relative": 0.935637 }, { "name": "video_fisheye_cam1", "l2": 5.548128, "mean_abs": 0.115395, "relative": 0.916781 }, { "name": "video_fisheye_cam2", "l2": 5.564328, "mean_abs": 0.112591, "relative": 0.888262 }, { "name": "video_fisheye_cam3", "l2": 5.562532, "mean_abs": 0.112465, "relative": 0.931379 }, { "name": "video_stereo_left", "l2": 5.279198, "mean_abs": 0.113952, "relative": 0.85788 }, { "name": "video_stereo_right", "l2": 6.663469, "mean_abs": 0.143632, "relative": 0.939409 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.093505, "mean_abs": 0.083502, "relative": 0.749234 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 681, "start_frame": 3405, "end_frame": 3424, "center_frame": 3414, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.781328, "mean_abs": 0.202049, "relative": 0.825097 }, { "name": "hand_right_joints", "l2": 10.891458, "mean_abs": 0.287713, "relative": 0.858538 }, { "name": "body_joints", "l2": 6.543168, "mean_abs": 0.089152, "relative": 0.88378 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012517, "mean_abs": 0.001892, "relative": 0.029539 }, { "name": "camera_rotation_matrix", "l2": 3.000391, "mean_abs": 0.206201, "relative": 0.921371 }, { "name": "imu_accel_gyro", "l2": 17.080627, "mean_abs": 0.987318, "relative": 0.921559 }, { "name": "depth_confidence", "l2": 11.19398, "mean_abs": 0.201247, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.167458, "mean_abs": 0.10676, "relative": 0.935716 }, { "name": "video_fisheye_cam1", "l2": 5.547014, "mean_abs": 0.115777, "relative": 0.916597 }, { "name": "video_fisheye_cam2", "l2": 5.569331, "mean_abs": 0.112888, "relative": 0.889061 }, { "name": "video_fisheye_cam3", "l2": 5.574585, "mean_abs": 0.112946, "relative": 0.933397 }, { "name": "video_stereo_left", "l2": 5.295119, "mean_abs": 0.114729, "relative": 0.860467 }, { "name": "video_stereo_right", "l2": 6.6713, "mean_abs": 0.14457, "relative": 0.940513 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.082937, "mean_abs": 0.082848, "relative": 0.745452 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 682, "start_frame": 3410, "end_frame": 3429, "center_frame": 3419, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.774035, "mean_abs": 0.201912, "relative": 0.824482 }, { "name": "hand_right_joints", "l2": 10.904467, "mean_abs": 0.28748, "relative": 0.859564 }, { "name": "body_joints", "l2": 6.561467, "mean_abs": 0.089431, "relative": 0.886252 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013711, "mean_abs": 0.001992, "relative": 0.032356 }, { "name": "camera_rotation_matrix", "l2": 3.000203, "mean_abs": 0.20622, "relative": 0.921314 }, { "name": "imu_accel_gyro", "l2": 17.075462, "mean_abs": 0.982842, "relative": 0.92128 }, { "name": "depth_confidence", "l2": 11.168252, "mean_abs": 0.200748, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.174285, "mean_abs": 0.106747, "relative": 0.936952 }, { "name": "video_fisheye_cam1", "l2": 5.5511, "mean_abs": 0.115845, "relative": 0.917272 }, { "name": "video_fisheye_cam2", "l2": 5.574954, "mean_abs": 0.113033, "relative": 0.889959 }, { "name": "video_fisheye_cam3", "l2": 5.608462, "mean_abs": 0.113159, "relative": 0.939069 }, { "name": "video_stereo_left", "l2": 5.309957, "mean_abs": 0.115003, "relative": 0.862879 }, { "name": "video_stereo_right", "l2": 6.686805, "mean_abs": 0.144926, "relative": 0.942699 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.273402, "mean_abs": 0.090644, "relative": 0.813616 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 683, "start_frame": 3415, "end_frame": 3434, "center_frame": 3424, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.772682, "mean_abs": 0.20158, "relative": 0.824367 }, { "name": "hand_right_joints", "l2": 10.898033, "mean_abs": 0.286456, "relative": 0.859057 }, { "name": "body_joints", "l2": 6.56278, "mean_abs": 0.089629, "relative": 0.886429 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015794, "mean_abs": 0.002061, "relative": 0.037273 }, { "name": "camera_rotation_matrix", "l2": 3.000156, "mean_abs": 0.206176, "relative": 0.921299 }, { "name": "imu_accel_gyro", "l2": 17.082905, "mean_abs": 0.9825, "relative": 0.921682 }, { "name": "depth_confidence", "l2": 11.149235, "mean_abs": 0.199872, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.171484, "mean_abs": 0.106859, "relative": 0.936445 }, { "name": "video_fisheye_cam1", "l2": 5.545551, "mean_abs": 0.115805, "relative": 0.916355 }, { "name": "video_fisheye_cam2", "l2": 5.580161, "mean_abs": 0.113051, "relative": 0.89079 }, { "name": "video_fisheye_cam3", "l2": 5.601367, "mean_abs": 0.113214, "relative": 0.937881 }, { "name": "video_stereo_left", "l2": 5.320807, "mean_abs": 0.115059, "relative": 0.864642 }, { "name": "video_stereo_right", "l2": 6.692446, "mean_abs": 0.145067, "relative": 0.943494 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.14394, "mean_abs": 0.082832, "relative": 0.767284 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 684, "start_frame": 3420, "end_frame": 3439, "center_frame": 3429, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.77243, "mean_abs": 0.201887, "relative": 0.824346 }, { "name": "hand_right_joints", "l2": 10.89364, "mean_abs": 0.287284, "relative": 0.85871 }, { "name": "body_joints", "l2": 6.555517, "mean_abs": 0.08967, "relative": 0.885448 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013965, "mean_abs": 0.001691, "relative": 0.032955 }, { "name": "camera_rotation_matrix", "l2": 3.000118, "mean_abs": 0.205839, "relative": 0.921288 }, { "name": "imu_accel_gyro", "l2": 17.029846, "mean_abs": 0.988741, "relative": 0.918819 }, { "name": "depth_confidence", "l2": 11.146868, "mean_abs": 0.199601, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.171652, "mean_abs": 0.106782, "relative": 0.936476 }, { "name": "video_fisheye_cam1", "l2": 5.543055, "mean_abs": 0.115536, "relative": 0.915943 }, { "name": "video_fisheye_cam2", "l2": 5.583109, "mean_abs": 0.112963, "relative": 0.89126 }, { "name": "video_fisheye_cam3", "l2": 5.60781, "mean_abs": 0.113157, "relative": 0.93896 }, { "name": "video_stereo_left", "l2": 5.332924, "mean_abs": 0.114937, "relative": 0.866611 }, { "name": "video_stereo_right", "l2": 6.692539, "mean_abs": 0.144523, "relative": 0.943508 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.061543, "mean_abs": 0.078359, "relative": 0.737795 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 685, "start_frame": 3425, "end_frame": 3444, "center_frame": 3434, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.783439, "mean_abs": 0.202173, "relative": 0.825275 }, { "name": "hand_right_joints", "l2": 10.922885, "mean_abs": 0.288435, "relative": 0.861016 }, { "name": "body_joints", "l2": 6.538794, "mean_abs": 0.089285, "relative": 0.88319 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012466, "mean_abs": 0.001675, "relative": 0.029417 }, { "name": "camera_rotation_matrix", "l2": 3.00065, "mean_abs": 0.205841, "relative": 0.921451 }, { "name": "imu_accel_gyro", "l2": 17.032213, "mean_abs": 0.990959, "relative": 0.918947 }, { "name": "depth_confidence", "l2": 11.143939, "mean_abs": 0.198955, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.172414, "mean_abs": 0.106689, "relative": 0.936614 }, { "name": "video_fisheye_cam1", "l2": 5.546143, "mean_abs": 0.115311, "relative": 0.916453 }, { "name": "video_fisheye_cam2", "l2": 5.585381, "mean_abs": 0.112824, "relative": 0.891623 }, { "name": "video_fisheye_cam3", "l2": 5.618217, "mean_abs": 0.113258, "relative": 0.940703 }, { "name": "video_stereo_left", "l2": 5.338706, "mean_abs": 0.114487, "relative": 0.86755 }, { "name": "video_stereo_right", "l2": 6.699014, "mean_abs": 0.144227, "relative": 0.94442 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.190882, "mean_abs": 0.084516, "relative": 0.784084 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 686, "start_frame": 3430, "end_frame": 3449, "center_frame": 3439, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.789341, "mean_abs": 0.200899, "relative": 0.825773 }, { "name": "hand_right_joints", "l2": 10.919186, "mean_abs": 0.287225, "relative": 0.860724 }, { "name": "body_joints", "l2": 6.530237, "mean_abs": 0.089066, "relative": 0.882034 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008705, "mean_abs": 0.001329, "relative": 0.020542 }, { "name": "camera_rotation_matrix", "l2": 3.000499, "mean_abs": 0.20571, "relative": 0.921404 }, { "name": "imu_accel_gyro", "l2": 17.027067, "mean_abs": 0.987424, "relative": 0.918669 }, { "name": "depth_confidence", "l2": 11.1558, "mean_abs": 0.199065, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.172065, "mean_abs": 0.106746, "relative": 0.93655 }, { "name": "video_fisheye_cam1", "l2": 5.548604, "mean_abs": 0.115479, "relative": 0.91686 }, { "name": "video_fisheye_cam2", "l2": 5.586379, "mean_abs": 0.112828, "relative": 0.891782 }, { "name": "video_fisheye_cam3", "l2": 5.620769, "mean_abs": 0.113605, "relative": 0.94113 }, { "name": "video_stereo_left", "l2": 5.341591, "mean_abs": 0.114408, "relative": 0.868019 }, { "name": "video_stereo_right", "l2": 6.701842, "mean_abs": 0.144418, "relative": 0.944819 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.07554, "mean_abs": 0.07813, "relative": 0.742805 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 687, "start_frame": 3435, "end_frame": 3454, "center_frame": 3444, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.792921, "mean_abs": 0.20078, "relative": 0.826075 }, { "name": "hand_right_joints", "l2": 10.939078, "mean_abs": 0.287004, "relative": 0.862292 }, { "name": "body_joints", "l2": 6.523863, "mean_abs": 0.088908, "relative": 0.881173 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007312, "mean_abs": 0.001135, "relative": 0.017254 }, { "name": "camera_rotation_matrix", "l2": 3.000025, "mean_abs": 0.205785, "relative": 0.921259 }, { "name": "imu_accel_gyro", "l2": 17.022614, "mean_abs": 0.984983, "relative": 0.918429 }, { "name": "depth_confidence", "l2": 11.161098, "mean_abs": 0.199327, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.175038, "mean_abs": 0.106649, "relative": 0.937089 }, { "name": "video_fisheye_cam1", "l2": 5.556722, "mean_abs": 0.115468, "relative": 0.918201 }, { "name": "video_fisheye_cam2", "l2": 5.586753, "mean_abs": 0.11281, "relative": 0.891842 }, { "name": "video_fisheye_cam3", "l2": 5.629772, "mean_abs": 0.113318, "relative": 0.942637 }, { "name": "video_stereo_left", "l2": 5.34409, "mean_abs": 0.114602, "relative": 0.868425 }, { "name": "video_stereo_right", "l2": 6.707183, "mean_abs": 0.144552, "relative": 0.945572 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.035303, "mean_abs": 0.076742, "relative": 0.728405 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 688, "start_frame": 3440, "end_frame": 3459, "center_frame": 3449, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.803744, "mean_abs": 0.200458, "relative": 0.826988 }, { "name": "hand_right_joints", "l2": 10.971761, "mean_abs": 0.286548, "relative": 0.864868 }, { "name": "body_joints", "l2": 6.522922, "mean_abs": 0.088402, "relative": 0.881046 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012609, "mean_abs": 0.00198, "relative": 0.029755 }, { "name": "camera_rotation_matrix", "l2": 3.000199, "mean_abs": 0.20598, "relative": 0.921313 }, { "name": "imu_accel_gyro", "l2": 17.062313, "mean_abs": 0.986378, "relative": 0.920571 }, { "name": "depth_confidence", "l2": 11.174833, "mean_abs": 0.199394, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.173953, "mean_abs": 0.106704, "relative": 0.936892 }, { "name": "video_fisheye_cam1", "l2": 5.56139, "mean_abs": 0.115537, "relative": 0.918973 }, { "name": "video_fisheye_cam2", "l2": 5.584725, "mean_abs": 0.112764, "relative": 0.891518 }, { "name": "video_fisheye_cam3", "l2": 5.628421, "mean_abs": 0.113485, "relative": 0.942411 }, { "name": "video_stereo_left", "l2": 5.341805, "mean_abs": 0.114473, "relative": 0.868054 }, { "name": "video_stereo_right", "l2": 6.708846, "mean_abs": 0.144404, "relative": 0.945807 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.068352, "mean_abs": 0.079004, "relative": 0.740232 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 689, "start_frame": 3445, "end_frame": 3464, "center_frame": 3454, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.807397, "mean_abs": 0.200371, "relative": 0.827296 }, { "name": "hand_right_joints", "l2": 10.988292, "mean_abs": 0.28702, "relative": 0.866171 }, { "name": "body_joints", "l2": 6.53838, "mean_abs": 0.088611, "relative": 0.883134 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016466, "mean_abs": 0.002464, "relative": 0.038857 }, { "name": "camera_rotation_matrix", "l2": 3.000351, "mean_abs": 0.206014, "relative": 0.921359 }, { "name": "imu_accel_gyro", "l2": 17.047491, "mean_abs": 0.984188, "relative": 0.919771 }, { "name": "depth_confidence", "l2": 11.191582, "mean_abs": 0.199307, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.169323, "mean_abs": 0.106551, "relative": 0.936054 }, { "name": "video_fisheye_cam1", "l2": 5.561261, "mean_abs": 0.11544, "relative": 0.918951 }, { "name": "video_fisheye_cam2", "l2": 5.583913, "mean_abs": 0.112756, "relative": 0.891389 }, { "name": "video_fisheye_cam3", "l2": 5.616213, "mean_abs": 0.113349, "relative": 0.940367 }, { "name": "video_stereo_left", "l2": 5.333801, "mean_abs": 0.114094, "relative": 0.866753 }, { "name": "video_stereo_right", "l2": 6.704225, "mean_abs": 0.14388, "relative": 0.945155 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.047677, "mean_abs": 0.080687, "relative": 0.732833 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 690, "start_frame": 3450, "end_frame": 3469, "center_frame": 3459, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.812188, "mean_abs": 0.201052, "relative": 0.8277 }, { "name": "hand_right_joints", "l2": 11.000694, "mean_abs": 0.288084, "relative": 0.867149 }, { "name": "body_joints", "l2": 6.547135, "mean_abs": 0.088999, "relative": 0.884316 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014887, "mean_abs": 0.002064, "relative": 0.035132 }, { "name": "camera_rotation_matrix", "l2": 3.000147, "mean_abs": 0.205974, "relative": 0.921296 }, { "name": "imu_accel_gyro", "l2": 17.070438, "mean_abs": 0.983923, "relative": 0.921009 }, { "name": "depth_confidence", "l2": 11.225869, "mean_abs": 0.199572, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.162374, "mean_abs": 0.106416, "relative": 0.934796 }, { "name": "video_fisheye_cam1", "l2": 5.564976, "mean_abs": 0.11553, "relative": 0.919565 }, { "name": "video_fisheye_cam2", "l2": 5.582752, "mean_abs": 0.112813, "relative": 0.891203 }, { "name": "video_fisheye_cam3", "l2": 5.593899, "mean_abs": 0.112933, "relative": 0.936631 }, { "name": "video_stereo_left", "l2": 5.33149, "mean_abs": 0.114041, "relative": 0.866378 }, { "name": "video_stereo_right", "l2": 6.714726, "mean_abs": 0.14421, "relative": 0.946635 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.124152, "mean_abs": 0.080603, "relative": 0.760202 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 691, "start_frame": 3455, "end_frame": 3474, "center_frame": 3464, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.823299, "mean_abs": 0.201763, "relative": 0.828637 }, { "name": "hand_right_joints", "l2": 11.006083, "mean_abs": 0.288039, "relative": 0.867574 }, { "name": "body_joints", "l2": 6.557521, "mean_abs": 0.089265, "relative": 0.885719 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013307, "mean_abs": 0.00184, "relative": 0.031402 }, { "name": "camera_rotation_matrix", "l2": 2.99978, "mean_abs": 0.205835, "relative": 0.921184 }, { "name": "imu_accel_gyro", "l2": 17.06687, "mean_abs": 0.985754, "relative": 0.920816 }, { "name": "depth_confidence", "l2": 11.242124, "mean_abs": 0.20021, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.155794, "mean_abs": 0.106358, "relative": 0.933604 }, { "name": "video_fisheye_cam1", "l2": 5.563785, "mean_abs": 0.115556, "relative": 0.919369 }, { "name": "video_fisheye_cam2", "l2": 5.578471, "mean_abs": 0.112814, "relative": 0.89052 }, { "name": "video_fisheye_cam3", "l2": 5.580484, "mean_abs": 0.112851, "relative": 0.934385 }, { "name": "video_stereo_left", "l2": 5.329124, "mean_abs": 0.113993, "relative": 0.865993 }, { "name": "video_stereo_right", "l2": 6.714358, "mean_abs": 0.14404, "relative": 0.946584 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.101694, "mean_abs": 0.081743, "relative": 0.752165 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 692, "start_frame": 3460, "end_frame": 3479, "center_frame": 3469, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.827687, "mean_abs": 0.202812, "relative": 0.829007 }, { "name": "hand_right_joints", "l2": 11.024171, "mean_abs": 0.289206, "relative": 0.869 }, { "name": "body_joints", "l2": 6.573583, "mean_abs": 0.089806, "relative": 0.887888 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007704, "mean_abs": 0.001089, "relative": 0.018181 }, { "name": "camera_rotation_matrix", "l2": 2.999846, "mean_abs": 0.205838, "relative": 0.921204 }, { "name": "imu_accel_gyro", "l2": 17.064865, "mean_abs": 0.982504, "relative": 0.920708 }, { "name": "depth_confidence", "l2": 11.24754, "mean_abs": 0.200225, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.14849, "mean_abs": 0.10623, "relative": 0.932281 }, { "name": "video_fisheye_cam1", "l2": 5.563648, "mean_abs": 0.115605, "relative": 0.919346 }, { "name": "video_fisheye_cam2", "l2": 5.576742, "mean_abs": 0.112807, "relative": 0.890244 }, { "name": "video_fisheye_cam3", "l2": 5.562748, "mean_abs": 0.112516, "relative": 0.931415 }, { "name": "video_stereo_left", "l2": 5.324522, "mean_abs": 0.113943, "relative": 0.865245 }, { "name": "video_stereo_right", "l2": 6.710156, "mean_abs": 0.144062, "relative": 0.945991 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.122749, "mean_abs": 0.084603, "relative": 0.7597 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 693, "start_frame": 3465, "end_frame": 3484, "center_frame": 3474, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.828828, "mean_abs": 0.202901, "relative": 0.829104 }, { "name": "hand_right_joints", "l2": 11.023143, "mean_abs": 0.28917, "relative": 0.868919 }, { "name": "body_joints", "l2": 6.583914, "mean_abs": 0.090067, "relative": 0.889284 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004115, "mean_abs": 0.000621, "relative": 0.009712 }, { "name": "camera_rotation_matrix", "l2": 3.000096, "mean_abs": 0.206155, "relative": 0.921281 }, { "name": "imu_accel_gyro", "l2": 17.048578, "mean_abs": 0.980694, "relative": 0.919829 }, { "name": "depth_confidence", "l2": 11.268714, "mean_abs": 0.200477, "relative": 0.000374 }, { "name": "video_fisheye_cam0", "l2": 5.14361, "mean_abs": 0.10629, "relative": 0.931398 }, { "name": "video_fisheye_cam1", "l2": 5.570669, "mean_abs": 0.116136, "relative": 0.920506 }, { "name": "video_fisheye_cam2", "l2": 5.574395, "mean_abs": 0.112955, "relative": 0.889869 }, { "name": "video_fisheye_cam3", "l2": 5.558735, "mean_abs": 0.112759, "relative": 0.930743 }, { "name": "video_stereo_left", "l2": 5.319178, "mean_abs": 0.114353, "relative": 0.864377 }, { "name": "video_stereo_right", "l2": 6.713856, "mean_abs": 0.144737, "relative": 0.946513 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.172406, "mean_abs": 0.08752, "relative": 0.777471 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 694, "start_frame": 3470, "end_frame": 3489, "center_frame": 3479, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.821974, "mean_abs": 0.204001, "relative": 0.828525 }, { "name": "hand_right_joints", "l2": 11.029341, "mean_abs": 0.289062, "relative": 0.869407 }, { "name": "body_joints", "l2": 6.591491, "mean_abs": 0.090477, "relative": 0.890307 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005963, "mean_abs": 0.000792, "relative": 0.014072 }, { "name": "camera_rotation_matrix", "l2": 2.999946, "mean_abs": 0.206083, "relative": 0.921235 }, { "name": "imu_accel_gyro", "l2": 17.059814, "mean_abs": 0.986411, "relative": 0.920436 }, { "name": "depth_confidence", "l2": 11.272629, "mean_abs": 0.199822, "relative": 0.000374 }, { "name": "video_fisheye_cam0", "l2": 5.13752, "mean_abs": 0.105998, "relative": 0.930295 }, { "name": "video_fisheye_cam1", "l2": 5.575107, "mean_abs": 0.116305, "relative": 0.921239 }, { "name": "video_fisheye_cam2", "l2": 5.571765, "mean_abs": 0.112863, "relative": 0.889449 }, { "name": "video_fisheye_cam3", "l2": 5.551685, "mean_abs": 0.112333, "relative": 0.929563 }, { "name": "video_stereo_left", "l2": 5.313992, "mean_abs": 0.114439, "relative": 0.863534 }, { "name": "video_stereo_right", "l2": 6.713218, "mean_abs": 0.14468, "relative": 0.946423 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.190351, "mean_abs": 0.087448, "relative": 0.783894 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 695, "start_frame": 3475, "end_frame": 3494, "center_frame": 3484, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.814569, "mean_abs": 0.203819, "relative": 0.827901 }, { "name": "hand_right_joints", "l2": 11.034707, "mean_abs": 0.289539, "relative": 0.86983 }, { "name": "body_joints", "l2": 6.602211, "mean_abs": 0.090757, "relative": 0.891755 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007526, "mean_abs": 0.001045, "relative": 0.01776 }, { "name": "camera_rotation_matrix", "l2": 2.999961, "mean_abs": 0.206009, "relative": 0.921239 }, { "name": "imu_accel_gyro", "l2": 17.05978, "mean_abs": 0.986444, "relative": 0.920434 }, { "name": "depth_confidence", "l2": 11.263003, "mean_abs": 0.199423, "relative": 0.000374 }, { "name": "video_fisheye_cam0", "l2": 5.140394, "mean_abs": 0.106272, "relative": 0.930815 }, { "name": "video_fisheye_cam1", "l2": 5.581516, "mean_abs": 0.11663, "relative": 0.922298 }, { "name": "video_fisheye_cam2", "l2": 5.575125, "mean_abs": 0.113139, "relative": 0.889986 }, { "name": "video_fisheye_cam3", "l2": 5.553386, "mean_abs": 0.113057, "relative": 0.929847 }, { "name": "video_stereo_left", "l2": 5.31658, "mean_abs": 0.11469, "relative": 0.863955 }, { "name": "video_stereo_right", "l2": 6.717138, "mean_abs": 0.144897, "relative": 0.946975 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.228284, "mean_abs": 0.09037, "relative": 0.797469 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 696, "start_frame": 3480, "end_frame": 3499, "center_frame": 3489, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.807247, "mean_abs": 0.203379, "relative": 0.827283 }, { "name": "hand_right_joints", "l2": 11.049137, "mean_abs": 0.290559, "relative": 0.870968 }, { "name": "body_joints", "l2": 6.602692, "mean_abs": 0.090728, "relative": 0.89182 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007288, "mean_abs": 0.001086, "relative": 0.0172 }, { "name": "camera_rotation_matrix", "l2": 3.000193, "mean_abs": 0.206006, "relative": 0.921311 }, { "name": "imu_accel_gyro", "l2": 17.042166, "mean_abs": 0.986175, "relative": 0.919484 }, { "name": "depth_confidence", "l2": 11.252183, "mean_abs": 0.19958, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.137918, "mean_abs": 0.106075, "relative": 0.930367 }, { "name": "video_fisheye_cam1", "l2": 5.553218, "mean_abs": 0.116699, "relative": 0.917622 }, { "name": "video_fisheye_cam2", "l2": 5.573969, "mean_abs": 0.113154, "relative": 0.889801 }, { "name": "video_fisheye_cam3", "l2": 5.544327, "mean_abs": 0.112702, "relative": 0.928331 }, { "name": "video_stereo_left", "l2": 5.311616, "mean_abs": 0.114652, "relative": 0.863148 }, { "name": "video_stereo_right", "l2": 6.664252, "mean_abs": 0.144728, "relative": 0.93952 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.27806, "mean_abs": 0.091888, "relative": 0.815284 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 697, "start_frame": 3485, "end_frame": 3504, "center_frame": 3494, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.809621, "mean_abs": 0.203613, "relative": 0.827483 }, { "name": "hand_right_joints", "l2": 11.058321, "mean_abs": 0.292294, "relative": 0.871692 }, { "name": "body_joints", "l2": 6.548098, "mean_abs": 0.091147, "relative": 0.884446 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008012, "mean_abs": 0.001064, "relative": 0.018908 }, { "name": "camera_rotation_matrix", "l2": 3.000075, "mean_abs": 0.206152, "relative": 0.921274 }, { "name": "imu_accel_gyro", "l2": 17.045942, "mean_abs": 0.991027, "relative": 0.919687 }, { "name": "depth_confidence", "l2": 11.220932, "mean_abs": 0.200026, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.135011, "mean_abs": 0.106054, "relative": 0.929841 }, { "name": "video_fisheye_cam1", "l2": 5.540721, "mean_abs": 0.116601, "relative": 0.915557 }, { "name": "video_fisheye_cam2", "l2": 5.572761, "mean_abs": 0.113358, "relative": 0.889608 }, { "name": "video_fisheye_cam3", "l2": 5.53424, "mean_abs": 0.112581, "relative": 0.926642 }, { "name": "video_stereo_left", "l2": 5.310799, "mean_abs": 0.115075, "relative": 0.863015 }, { "name": "video_stereo_right", "l2": 6.626451, "mean_abs": 0.144829, "relative": 0.934191 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.20606, "mean_abs": 0.088285, "relative": 0.789516 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 698, "start_frame": 3490, "end_frame": 3509, "center_frame": 3499, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.815164, "mean_abs": 0.203363, "relative": 0.827951 }, { "name": "hand_right_joints", "l2": 11.043054, "mean_abs": 0.292801, "relative": 0.870488 }, { "name": "body_joints", "l2": 6.46485, "mean_abs": 0.091973, "relative": 0.873202 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021521, "mean_abs": 0.002963, "relative": 0.050787 }, { "name": "camera_rotation_matrix", "l2": 3.00074, "mean_abs": 0.207916, "relative": 0.921479 }, { "name": "imu_accel_gyro", "l2": 17.033991, "mean_abs": 1.007591, "relative": 0.919042 }, { "name": "depth_confidence", "l2": 11.130414, "mean_abs": 0.201166, "relative": 0.000369 }, { "name": "video_fisheye_cam0", "l2": 5.137795, "mean_abs": 0.106693, "relative": 0.930345 }, { "name": "video_fisheye_cam1", "l2": 5.52187, "mean_abs": 0.116962, "relative": 0.912442 }, { "name": "video_fisheye_cam2", "l2": 5.565613, "mean_abs": 0.114091, "relative": 0.888467 }, { "name": "video_fisheye_cam3", "l2": 5.524346, "mean_abs": 0.113314, "relative": 0.924985 }, { "name": "video_stereo_left", "l2": 5.305381, "mean_abs": 0.117155, "relative": 0.862135 }, { "name": "video_stereo_right", "l2": 6.572094, "mean_abs": 0.146623, "relative": 0.926527 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.033753, "mean_abs": 0.079046, "relative": 0.72785 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 699, "start_frame": 3495, "end_frame": 3514, "center_frame": 3504, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.826247, "mean_abs": 0.203676, "relative": 0.828886 }, { "name": "hand_right_joints", "l2": 10.981949, "mean_abs": 0.290869, "relative": 0.865671 }, { "name": "body_joints", "l2": 6.366473, "mean_abs": 0.092697, "relative": 0.859914 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.046538, "mean_abs": 0.006201, "relative": 0.109823 }, { "name": "camera_rotation_matrix", "l2": 3.00293, "mean_abs": 0.209952, "relative": 0.922151 }, { "name": "imu_accel_gyro", "l2": 17.065866, "mean_abs": 1.019171, "relative": 0.920762 }, { "name": "depth_confidence", "l2": 11.075548, "mean_abs": 0.201884, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.141284, "mean_abs": 0.107396, "relative": 0.930977 }, { "name": "video_fisheye_cam1", "l2": 5.486444, "mean_abs": 0.117383, "relative": 0.906588 }, { "name": "video_fisheye_cam2", "l2": 5.551041, "mean_abs": 0.114995, "relative": 0.886141 }, { "name": "video_fisheye_cam3", "l2": 5.514952, "mean_abs": 0.114047, "relative": 0.923412 }, { "name": "video_stereo_left", "l2": 5.310318, "mean_abs": 0.119535, "relative": 0.862937 }, { "name": "video_stereo_right", "l2": 6.475953, "mean_abs": 0.147331, "relative": 0.912973 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.108133, "mean_abs": 0.080425, "relative": 0.754469 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 700, "start_frame": 3500, "end_frame": 3519, "center_frame": 3509, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.836484, "mean_abs": 0.203674, "relative": 0.829749 }, { "name": "hand_right_joints", "l2": 10.975532, "mean_abs": 0.287785, "relative": 0.865166 }, { "name": "body_joints", "l2": 6.215194, "mean_abs": 0.092093, "relative": 0.839481 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.077263, "mean_abs": 0.009779, "relative": 0.182329 }, { "name": "camera_rotation_matrix", "l2": 3.007141, "mean_abs": 0.211317, "relative": 0.923444 }, { "name": "imu_accel_gyro", "l2": 17.073359, "mean_abs": 1.029418, "relative": 0.921166 }, { "name": "depth_confidence", "l2": 11.001656, "mean_abs": 0.203235, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.144814, "mean_abs": 0.107951, "relative": 0.931616 }, { "name": "video_fisheye_cam1", "l2": 5.457901, "mean_abs": 0.118341, "relative": 0.901872 }, { "name": "video_fisheye_cam2", "l2": 5.530491, "mean_abs": 0.116046, "relative": 0.882861 }, { "name": "video_fisheye_cam3", "l2": 5.485538, "mean_abs": 0.114735, "relative": 0.918487 }, { "name": "video_stereo_left", "l2": 5.286496, "mean_abs": 0.12073, "relative": 0.859066 }, { "name": "video_stereo_right", "l2": 6.416508, "mean_abs": 0.148533, "relative": 0.904593 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.133935, "mean_abs": 0.082466, "relative": 0.763703 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 701, "start_frame": 3505, "end_frame": 3524, "center_frame": 3514, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.857885, "mean_abs": 0.202989, "relative": 0.831555 }, { "name": "hand_right_joints", "l2": 11.032827, "mean_abs": 0.284096, "relative": 0.869682 }, { "name": "body_joints", "l2": 6.038343, "mean_abs": 0.090027, "relative": 0.815594 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.09889, "mean_abs": 0.012007, "relative": 0.233366 }, { "name": "camera_rotation_matrix", "l2": 3.007729, "mean_abs": 0.209763, "relative": 0.923625 }, { "name": "imu_accel_gyro", "l2": 17.066299, "mean_abs": 1.031103, "relative": 0.920786 }, { "name": "depth_confidence", "l2": 10.955533, "mean_abs": 0.203305, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.131458, "mean_abs": 0.107493, "relative": 0.929197 }, { "name": "video_fisheye_cam1", "l2": 5.413602, "mean_abs": 0.117549, "relative": 0.894552 }, { "name": "video_fisheye_cam2", "l2": 5.503974, "mean_abs": 0.115877, "relative": 0.878628 }, { "name": "video_fisheye_cam3", "l2": 5.428934, "mean_abs": 0.112608, "relative": 0.90901 }, { "name": "video_stereo_left", "l2": 5.248232, "mean_abs": 0.119785, "relative": 0.852848 }, { "name": "video_stereo_right", "l2": 6.282777, "mean_abs": 0.145456, "relative": 0.88574 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.192176, "mean_abs": 0.084371, "relative": 0.784547 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 702, "start_frame": 3510, "end_frame": 3529, "center_frame": 3519, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.913565, "mean_abs": 0.20847, "relative": 0.836251 }, { "name": "hand_right_joints", "l2": 11.180534, "mean_abs": 0.288286, "relative": 0.881325 }, { "name": "body_joints", "l2": 5.883197, "mean_abs": 0.088584, "relative": 0.794639 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.103959, "mean_abs": 0.012224, "relative": 0.245328 }, { "name": "camera_rotation_matrix", "l2": 3.004782, "mean_abs": 0.206257, "relative": 0.92272 }, { "name": "imu_accel_gyro", "l2": 17.051331, "mean_abs": 1.030985, "relative": 0.919978 }, { "name": "depth_confidence", "l2": 10.906112, "mean_abs": 0.200734, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.123476, "mean_abs": 0.106848, "relative": 0.927752 }, { "name": "video_fisheye_cam1", "l2": 5.371353, "mean_abs": 0.115803, "relative": 0.887571 }, { "name": "video_fisheye_cam2", "l2": 5.479977, "mean_abs": 0.114358, "relative": 0.874797 }, { "name": "video_fisheye_cam3", "l2": 5.386453, "mean_abs": 0.111333, "relative": 0.901897 }, { "name": "video_stereo_left", "l2": 5.188178, "mean_abs": 0.116821, "relative": 0.843089 }, { "name": "video_stereo_right", "l2": 6.143345, "mean_abs": 0.140818, "relative": 0.866083 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.117237, "mean_abs": 0.083907, "relative": 0.757727 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 703, "start_frame": 3515, "end_frame": 3534, "center_frame": 3524, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.949187, "mean_abs": 0.214823, "relative": 0.839256 }, { "name": "hand_right_joints", "l2": 11.456828, "mean_abs": 0.296228, "relative": 0.903105 }, { "name": "body_joints", "l2": 5.731036, "mean_abs": 0.086668, "relative": 0.774086 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.098501, "mean_abs": 0.011654, "relative": 0.232448 }, { "name": "camera_rotation_matrix", "l2": 3.003984, "mean_abs": 0.202752, "relative": 0.922475 }, { "name": "imu_accel_gyro", "l2": 17.111002, "mean_abs": 1.054168, "relative": 0.923197 }, { "name": "depth_confidence", "l2": 10.817523, "mean_abs": 0.199307, "relative": 0.000359 }, { "name": "video_fisheye_cam0", "l2": 5.106818, "mean_abs": 0.106189, "relative": 0.924735 }, { "name": "video_fisheye_cam1", "l2": 5.320819, "mean_abs": 0.113178, "relative": 0.87922 }, { "name": "video_fisheye_cam2", "l2": 5.46794, "mean_abs": 0.112521, "relative": 0.872875 }, { "name": "video_fisheye_cam3", "l2": 5.305941, "mean_abs": 0.109288, "relative": 0.888416 }, { "name": "video_stereo_left", "l2": 5.160637, "mean_abs": 0.114276, "relative": 0.838614 }, { "name": "video_stereo_right", "l2": 6.053789, "mean_abs": 0.13672, "relative": 0.853457 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.96814, "mean_abs": 0.07645, "relative": 0.704368 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 704, "start_frame": 3520, "end_frame": 3539, "center_frame": 3529, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.02913, "mean_abs": 0.220264, "relative": 0.846 }, { "name": "hand_right_joints", "l2": 11.595881, "mean_abs": 0.295798, "relative": 0.914066 }, { "name": "body_joints", "l2": 5.625551, "mean_abs": 0.085156, "relative": 0.759839 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.085898, "mean_abs": 0.010562, "relative": 0.202707 }, { "name": "camera_rotation_matrix", "l2": 3.006009, "mean_abs": 0.206969, "relative": 0.923097 }, { "name": "imu_accel_gyro", "l2": 17.169426, "mean_abs": 1.060282, "relative": 0.92635 }, { "name": "depth_confidence", "l2": 10.809529, "mean_abs": 0.200836, "relative": 0.000359 }, { "name": "video_fisheye_cam0", "l2": 5.101666, "mean_abs": 0.106665, "relative": 0.923803 }, { "name": "video_fisheye_cam1", "l2": 5.287625, "mean_abs": 0.113844, "relative": 0.873735 }, { "name": "video_fisheye_cam2", "l2": 5.492073, "mean_abs": 0.113744, "relative": 0.876728 }, { "name": "video_fisheye_cam3", "l2": 5.309074, "mean_abs": 0.109584, "relative": 0.888941 }, { "name": "video_stereo_left", "l2": 5.185826, "mean_abs": 0.116826, "relative": 0.842707 }, { "name": "video_stereo_right", "l2": 6.099546, "mean_abs": 0.140771, "relative": 0.859908 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.963349, "mean_abs": 0.073966, "relative": 0.702653 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 705, "start_frame": 3525, "end_frame": 3544, "center_frame": 3534, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.078417, "mean_abs": 0.222471, "relative": 0.850157 }, { "name": "hand_right_joints", "l2": 11.824104, "mean_abs": 0.29422, "relative": 0.932056 }, { "name": "body_joints", "l2": 5.584156, "mean_abs": 0.084329, "relative": 0.754247 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.075468, "mean_abs": 0.009682, "relative": 0.178093 }, { "name": "camera_rotation_matrix", "l2": 3.006994, "mean_abs": 0.209775, "relative": 0.923399 }, { "name": "imu_accel_gyro", "l2": 17.173134, "mean_abs": 1.07244, "relative": 0.92655 }, { "name": "depth_confidence", "l2": 10.83474, "mean_abs": 0.202347, "relative": 0.00036 }, { "name": "video_fisheye_cam0", "l2": 5.097636, "mean_abs": 0.106922, "relative": 0.923073 }, { "name": "video_fisheye_cam1", "l2": 5.269115, "mean_abs": 0.114043, "relative": 0.870677 }, { "name": "video_fisheye_cam2", "l2": 5.519236, "mean_abs": 0.114458, "relative": 0.881064 }, { "name": "video_fisheye_cam3", "l2": 5.338339, "mean_abs": 0.110751, "relative": 0.89384 }, { "name": "video_stereo_left", "l2": 5.203235, "mean_abs": 0.117434, "relative": 0.845536 }, { "name": "video_stereo_right", "l2": 6.126861, "mean_abs": 0.141694, "relative": 0.863759 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.321166, "mean_abs": 0.096187, "relative": 0.83071 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 706, "start_frame": 3530, "end_frame": 3549, "center_frame": 3539, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.10893, "mean_abs": 0.223445, "relative": 0.852731 }, { "name": "hand_right_joints", "l2": 12.024247, "mean_abs": 0.294311, "relative": 0.947832 }, { "name": "body_joints", "l2": 5.578873, "mean_abs": 0.083547, "relative": 0.753534 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.063123, "mean_abs": 0.008456, "relative": 0.14896 }, { "name": "camera_rotation_matrix", "l2": 3.003903, "mean_abs": 0.209794, "relative": 0.92245 }, { "name": "imu_accel_gyro", "l2": 17.159849, "mean_abs": 1.072454, "relative": 0.925833 }, { "name": "depth_confidence", "l2": 10.835789, "mean_abs": 0.201752, "relative": 0.00036 }, { "name": "video_fisheye_cam0", "l2": 5.096083, "mean_abs": 0.106709, "relative": 0.922792 }, { "name": "video_fisheye_cam1", "l2": 5.246034, "mean_abs": 0.112975, "relative": 0.866863 }, { "name": "video_fisheye_cam2", "l2": 5.553493, "mean_abs": 0.114565, "relative": 0.886533 }, { "name": "video_fisheye_cam3", "l2": 5.344934, "mean_abs": 0.110784, "relative": 0.894945 }, { "name": "video_stereo_left", "l2": 5.216774, "mean_abs": 0.116479, "relative": 0.847736 }, { "name": "video_stereo_right", "l2": 6.129104, "mean_abs": 0.139965, "relative": 0.864075 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.325566, "mean_abs": 0.094372, "relative": 0.832285 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 707, "start_frame": 3535, "end_frame": 3554, "center_frame": 3544, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.183475, "mean_abs": 0.223818, "relative": 0.85902 }, { "name": "hand_right_joints", "l2": 12.166802, "mean_abs": 0.295677, "relative": 0.959069 }, { "name": "body_joints", "l2": 5.560982, "mean_abs": 0.081945, "relative": 0.751117 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.043637, "mean_abs": 0.005969, "relative": 0.102977 }, { "name": "camera_rotation_matrix", "l2": 2.999641, "mean_abs": 0.207537, "relative": 0.921141 }, { "name": "imu_accel_gyro", "l2": 17.08699, "mean_abs": 1.058398, "relative": 0.921902 }, { "name": "depth_confidence", "l2": 10.813914, "mean_abs": 0.199535, "relative": 0.000359 }, { "name": "video_fisheye_cam0", "l2": 5.09589, "mean_abs": 0.106109, "relative": 0.922757 }, { "name": "video_fisheye_cam1", "l2": 5.232704, "mean_abs": 0.11109, "relative": 0.86466 }, { "name": "video_fisheye_cam2", "l2": 5.586624, "mean_abs": 0.113708, "relative": 0.891821 }, { "name": "video_fisheye_cam3", "l2": 5.382703, "mean_abs": 0.110692, "relative": 0.901269 }, { "name": "video_stereo_left", "l2": 5.222302, "mean_abs": 0.114044, "relative": 0.848635 }, { "name": "video_stereo_right", "l2": 6.138393, "mean_abs": 0.136507, "relative": 0.865385 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.309823, "mean_abs": 0.094335, "relative": 0.826651 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 708, "start_frame": 3540, "end_frame": 3559, "center_frame": 3549, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.201453, "mean_abs": 0.221733, "relative": 0.860536 }, { "name": "hand_right_joints", "l2": 12.315664, "mean_abs": 0.299859, "relative": 0.970804 }, { "name": "body_joints", "l2": 5.556938, "mean_abs": 0.08161, "relative": 0.750571 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019084, "mean_abs": 0.003092, "relative": 0.045035 }, { "name": "camera_rotation_matrix", "l2": 3.000295, "mean_abs": 0.208037, "relative": 0.921342 }, { "name": "imu_accel_gyro", "l2": 17.077177, "mean_abs": 1.050188, "relative": 0.921373 }, { "name": "depth_confidence", "l2": 10.879281, "mean_abs": 0.199488, "relative": 0.000361 }, { "name": "video_fisheye_cam0", "l2": 5.095869, "mean_abs": 0.105883, "relative": 0.922753 }, { "name": "video_fisheye_cam1", "l2": 5.249277, "mean_abs": 0.111639, "relative": 0.867399 }, { "name": "video_fisheye_cam2", "l2": 5.608554, "mean_abs": 0.113937, "relative": 0.895322 }, { "name": "video_fisheye_cam3", "l2": 5.467502, "mean_abs": 0.110966, "relative": 0.915467 }, { "name": "video_stereo_left", "l2": 5.2241, "mean_abs": 0.113455, "relative": 0.848927 }, { "name": "video_stereo_right", "l2": 6.196839, "mean_abs": 0.136796, "relative": 0.873624 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.190581, "mean_abs": 0.088591, "relative": 0.783976 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011819, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 709, "start_frame": 3545, "end_frame": 3564, "center_frame": 3554, "action": "Move kettle away", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.223863, "mean_abs": 0.220797, "relative": 0.862426 }, { "name": "hand_right_joints", "l2": 12.354643, "mean_abs": 0.29701, "relative": 0.973876 }, { "name": "body_joints", "l2": 5.576249, "mean_abs": 0.082003, "relative": 0.753179 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.040636, "mean_abs": 0.005889, "relative": 0.095896 }, { "name": "camera_rotation_matrix", "l2": 3.001609, "mean_abs": 0.210735, "relative": 0.921745 }, { "name": "imu_accel_gyro", "l2": 17.133047, "mean_abs": 1.063905, "relative": 0.924387 }, { "name": "depth_confidence", "l2": 10.93184, "mean_abs": 0.201267, "relative": 0.000363 }, { "name": "video_fisheye_cam0", "l2": 5.090643, "mean_abs": 0.105943, "relative": 0.921807 }, { "name": "video_fisheye_cam1", "l2": 5.276838, "mean_abs": 0.112751, "relative": 0.871953 }, { "name": "video_fisheye_cam2", "l2": 5.618662, "mean_abs": 0.114557, "relative": 0.896936 }, { "name": "video_fisheye_cam3", "l2": 5.522224, "mean_abs": 0.111564, "relative": 0.92463 }, { "name": "video_stereo_left", "l2": 5.219533, "mean_abs": 0.114095, "relative": 0.848184 }, { "name": "video_stereo_right", "l2": 6.232917, "mean_abs": 0.137884, "relative": 0.878711 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.121402, "mean_abs": 0.082996, "relative": 0.759218 }, { "name": "caption_objects_interaction_text", "l2": 1.869497, "mean_abs": 0.016509, "relative": 0.858428 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 710, "start_frame": 3550, "end_frame": 3569, "center_frame": 3559, "action": "", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.218198, "mean_abs": 0.222327, "relative": 0.861949 }, { "name": "hand_right_joints", "l2": 12.379872, "mean_abs": 0.297978, "relative": 0.975865 }, { "name": "body_joints", "l2": 5.630913, "mean_abs": 0.08441, "relative": 0.760563 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.072909, "mean_abs": 0.010112, "relative": 0.172055 }, { "name": "camera_rotation_matrix", "l2": 3.002102, "mean_abs": 0.212546, "relative": 0.921897 }, { "name": "imu_accel_gyro", "l2": 17.126516, "mean_abs": 1.060268, "relative": 0.924035 }, { "name": "depth_confidence", "l2": 10.978422, "mean_abs": 0.202606, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.088985, "mean_abs": 0.106104, "relative": 0.921506 }, { "name": "video_fisheye_cam1", "l2": 5.313544, "mean_abs": 0.114184, "relative": 0.878018 }, { "name": "video_fisheye_cam2", "l2": 5.622643, "mean_abs": 0.115327, "relative": 0.897571 }, { "name": "video_fisheye_cam3", "l2": 5.596084, "mean_abs": 0.11403, "relative": 0.936997 }, { "name": "video_stereo_left", "l2": 5.227044, "mean_abs": 0.115294, "relative": 0.849405 }, { "name": "video_stereo_right", "l2": 6.259712, "mean_abs": 0.139921, "relative": 0.882488 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.052831, "mean_abs": 0.079941, "relative": 0.734677 }, { "name": "caption_objects_interaction_text", "l2": 1.869497, "mean_abs": 0.016673, "relative": 0.858428 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 711, "start_frame": 3555, "end_frame": 3574, "center_frame": 3564, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.196583, "mean_abs": 0.224319, "relative": 0.860125 }, { "name": "hand_right_joints", "l2": 12.207815, "mean_abs": 0.300015, "relative": 0.962302 }, { "name": "body_joints", "l2": 5.673817, "mean_abs": 0.086551, "relative": 0.766358 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.112125, "mean_abs": 0.015487, "relative": 0.2646 }, { "name": "camera_rotation_matrix", "l2": 3.002616, "mean_abs": 0.214446, "relative": 0.922055 }, { "name": "imu_accel_gyro", "l2": 17.109079, "mean_abs": 1.055228, "relative": 0.923094 }, { "name": "depth_confidence", "l2": 11.090134, "mean_abs": 0.205002, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.091981, "mean_abs": 0.106434, "relative": 0.922049 }, { "name": "video_fisheye_cam1", "l2": 5.355191, "mean_abs": 0.115305, "relative": 0.8849 }, { "name": "video_fisheye_cam2", "l2": 5.634975, "mean_abs": 0.116064, "relative": 0.89954 }, { "name": "video_fisheye_cam3", "l2": 5.662355, "mean_abs": 0.115725, "relative": 0.948093 }, { "name": "video_stereo_left", "l2": 5.246967, "mean_abs": 0.116755, "relative": 0.852643 }, { "name": "video_stereo_right", "l2": 6.28648, "mean_abs": 0.142034, "relative": 0.886262 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.077166, "mean_abs": 0.078628, "relative": 0.743386 }, { "name": "caption_objects_interaction_text", "l2": 1.869497, "mean_abs": 0.016459, "relative": 0.858428 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 712, "start_frame": 3560, "end_frame": 3579, "center_frame": 3569, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.170092, "mean_abs": 0.227337, "relative": 0.857891 }, { "name": "hand_right_joints", "l2": 11.880032, "mean_abs": 0.30205, "relative": 0.936464 }, { "name": "body_joints", "l2": 5.748497, "mean_abs": 0.088569, "relative": 0.776445 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.136412, "mean_abs": 0.019161, "relative": 0.321913 }, { "name": "camera_rotation_matrix", "l2": 3.003258, "mean_abs": 0.215569, "relative": 0.922252 }, { "name": "imu_accel_gyro", "l2": 17.122469, "mean_abs": 1.065022, "relative": 0.923816 }, { "name": "depth_confidence", "l2": 11.191556, "mean_abs": 0.206969, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.093909, "mean_abs": 0.106588, "relative": 0.922398 }, { "name": "video_fisheye_cam1", "l2": 5.401577, "mean_abs": 0.11619, "relative": 0.892565 }, { "name": "video_fisheye_cam2", "l2": 5.647666, "mean_abs": 0.116373, "relative": 0.901566 }, { "name": "video_fisheye_cam3", "l2": 5.70432, "mean_abs": 0.116552, "relative": 0.95512 }, { "name": "video_stereo_left", "l2": 5.25991, "mean_abs": 0.117354, "relative": 0.854746 }, { "name": "video_stereo_right", "l2": 6.32104, "mean_abs": 0.143753, "relative": 0.891134 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.066071, "mean_abs": 0.078469, "relative": 0.739416 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 713, "start_frame": 3565, "end_frame": 3584, "center_frame": 3574, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.138498, "mean_abs": 0.22872, "relative": 0.855226 }, { "name": "hand_right_joints", "l2": 11.56449, "mean_abs": 0.305468, "relative": 0.911591 }, { "name": "body_joints", "l2": 5.796774, "mean_abs": 0.091591, "relative": 0.782966 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.151957, "mean_abs": 0.020978, "relative": 0.358597 }, { "name": "camera_rotation_matrix", "l2": 3.002949, "mean_abs": 0.216677, "relative": 0.922157 }, { "name": "imu_accel_gyro", "l2": 17.1775, "mean_abs": 1.053478, "relative": 0.926785 }, { "name": "depth_confidence", "l2": 11.255281, "mean_abs": 0.208307, "relative": 0.000374 }, { "name": "video_fisheye_cam0", "l2": 5.09849, "mean_abs": 0.106929, "relative": 0.923227 }, { "name": "video_fisheye_cam1", "l2": 5.431816, "mean_abs": 0.117667, "relative": 0.897562 }, { "name": "video_fisheye_cam2", "l2": 5.673784, "mean_abs": 0.11712, "relative": 0.905735 }, { "name": "video_fisheye_cam3", "l2": 5.729571, "mean_abs": 0.116819, "relative": 0.959348 }, { "name": "video_stereo_left", "l2": 5.275106, "mean_abs": 0.118191, "relative": 0.857215 }, { "name": "video_stereo_right", "l2": 6.391393, "mean_abs": 0.147036, "relative": 0.901052 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.007281, "mean_abs": 0.077956, "relative": 0.718376 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 714, "start_frame": 3570, "end_frame": 3589, "center_frame": 3579, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.102632, "mean_abs": 0.22994, "relative": 0.8522 }, { "name": "hand_right_joints", "l2": 11.202084, "mean_abs": 0.299948, "relative": 0.883024 }, { "name": "body_joints", "l2": 5.799545, "mean_abs": 0.092022, "relative": 0.78334 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.156683, "mean_abs": 0.020827, "relative": 0.36975 }, { "name": "camera_rotation_matrix", "l2": 3.005764, "mean_abs": 0.21656, "relative": 0.923021 }, { "name": "imu_accel_gyro", "l2": 17.093744, "mean_abs": 1.003464, "relative": 0.922266 }, { "name": "depth_confidence", "l2": 11.306493, "mean_abs": 0.21012, "relative": 0.000375 }, { "name": "video_fisheye_cam0", "l2": 5.115029, "mean_abs": 0.107608, "relative": 0.926222 }, { "name": "video_fisheye_cam1", "l2": 5.468221, "mean_abs": 0.11934, "relative": 0.903577 }, { "name": "video_fisheye_cam2", "l2": 5.712185, "mean_abs": 0.118432, "relative": 0.911865 }, { "name": "video_fisheye_cam3", "l2": 5.746737, "mean_abs": 0.117345, "relative": 0.962222 }, { "name": "video_stereo_left", "l2": 5.352826, "mean_abs": 0.121255, "relative": 0.869845 }, { "name": "video_stereo_right", "l2": 6.487796, "mean_abs": 0.150108, "relative": 0.914643 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.006554, "mean_abs": 0.08029, "relative": 0.718116 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 715, "start_frame": 3575, "end_frame": 3594, "center_frame": 3584, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.06851, "mean_abs": 0.22673, "relative": 0.849322 }, { "name": "hand_right_joints", "l2": 10.759057, "mean_abs": 0.288306, "relative": 0.848101 }, { "name": "body_joints", "l2": 5.793384, "mean_abs": 0.09012, "relative": 0.782508 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.157116, "mean_abs": 0.019953, "relative": 0.370771 }, { "name": "camera_rotation_matrix", "l2": 3.005846, "mean_abs": 0.218747, "relative": 0.923046 }, { "name": "imu_accel_gyro", "l2": 16.967915, "mean_abs": 1.012011, "relative": 0.915477 }, { "name": "depth_confidence", "l2": 11.280445, "mean_abs": 0.211206, "relative": 0.000374 }, { "name": "video_fisheye_cam0", "l2": 5.128893, "mean_abs": 0.107929, "relative": 0.928733 }, { "name": "video_fisheye_cam1", "l2": 5.49605, "mean_abs": 0.12012, "relative": 0.908176 }, { "name": "video_fisheye_cam2", "l2": 5.752031, "mean_abs": 0.119658, "relative": 0.918226 }, { "name": "video_fisheye_cam3", "l2": 5.77474, "mean_abs": 0.118677, "relative": 0.966911 }, { "name": "video_stereo_left", "l2": 5.428666, "mean_abs": 0.124077, "relative": 0.882169 }, { "name": "video_stereo_right", "l2": 6.548841, "mean_abs": 0.151169, "relative": 0.923249 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.947261, "mean_abs": 0.074331, "relative": 0.696896 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 716, "start_frame": 3580, "end_frame": 3599, "center_frame": 3589, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.037255, "mean_abs": 0.223033, "relative": 0.846685 }, { "name": "hand_right_joints", "l2": 10.558446, "mean_abs": 0.284339, "relative": 0.832288 }, { "name": "body_joints", "l2": 6.04263, "mean_abs": 0.096546, "relative": 0.816173 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.16675, "mean_abs": 0.020742, "relative": 0.393505 }, { "name": "camera_rotation_matrix", "l2": 3.007474, "mean_abs": 0.22017, "relative": 0.923546 }, { "name": "imu_accel_gyro", "l2": 17.075499, "mean_abs": 1.034118, "relative": 0.921282 }, { "name": "depth_confidence", "l2": 11.227948, "mean_abs": 0.212255, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.14281, "mean_abs": 0.108127, "relative": 0.931253 }, { "name": "video_fisheye_cam1", "l2": 5.539682, "mean_abs": 0.122122, "relative": 0.915386 }, { "name": "video_fisheye_cam2", "l2": 5.765942, "mean_abs": 0.120082, "relative": 0.920447 }, { "name": "video_fisheye_cam3", "l2": 5.806429, "mean_abs": 0.120364, "relative": 0.972217 }, { "name": "video_stereo_left", "l2": 5.488305, "mean_abs": 0.127166, "relative": 0.89186 }, { "name": "video_stereo_right", "l2": 6.682951, "mean_abs": 0.156124, "relative": 0.942156 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.995108, "mean_abs": 0.079579, "relative": 0.714019 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 717, "start_frame": 3585, "end_frame": 3604, "center_frame": 3594, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.979871, "mean_abs": 0.222929, "relative": 0.841845 }, { "name": "hand_right_joints", "l2": 10.922304, "mean_abs": 0.29801, "relative": 0.86097 }, { "name": "body_joints", "l2": 6.152764, "mean_abs": 0.102392, "relative": 0.831049 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08169, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.16893, "mean_abs": 0.019341, "relative": 0.398652 }, { "name": "camera_rotation_matrix", "l2": 3.005429, "mean_abs": 0.220155, "relative": 0.922918 }, { "name": "imu_accel_gyro", "l2": 16.990093, "mean_abs": 1.033702, "relative": 0.916674 }, { "name": "depth_confidence", "l2": 11.1429, "mean_abs": 0.211548, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.154872, "mean_abs": 0.108271, "relative": 0.933437 }, { "name": "video_fisheye_cam1", "l2": 5.524857, "mean_abs": 0.122374, "relative": 0.912936 }, { "name": "video_fisheye_cam2", "l2": 5.802081, "mean_abs": 0.120412, "relative": 0.926216 }, { "name": "video_fisheye_cam3", "l2": 5.795117, "mean_abs": 0.120634, "relative": 0.970322 }, { "name": "video_stereo_left", "l2": 5.531917, "mean_abs": 0.128076, "relative": 0.898948 }, { "name": "video_stereo_right", "l2": 6.785336, "mean_abs": 0.159222, "relative": 0.95659 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.025691, "mean_abs": 0.076949, "relative": 0.724964 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 718, "start_frame": 3590, "end_frame": 3609, "center_frame": 3599, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.983228, "mean_abs": 0.223904, "relative": 0.842128 }, { "name": "hand_right_joints", "l2": 11.145322, "mean_abs": 0.298382, "relative": 0.87855 }, { "name": "body_joints", "l2": 6.171595, "mean_abs": 0.103144, "relative": 0.833592 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.158716, "mean_abs": 0.018846, "relative": 0.374547 }, { "name": "camera_rotation_matrix", "l2": 3.005218, "mean_abs": 0.220078, "relative": 0.922854 }, { "name": "imu_accel_gyro", "l2": 17.033531, "mean_abs": 1.028644, "relative": 0.919018 }, { "name": "depth_confidence", "l2": 11.091823, "mean_abs": 0.209997, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.170648, "mean_abs": 0.108536, "relative": 0.936294 }, { "name": "video_fisheye_cam1", "l2": 5.517512, "mean_abs": 0.122175, "relative": 0.911722 }, { "name": "video_fisheye_cam2", "l2": 5.829175, "mean_abs": 0.12066, "relative": 0.930541 }, { "name": "video_fisheye_cam3", "l2": 5.766096, "mean_abs": 0.120393, "relative": 0.965463 }, { "name": "video_stereo_left", "l2": 5.540886, "mean_abs": 0.12654, "relative": 0.900405 }, { "name": "video_stereo_right", "l2": 6.851971, "mean_abs": 0.160345, "relative": 0.965984 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.031438, "mean_abs": 0.079605, "relative": 0.727021 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 719, "start_frame": 3595, "end_frame": 3614, "center_frame": 3604, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.941909, "mean_abs": 0.216853, "relative": 0.838642 }, { "name": "hand_right_joints", "l2": 11.354346, "mean_abs": 0.297991, "relative": 0.895026 }, { "name": "body_joints", "l2": 6.188421, "mean_abs": 0.101488, "relative": 0.835865 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.125266, "mean_abs": 0.015695, "relative": 0.29561 }, { "name": "camera_rotation_matrix", "l2": 3.004086, "mean_abs": 0.21963, "relative": 0.922506 }, { "name": "imu_accel_gyro", "l2": 17.086323, "mean_abs": 1.04825, "relative": 0.921866 }, { "name": "depth_confidence", "l2": 11.016649, "mean_abs": 0.207451, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.17485, "mean_abs": 0.108536, "relative": 0.937055 }, { "name": "video_fisheye_cam1", "l2": 5.508018, "mean_abs": 0.121284, "relative": 0.910154 }, { "name": "video_fisheye_cam2", "l2": 5.809738, "mean_abs": 0.11978, "relative": 0.927438 }, { "name": "video_fisheye_cam3", "l2": 5.744019, "mean_abs": 0.119473, "relative": 0.961767 }, { "name": "video_stereo_left", "l2": 5.523543, "mean_abs": 0.123766, "relative": 0.897587 }, { "name": "video_stereo_right", "l2": 6.861021, "mean_abs": 0.158238, "relative": 0.96726 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.127178, "mean_abs": 0.088795, "relative": 0.761285 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 720, "start_frame": 3600, "end_frame": 3619, "center_frame": 3609, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.872911, "mean_abs": 0.205231, "relative": 0.832822 }, { "name": "hand_right_joints", "l2": 11.670125, "mean_abs": 0.290769, "relative": 0.919918 }, { "name": "body_joints", "l2": 6.114395, "mean_abs": 0.094127, "relative": 0.825866 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.081715, "mean_abs": 0.011455, "relative": 0.192836 }, { "name": "camera_rotation_matrix", "l2": 3.001383, "mean_abs": 0.216213, "relative": 0.921676 }, { "name": "imu_accel_gyro", "l2": 17.205627, "mean_abs": 1.099983, "relative": 0.928303 }, { "name": "depth_confidence", "l2": 10.956614, "mean_abs": 0.203277, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.176317, "mean_abs": 0.107915, "relative": 0.93732 }, { "name": "video_fisheye_cam1", "l2": 5.492452, "mean_abs": 0.11855, "relative": 0.907581 }, { "name": "video_fisheye_cam2", "l2": 5.777358, "mean_abs": 0.118015, "relative": 0.922269 }, { "name": "video_fisheye_cam3", "l2": 5.716779, "mean_abs": 0.117176, "relative": 0.957206 }, { "name": "video_stereo_left", "l2": 5.488258, "mean_abs": 0.120368, "relative": 0.891853 }, { "name": "video_stereo_right", "l2": 6.835394, "mean_abs": 0.153244, "relative": 0.963647 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.217721, "mean_abs": 0.099938, "relative": 0.793689 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 721, "start_frame": 3605, "end_frame": 3624, "center_frame": 3614, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.870651, "mean_abs": 0.195917, "relative": 0.832632 }, { "name": "hand_right_joints", "l2": 11.675878, "mean_abs": 0.291918, "relative": 0.920372 }, { "name": "body_joints", "l2": 6.070394, "mean_abs": 0.092097, "relative": 0.819923 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.03206, "mean_abs": 0.004627, "relative": 0.075657 }, { "name": "camera_rotation_matrix", "l2": 3.001416, "mean_abs": 0.215763, "relative": 0.921686 }, { "name": "imu_accel_gyro", "l2": 17.232876, "mean_abs": 1.088005, "relative": 0.929773 }, { "name": "depth_confidence", "l2": 10.920128, "mean_abs": 0.200261, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.201816, "mean_abs": 0.108928, "relative": 0.941938 }, { "name": "video_fisheye_cam1", "l2": 5.460385, "mean_abs": 0.117629, "relative": 0.902283 }, { "name": "video_fisheye_cam2", "l2": 5.749439, "mean_abs": 0.117971, "relative": 0.917812 }, { "name": "video_fisheye_cam3", "l2": 5.670955, "mean_abs": 0.115924, "relative": 0.949533 }, { "name": "video_stereo_left", "l2": 5.446223, "mean_abs": 0.119693, "relative": 0.885022 }, { "name": "video_stereo_right", "l2": 6.806857, "mean_abs": 0.151599, "relative": 0.959624 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.133782, "mean_abs": 0.094099, "relative": 0.763649 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 722, "start_frame": 3610, "end_frame": 3629, "center_frame": 3619, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.820341, "mean_abs": 0.205037, "relative": 0.828388 }, { "name": "hand_right_joints", "l2": 11.285937, "mean_abs": 0.296237, "relative": 0.889634 }, { "name": "body_joints", "l2": 6.321201, "mean_abs": 0.105561, "relative": 0.8538 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.076009, "mean_abs": 0.010725, "relative": 0.179371 }, { "name": "camera_rotation_matrix", "l2": 3.011894, "mean_abs": 0.225341, "relative": 0.924904 }, { "name": "imu_accel_gyro", "l2": 17.197495, "mean_abs": 1.107158, "relative": 0.927864 }, { "name": "depth_confidence", "l2": 11.033708, "mean_abs": 0.207545, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.215998, "mean_abs": 0.110699, "relative": 0.944506 }, { "name": "video_fisheye_cam1", "l2": 5.510409, "mean_abs": 0.122324, "relative": 0.910549 }, { "name": "video_fisheye_cam2", "l2": 5.709402, "mean_abs": 0.119273, "relative": 0.911421 }, { "name": "video_fisheye_cam3", "l2": 5.677142, "mean_abs": 0.119465, "relative": 0.950569 }, { "name": "video_stereo_left", "l2": 5.437721, "mean_abs": 0.125545, "relative": 0.883641 }, { "name": "video_stereo_right", "l2": 6.912055, "mean_abs": 0.161003, "relative": 0.974455 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.196137, "mean_abs": 0.097013, "relative": 0.785965 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 723, "start_frame": 3615, "end_frame": 3634, "center_frame": 3624, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.836621, "mean_abs": 0.212949, "relative": 0.829761 }, { "name": "hand_right_joints", "l2": 11.004817, "mean_abs": 0.291133, "relative": 0.867474 }, { "name": "body_joints", "l2": 6.42348, "mean_abs": 0.113846, "relative": 0.867614 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08169, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.141304, "mean_abs": 0.018228, "relative": 0.333457 }, { "name": "camera_rotation_matrix", "l2": 3.025504, "mean_abs": 0.230417, "relative": 0.929083 }, { "name": "imu_accel_gyro", "l2": 17.231344, "mean_abs": 1.134982, "relative": 0.92969 }, { "name": "depth_confidence", "l2": 11.092676, "mean_abs": 0.211806, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.236544, "mean_abs": 0.111788, "relative": 0.948226 }, { "name": "video_fisheye_cam1", "l2": 5.579712, "mean_abs": 0.1256, "relative": 0.922 }, { "name": "video_fisheye_cam2", "l2": 5.654045, "mean_abs": 0.119578, "relative": 0.902584 }, { "name": "video_fisheye_cam3", "l2": 5.688493, "mean_abs": 0.121033, "relative": 0.95247 }, { "name": "video_stereo_left", "l2": 5.43502, "mean_abs": 0.12743, "relative": 0.883202 }, { "name": "video_stereo_right", "l2": 6.921629, "mean_abs": 0.163886, "relative": 0.975805 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.245365, "mean_abs": 0.097644, "relative": 0.803582 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 724, "start_frame": 3620, "end_frame": 3639, "center_frame": 3629, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.001982, "mean_abs": 0.220427, "relative": 0.84371 }, { "name": "hand_right_joints", "l2": 10.672816, "mean_abs": 0.278534, "relative": 0.841303 }, { "name": "body_joints", "l2": 6.301569, "mean_abs": 0.112094, "relative": 0.851148 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081689, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.186047, "mean_abs": 0.022619, "relative": 0.439046 }, { "name": "camera_rotation_matrix", "l2": 3.024918, "mean_abs": 0.226536, "relative": 0.928903 }, { "name": "imu_accel_gyro", "l2": 17.091558, "mean_abs": 1.092668, "relative": 0.922148 }, { "name": "depth_confidence", "l2": 11.019722, "mean_abs": 0.2107, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.244449, "mean_abs": 0.111538, "relative": 0.949658 }, { "name": "video_fisheye_cam1", "l2": 5.575803, "mean_abs": 0.125303, "relative": 0.921354 }, { "name": "video_fisheye_cam2", "l2": 5.565339, "mean_abs": 0.11791, "relative": 0.888424 }, { "name": "video_fisheye_cam3", "l2": 5.657937, "mean_abs": 0.120039, "relative": 0.947353 }, { "name": "video_stereo_left", "l2": 5.351764, "mean_abs": 0.125282, "relative": 0.869672 }, { "name": "video_stereo_right", "l2": 6.852306, "mean_abs": 0.162746, "relative": 0.966031 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.105013, "mean_abs": 0.08423, "relative": 0.753353 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 725, "start_frame": 3625, "end_frame": 3644, "center_frame": 3634, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.405947, "mean_abs": 0.236368, "relative": 0.877786 }, { "name": "hand_right_joints", "l2": 10.435174, "mean_abs": 0.268514, "relative": 0.822571 }, { "name": "body_joints", "l2": 6.157701, "mean_abs": 0.106981, "relative": 0.831716 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.192767, "mean_abs": 0.024357, "relative": 0.454902 }, { "name": "camera_rotation_matrix", "l2": 3.01609, "mean_abs": 0.219783, "relative": 0.926192 }, { "name": "imu_accel_gyro", "l2": 16.832207, "mean_abs": 1.068953, "relative": 0.908156 }, { "name": "depth_confidence", "l2": 10.882925, "mean_abs": 0.2084, "relative": 0.000361 }, { "name": "video_fisheye_cam0", "l2": 5.225103, "mean_abs": 0.110404, "relative": 0.946154 }, { "name": "video_fisheye_cam1", "l2": 5.556507, "mean_abs": 0.123532, "relative": 0.918166 }, { "name": "video_fisheye_cam2", "l2": 5.48625, "mean_abs": 0.115873, "relative": 0.875798 }, { "name": "video_fisheye_cam3", "l2": 5.587066, "mean_abs": 0.116132, "relative": 0.935487 }, { "name": "video_stereo_left", "l2": 5.269758, "mean_abs": 0.122653, "relative": 0.856346 }, { "name": "video_stereo_right", "l2": 6.811649, "mean_abs": 0.161314, "relative": 0.9603 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.14437, "mean_abs": 0.082314, "relative": 0.767438 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 726, "start_frame": 3630, "end_frame": 3649, "center_frame": 3639, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.571102, "mean_abs": 0.243105, "relative": 0.891718 }, { "name": "hand_right_joints", "l2": 10.720192, "mean_abs": 0.28058, "relative": 0.845038 }, { "name": "body_joints", "l2": 6.010962, "mean_abs": 0.100357, "relative": 0.811896 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.184926, "mean_abs": 0.027303, "relative": 0.4364 }, { "name": "camera_rotation_matrix", "l2": 3.004801, "mean_abs": 0.212128, "relative": 0.922726 }, { "name": "imu_accel_gyro", "l2": 17.131737, "mean_abs": 1.152245, "relative": 0.924316 }, { "name": "depth_confidence", "l2": 10.760283, "mean_abs": 0.205479, "relative": 0.000357 }, { "name": "video_fisheye_cam0", "l2": 5.210489, "mean_abs": 0.109186, "relative": 0.943508 }, { "name": "video_fisheye_cam1", "l2": 5.507482, "mean_abs": 0.11973, "relative": 0.910065 }, { "name": "video_fisheye_cam2", "l2": 5.415634, "mean_abs": 0.113003, "relative": 0.864525 }, { "name": "video_fisheye_cam3", "l2": 5.563146, "mean_abs": 0.115451, "relative": 0.931482 }, { "name": "video_stereo_left", "l2": 5.275227, "mean_abs": 0.121835, "relative": 0.857235 }, { "name": "video_stereo_right", "l2": 6.734125, "mean_abs": 0.158287, "relative": 0.94937 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.116174, "mean_abs": 0.0804, "relative": 0.757347 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 727, "start_frame": 3635, "end_frame": 3654, "center_frame": 3644, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.733851, "mean_abs": 0.243677, "relative": 0.905446 }, { "name": "hand_right_joints", "l2": 10.936397, "mean_abs": 0.278323, "relative": 0.862081 }, { "name": "body_joints", "l2": 5.854798, "mean_abs": 0.096488, "relative": 0.790803 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.13461, "mean_abs": 0.02094, "relative": 0.31766 }, { "name": "camera_rotation_matrix", "l2": 3.020042, "mean_abs": 0.216524, "relative": 0.927406 }, { "name": "imu_accel_gyro", "l2": 17.429419, "mean_abs": 1.21233, "relative": 0.940377 }, { "name": "depth_confidence", "l2": 10.626938, "mean_abs": 0.2033, "relative": 0.000353 }, { "name": "video_fisheye_cam0", "l2": 5.210872, "mean_abs": 0.110092, "relative": 0.943578 }, { "name": "video_fisheye_cam1", "l2": 5.431461, "mean_abs": 0.119137, "relative": 0.897503 }, { "name": "video_fisheye_cam2", "l2": 5.400574, "mean_abs": 0.112724, "relative": 0.862121 }, { "name": "video_fisheye_cam3", "l2": 5.567666, "mean_abs": 0.116765, "relative": 0.932238 }, { "name": "video_stereo_left", "l2": 5.296287, "mean_abs": 0.12278, "relative": 0.860657 }, { "name": "video_stereo_right", "l2": 6.625648, "mean_abs": 0.155924, "relative": 0.934077 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.397372, "mean_abs": 0.100512, "relative": 0.857984 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 728, "start_frame": 3640, "end_frame": 3659, "center_frame": 3649, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.020975, "mean_abs": 0.235772, "relative": 0.929666 }, { "name": "hand_right_joints", "l2": 11.301028, "mean_abs": 0.286477, "relative": 0.890823 }, { "name": "body_joints", "l2": 5.704309, "mean_abs": 0.092792, "relative": 0.770476 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.077216, "mean_abs": 0.012357, "relative": 0.182219 }, { "name": "camera_rotation_matrix", "l2": 3.01478, "mean_abs": 0.215665, "relative": 0.92579 }, { "name": "imu_accel_gyro", "l2": 17.332733, "mean_abs": 1.175119, "relative": 0.935161 }, { "name": "depth_confidence", "l2": 10.446699, "mean_abs": 0.199308, "relative": 0.000347 }, { "name": "video_fisheye_cam0", "l2": 5.216974, "mean_abs": 0.110225, "relative": 0.944682 }, { "name": "video_fisheye_cam1", "l2": 5.369589, "mean_abs": 0.118373, "relative": 0.887279 }, { "name": "video_fisheye_cam2", "l2": 5.424297, "mean_abs": 0.113387, "relative": 0.865908 }, { "name": "video_fisheye_cam3", "l2": 5.565084, "mean_abs": 0.116518, "relative": 0.931806 }, { "name": "video_stereo_left", "l2": 5.330899, "mean_abs": 0.122783, "relative": 0.866282 }, { "name": "video_stereo_right", "l2": 6.620018, "mean_abs": 0.155482, "relative": 0.933284 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.247265, "mean_abs": 0.092404, "relative": 0.804262 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 729, "start_frame": 3645, "end_frame": 3664, "center_frame": 3654, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.004388, "mean_abs": 0.231746, "relative": 0.928267 }, { "name": "hand_right_joints", "l2": 11.408207, "mean_abs": 0.291042, "relative": 0.899272 }, { "name": "body_joints", "l2": 5.731597, "mean_abs": 0.090962, "relative": 0.774162 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.102339, "mean_abs": 0.014721, "relative": 0.241506 }, { "name": "camera_rotation_matrix", "l2": 3.007221, "mean_abs": 0.211332, "relative": 0.923469 }, { "name": "imu_accel_gyro", "l2": 17.262846, "mean_abs": 1.140485, "relative": 0.93139 }, { "name": "depth_confidence", "l2": 10.371467, "mean_abs": 0.196309, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.213131, "mean_abs": 0.109581, "relative": 0.943987 }, { "name": "video_fisheye_cam1", "l2": 5.351747, "mean_abs": 0.117809, "relative": 0.884331 }, { "name": "video_fisheye_cam2", "l2": 5.473626, "mean_abs": 0.114593, "relative": 0.873783 }, { "name": "video_fisheye_cam3", "l2": 5.553124, "mean_abs": 0.11478, "relative": 0.929804 }, { "name": "video_stereo_left", "l2": 5.378601, "mean_abs": 0.122531, "relative": 0.874033 }, { "name": "video_stereo_right", "l2": 6.632211, "mean_abs": 0.153707, "relative": 0.935003 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.219102, "mean_abs": 0.089629, "relative": 0.794183 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 730, "start_frame": 3650, "end_frame": 3669, "center_frame": 3659, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.923434, "mean_abs": 0.230339, "relative": 0.921438 }, { "name": "hand_right_joints", "l2": 11.355847, "mean_abs": 0.292207, "relative": 0.895145 }, { "name": "body_joints", "l2": 5.783721, "mean_abs": 0.0919, "relative": 0.781202 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.172764, "mean_abs": 0.023723, "relative": 0.407699 }, { "name": "camera_rotation_matrix", "l2": 3.005065, "mean_abs": 0.209607, "relative": 0.922807 }, { "name": "imu_accel_gyro", "l2": 17.159382, "mean_abs": 1.157949, "relative": 0.925808 }, { "name": "depth_confidence", "l2": 10.488904, "mean_abs": 0.195836, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.208563, "mean_abs": 0.108971, "relative": 0.943159 }, { "name": "video_fisheye_cam1", "l2": 5.375687, "mean_abs": 0.11738, "relative": 0.888287 }, { "name": "video_fisheye_cam2", "l2": 5.518198, "mean_abs": 0.114932, "relative": 0.880898 }, { "name": "video_fisheye_cam3", "l2": 5.582208, "mean_abs": 0.114992, "relative": 0.934673 }, { "name": "video_stereo_left", "l2": 5.444964, "mean_abs": 0.121919, "relative": 0.884818 }, { "name": "video_stereo_right", "l2": 6.736882, "mean_abs": 0.153398, "relative": 0.949759 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.25727, "mean_abs": 0.091636, "relative": 0.807843 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 731, "start_frame": 3655, "end_frame": 3674, "center_frame": 3664, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.832969, "mean_abs": 0.228851, "relative": 0.913807 }, { "name": "hand_right_joints", "l2": 11.32666, "mean_abs": 0.293244, "relative": 0.892844 }, { "name": "body_joints", "l2": 5.766539, "mean_abs": 0.091993, "relative": 0.778882 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.201657, "mean_abs": 0.028441, "relative": 0.475882 }, { "name": "camera_rotation_matrix", "l2": 3.008541, "mean_abs": 0.213381, "relative": 0.923874 }, { "name": "imu_accel_gyro", "l2": 17.313824, "mean_abs": 1.188908, "relative": 0.93414 }, { "name": "depth_confidence", "l2": 10.622103, "mean_abs": 0.200648, "relative": 0.000353 }, { "name": "video_fisheye_cam0", "l2": 5.212463, "mean_abs": 0.109753, "relative": 0.943866 }, { "name": "video_fisheye_cam1", "l2": 5.428332, "mean_abs": 0.118619, "relative": 0.896986 }, { "name": "video_fisheye_cam2", "l2": 5.544252, "mean_abs": 0.115301, "relative": 0.885057 }, { "name": "video_fisheye_cam3", "l2": 5.635149, "mean_abs": 0.118134, "relative": 0.943538 }, { "name": "video_stereo_left", "l2": 5.472278, "mean_abs": 0.124136, "relative": 0.889256 }, { "name": "video_stereo_right", "l2": 6.796846, "mean_abs": 0.155634, "relative": 0.958213 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.030748, "mean_abs": 0.081769, "relative": 0.726774 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 732, "start_frame": 3660, "end_frame": 3679, "center_frame": 3669, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.681574, "mean_abs": 0.232025, "relative": 0.901036 }, { "name": "hand_right_joints", "l2": 10.901778, "mean_abs": 0.287774, "relative": 0.859352 }, { "name": "body_joints", "l2": 5.882945, "mean_abs": 0.098034, "relative": 0.794605 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.159415, "mean_abs": 0.023416, "relative": 0.376196 }, { "name": "camera_rotation_matrix", "l2": 3.018147, "mean_abs": 0.221314, "relative": 0.926824 }, { "name": "imu_accel_gyro", "l2": 17.387871, "mean_abs": 1.199002, "relative": 0.938136 }, { "name": "depth_confidence", "l2": 10.845385, "mean_abs": 0.205136, "relative": 0.00036 }, { "name": "video_fisheye_cam0", "l2": 5.227896, "mean_abs": 0.111242, "relative": 0.94666 }, { "name": "video_fisheye_cam1", "l2": 5.526035, "mean_abs": 0.122098, "relative": 0.913131 }, { "name": "video_fisheye_cam2", "l2": 5.557601, "mean_abs": 0.116824, "relative": 0.887188 }, { "name": "video_fisheye_cam3", "l2": 5.666972, "mean_abs": 0.120241, "relative": 0.948866 }, { "name": "video_stereo_left", "l2": 5.505464, "mean_abs": 0.127786, "relative": 0.894649 }, { "name": "video_stereo_right", "l2": 6.904735, "mean_abs": 0.161391, "relative": 0.973423 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.046321, "mean_abs": 0.081342, "relative": 0.732348 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011522, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 733, "start_frame": 3665, "end_frame": 3684, "center_frame": 3674, "action": "Place kettle on table", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.454657, "mean_abs": 0.233308, "relative": 0.881895 }, { "name": "hand_right_joints", "l2": 10.5185, "mean_abs": 0.268873, "relative": 0.829139 }, { "name": "body_joints", "l2": 5.863424, "mean_abs": 0.100922, "relative": 0.791968 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.094972, "mean_abs": 0.014442, "relative": 0.224122 }, { "name": "camera_rotation_matrix", "l2": 3.019696, "mean_abs": 0.222664, "relative": 0.9273 }, { "name": "imu_accel_gyro", "l2": 17.282522, "mean_abs": 1.15165, "relative": 0.932452 }, { "name": "depth_confidence", "l2": 10.950925, "mean_abs": 0.207684, "relative": 0.000363 }, { "name": "video_fisheye_cam0", "l2": 5.219563, "mean_abs": 0.110734, "relative": 0.945151 }, { "name": "video_fisheye_cam1", "l2": 5.601632, "mean_abs": 0.123149, "relative": 0.925622 }, { "name": "video_fisheye_cam2", "l2": 5.543253, "mean_abs": 0.116931, "relative": 0.884898 }, { "name": "video_fisheye_cam3", "l2": 5.638107, "mean_abs": 0.118835, "relative": 0.944033 }, { "name": "video_stereo_left", "l2": 5.474235, "mean_abs": 0.127452, "relative": 0.889574 }, { "name": "video_stereo_right", "l2": 6.942605, "mean_abs": 0.163879, "relative": 0.978762 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.007501, "mean_abs": 0.079692, "relative": 0.718455 }, { "name": "caption_objects_interaction_text", "l2": 1.892734, "mean_abs": 0.015721, "relative": 0.869098 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 734, "start_frame": 3670, "end_frame": 3689, "center_frame": 3679, "action": "", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.453823, "mean_abs": 0.243848, "relative": 0.881825 }, { "name": "hand_right_joints", "l2": 10.462669, "mean_abs": 0.277823, "relative": 0.824738 }, { "name": "body_joints", "l2": 5.86855, "mean_abs": 0.102265, "relative": 0.79266 }, { "name": "body_contacts", "l2": 3.464562, "mean_abs": 0.08172, "relative": 0.999758 }, { "name": "camera_translation", "l2": 0.121427, "mean_abs": 0.014781, "relative": 0.28655 }, { "name": "camera_rotation_matrix", "l2": 3.008755, "mean_abs": 0.217271, "relative": 0.92394 }, { "name": "imu_accel_gyro", "l2": 17.191969, "mean_abs": 1.104181, "relative": 0.927566 }, { "name": "depth_confidence", "l2": 10.999474, "mean_abs": 0.206314, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.199091, "mean_abs": 0.109347, "relative": 0.941444 }, { "name": "video_fisheye_cam1", "l2": 5.643026, "mean_abs": 0.122697, "relative": 0.932462 }, { "name": "video_fisheye_cam2", "l2": 5.50681, "mean_abs": 0.115795, "relative": 0.87908 }, { "name": "video_fisheye_cam3", "l2": 5.559561, "mean_abs": 0.115512, "relative": 0.930881 }, { "name": "video_stereo_left", "l2": 5.396467, "mean_abs": 0.124269, "relative": 0.876937 }, { "name": "video_stereo_right", "l2": 6.871435, "mean_abs": 0.161413, "relative": 0.968728 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.034411, "mean_abs": 0.083094, "relative": 0.728085 }, { "name": "caption_objects_interaction_text", "l2": 1.892734, "mean_abs": 0.015821, "relative": 0.869098 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 735, "start_frame": 3675, "end_frame": 3694, "center_frame": 3684, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.386859, "mean_abs": 0.246821, "relative": 0.876176 }, { "name": "hand_right_joints", "l2": 10.149512, "mean_abs": 0.270304, "relative": 0.800053 }, { "name": "body_joints", "l2": 5.75243, "mean_abs": 0.097556, "relative": 0.776976 }, { "name": "body_contacts", "l2": 3.464359, "mean_abs": 0.081715, "relative": 0.999699 }, { "name": "camera_translation", "l2": 0.223431, "mean_abs": 0.024606, "relative": 0.527266 }, { "name": "camera_rotation_matrix", "l2": 3.000604, "mean_abs": 0.208012, "relative": 0.921437 }, { "name": "imu_accel_gyro", "l2": 17.088476, "mean_abs": 1.06134, "relative": 0.921982 }, { "name": "depth_confidence", "l2": 11.027709, "mean_abs": 0.203608, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.171185, "mean_abs": 0.108326, "relative": 0.936391 }, { "name": "video_fisheye_cam1", "l2": 5.648923, "mean_abs": 0.120622, "relative": 0.933437 }, { "name": "video_fisheye_cam2", "l2": 5.449068, "mean_abs": 0.113239, "relative": 0.869863 }, { "name": "video_fisheye_cam3", "l2": 5.49126, "mean_abs": 0.113743, "relative": 0.919445 }, { "name": "video_stereo_left", "l2": 5.332734, "mean_abs": 0.120714, "relative": 0.86658 }, { "name": "video_stereo_right", "l2": 6.801427, "mean_abs": 0.157078, "relative": 0.958858 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.08676, "mean_abs": 0.084486, "relative": 0.74682 }, { "name": "caption_objects_interaction_text", "l2": 1.892734, "mean_abs": 0.01556, "relative": 0.869098 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 736, "start_frame": 3680, "end_frame": 3699, "center_frame": 3689, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.273278, "mean_abs": 0.246571, "relative": 0.866595 }, { "name": "hand_right_joints", "l2": 9.643824, "mean_abs": 0.25455, "relative": 0.760191 }, { "name": "body_joints", "l2": 5.66047, "mean_abs": 0.092097, "relative": 0.764555 }, { "name": "body_contacts", "l2": 3.464359, "mean_abs": 0.0817, "relative": 0.999699 }, { "name": "camera_translation", "l2": 0.274992, "mean_abs": 0.030099, "relative": 0.648943 }, { "name": "camera_rotation_matrix", "l2": 3.004885, "mean_abs": 0.21335, "relative": 0.922751 }, { "name": "imu_accel_gyro", "l2": 17.275188, "mean_abs": 1.125332, "relative": 0.932056 }, { "name": "depth_confidence", "l2": 11.053084, "mean_abs": 0.203035, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.143452, "mean_abs": 0.108654, "relative": 0.931369 }, { "name": "video_fisheye_cam1", "l2": 5.633489, "mean_abs": 0.119938, "relative": 0.930887 }, { "name": "video_fisheye_cam2", "l2": 5.427555, "mean_abs": 0.112703, "relative": 0.866428 }, { "name": "video_fisheye_cam3", "l2": 5.4122, "mean_abs": 0.113858, "relative": 0.906208 }, { "name": "video_stereo_left", "l2": 5.313315, "mean_abs": 0.119698, "relative": 0.863424 }, { "name": "video_stereo_right", "l2": 6.768639, "mean_abs": 0.153443, "relative": 0.954236 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.093786, "mean_abs": 0.083775, "relative": 0.749335 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 737, "start_frame": 3685, "end_frame": 3704, "center_frame": 3694, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.259523, "mean_abs": 0.245746, "relative": 0.865435 }, { "name": "hand_right_joints", "l2": 9.553101, "mean_abs": 0.247089, "relative": 0.75304 }, { "name": "body_joints", "l2": 5.629719, "mean_abs": 0.091331, "relative": 0.760402 }, { "name": "body_contacts", "l2": 3.464359, "mean_abs": 0.081701, "relative": 0.999699 }, { "name": "camera_translation", "l2": 0.251732, "mean_abs": 0.02841, "relative": 0.594053 }, { "name": "camera_rotation_matrix", "l2": 3.006952, "mean_abs": 0.215115, "relative": 0.923386 }, { "name": "imu_accel_gyro", "l2": 17.357256, "mean_abs": 1.140812, "relative": 0.936484 }, { "name": "depth_confidence", "l2": 11.03524, "mean_abs": 0.203496, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.125659, "mean_abs": 0.108389, "relative": 0.928147 }, { "name": "video_fisheye_cam1", "l2": 5.617482, "mean_abs": 0.119793, "relative": 0.928241 }, { "name": "video_fisheye_cam2", "l2": 5.443612, "mean_abs": 0.112981, "relative": 0.868992 }, { "name": "video_fisheye_cam3", "l2": 5.381271, "mean_abs": 0.112638, "relative": 0.901029 }, { "name": "video_stereo_left", "l2": 5.325149, "mean_abs": 0.120241, "relative": 0.865347 }, { "name": "video_stereo_right", "l2": 6.796643, "mean_abs": 0.154328, "relative": 0.958184 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.069221, "mean_abs": 0.0823, "relative": 0.740543 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 738, "start_frame": 3690, "end_frame": 3709, "center_frame": 3699, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.361548, "mean_abs": 0.242441, "relative": 0.874041 }, { "name": "hand_right_joints", "l2": 9.600746, "mean_abs": 0.243348, "relative": 0.756796 }, { "name": "body_joints", "l2": 5.56609, "mean_abs": 0.089013, "relative": 0.751807 }, { "name": "body_contacts", "l2": 3.464142, "mean_abs": 0.081705, "relative": 0.999636 }, { "name": "camera_translation", "l2": 0.159784, "mean_abs": 0.019826, "relative": 0.377069 }, { "name": "camera_rotation_matrix", "l2": 3.009008, "mean_abs": 0.217377, "relative": 0.924017 }, { "name": "imu_accel_gyro", "l2": 17.298576, "mean_abs": 1.12292, "relative": 0.933318 }, { "name": "depth_confidence", "l2": 11.003507, "mean_abs": 0.203697, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.110317, "mean_abs": 0.107719, "relative": 0.925369 }, { "name": "video_fisheye_cam1", "l2": 5.570472, "mean_abs": 0.11967, "relative": 0.920474 }, { "name": "video_fisheye_cam2", "l2": 5.448748, "mean_abs": 0.113459, "relative": 0.869812 }, { "name": "video_fisheye_cam3", "l2": 5.426629, "mean_abs": 0.11405, "relative": 0.908624 }, { "name": "video_stereo_left", "l2": 5.339603, "mean_abs": 0.121215, "relative": 0.867696 }, { "name": "video_stereo_right", "l2": 6.796964, "mean_abs": 0.155142, "relative": 0.958229 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.129388, "mean_abs": 0.0913, "relative": 0.762076 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 739, "start_frame": 3695, "end_frame": 3714, "center_frame": 3704, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.399364, "mean_abs": 0.241751, "relative": 0.877231 }, { "name": "hand_right_joints", "l2": 9.716628, "mean_abs": 0.252212, "relative": 0.76593 }, { "name": "body_joints", "l2": 5.606276, "mean_abs": 0.090448, "relative": 0.757235 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.080423, "mean_abs": 0.01141, "relative": 0.189788 }, { "name": "camera_rotation_matrix", "l2": 3.014403, "mean_abs": 0.218601, "relative": 0.925674 }, { "name": "imu_accel_gyro", "l2": 17.22192, "mean_abs": 1.093075, "relative": 0.929182 }, { "name": "depth_confidence", "l2": 10.897362, "mean_abs": 0.203324, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.106243, "mean_abs": 0.108013, "relative": 0.924631 }, { "name": "video_fisheye_cam1", "l2": 5.525106, "mean_abs": 0.120773, "relative": 0.912977 }, { "name": "video_fisheye_cam2", "l2": 5.426578, "mean_abs": 0.114215, "relative": 0.866272 }, { "name": "video_fisheye_cam3", "l2": 5.459506, "mean_abs": 0.1144, "relative": 0.914128 }, { "name": "video_stereo_left", "l2": 5.33169, "mean_abs": 0.122077, "relative": 0.86641 }, { "name": "video_stereo_right", "l2": 6.71486, "mean_abs": 0.154536, "relative": 0.946654 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.148219, "mean_abs": 0.088103, "relative": 0.768815 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 740, "start_frame": 3700, "end_frame": 3719, "center_frame": 3709, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.425212, "mean_abs": 0.243778, "relative": 0.879411 }, { "name": "hand_right_joints", "l2": 10.452446, "mean_abs": 0.279614, "relative": 0.823932 }, { "name": "body_joints", "l2": 5.71369, "mean_abs": 0.095101, "relative": 0.771743 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.095062, "mean_abs": 0.014863, "relative": 0.224332 }, { "name": "camera_rotation_matrix", "l2": 3.018475, "mean_abs": 0.218008, "relative": 0.926925 }, { "name": "imu_accel_gyro", "l2": 17.275236, "mean_abs": 1.133153, "relative": 0.932058 }, { "name": "depth_confidence", "l2": 10.759188, "mean_abs": 0.202937, "relative": 0.000357 }, { "name": "video_fisheye_cam0", "l2": 5.115395, "mean_abs": 0.108601, "relative": 0.926289 }, { "name": "video_fisheye_cam1", "l2": 5.490909, "mean_abs": 0.120916, "relative": 0.907326 }, { "name": "video_fisheye_cam2", "l2": 5.406214, "mean_abs": 0.114559, "relative": 0.863022 }, { "name": "video_fisheye_cam3", "l2": 5.520111, "mean_abs": 0.115456, "relative": 0.924276 }, { "name": "video_stereo_left", "l2": 5.308529, "mean_abs": 0.122821, "relative": 0.862647 }, { "name": "video_stereo_right", "l2": 6.632843, "mean_abs": 0.154858, "relative": 0.935092 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.361623, "mean_abs": 0.097295, "relative": 0.845189 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 741, "start_frame": 3705, "end_frame": 3724, "center_frame": 3714, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.453043, "mean_abs": 0.24084, "relative": 0.881759 }, { "name": "hand_right_joints", "l2": 10.967984, "mean_abs": 0.296943, "relative": 0.864571 }, { "name": "body_joints", "l2": 5.817159, "mean_abs": 0.097491, "relative": 0.785719 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.127576, "mean_abs": 0.019545, "relative": 0.301061 }, { "name": "camera_rotation_matrix", "l2": 3.011191, "mean_abs": 0.213579, "relative": 0.924688 }, { "name": "imu_accel_gyro", "l2": 17.312454, "mean_abs": 1.160553, "relative": 0.934067 }, { "name": "depth_confidence", "l2": 10.648272, "mean_abs": 0.201343, "relative": 0.000353 }, { "name": "video_fisheye_cam0", "l2": 5.121008, "mean_abs": 0.108116, "relative": 0.927305 }, { "name": "video_fisheye_cam1", "l2": 5.407676, "mean_abs": 0.118654, "relative": 0.893573 }, { "name": "video_fisheye_cam2", "l2": 5.378909, "mean_abs": 0.113333, "relative": 0.858663 }, { "name": "video_fisheye_cam3", "l2": 5.56827, "mean_abs": 0.11513, "relative": 0.93234 }, { "name": "video_stereo_left", "l2": 5.228595, "mean_abs": 0.119693, "relative": 0.849657 }, { "name": "video_stereo_right", "l2": 6.519946, "mean_abs": 0.150958, "relative": 0.919176 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.133963, "mean_abs": 0.088307, "relative": 0.763713 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 742, "start_frame": 3710, "end_frame": 3729, "center_frame": 3719, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.389496, "mean_abs": 0.235146, "relative": 0.876398 }, { "name": "hand_right_joints", "l2": 11.284374, "mean_abs": 0.311959, "relative": 0.889511 }, { "name": "body_joints", "l2": 5.924999, "mean_abs": 0.097551, "relative": 0.800285 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.136129, "mean_abs": 0.01929, "relative": 0.321245 }, { "name": "camera_rotation_matrix", "l2": 3.01258, "mean_abs": 0.207889, "relative": 0.925114 }, { "name": "imu_accel_gyro", "l2": 17.306814, "mean_abs": 1.206018, "relative": 0.933762 }, { "name": "depth_confidence", "l2": 10.47426, "mean_abs": 0.19809, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.135387, "mean_abs": 0.107784, "relative": 0.929909 }, { "name": "video_fisheye_cam1", "l2": 5.329093, "mean_abs": 0.116638, "relative": 0.880588 }, { "name": "video_fisheye_cam2", "l2": 5.383187, "mean_abs": 0.112354, "relative": 0.859346 }, { "name": "video_fisheye_cam3", "l2": 5.594497, "mean_abs": 0.115287, "relative": 0.936731 }, { "name": "video_stereo_left", "l2": 5.286863, "mean_abs": 0.121785, "relative": 0.859126 }, { "name": "video_stereo_right", "l2": 6.490727, "mean_abs": 0.151834, "relative": 0.915056 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.185066, "mean_abs": 0.088996, "relative": 0.782002 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 743, "start_frame": 3715, "end_frame": 3734, "center_frame": 3724, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.362472, "mean_abs": 0.235002, "relative": 0.874119 }, { "name": "hand_right_joints", "l2": 11.669743, "mean_abs": 0.317664, "relative": 0.919888 }, { "name": "body_joints", "l2": 5.868886, "mean_abs": 0.092687, "relative": 0.792706 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.126655, "mean_abs": 0.01656, "relative": 0.298887 }, { "name": "camera_rotation_matrix", "l2": 3.019829, "mean_abs": 0.209951, "relative": 0.927341 }, { "name": "imu_accel_gyro", "l2": 17.556612, "mean_abs": 1.278094, "relative": 0.94724 }, { "name": "depth_confidence", "l2": 10.39003, "mean_abs": 0.198414, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.147449, "mean_abs": 0.108926, "relative": 0.932093 }, { "name": "video_fisheye_cam1", "l2": 5.265533, "mean_abs": 0.116256, "relative": 0.870085 }, { "name": "video_fisheye_cam2", "l2": 5.395889, "mean_abs": 0.11297, "relative": 0.861373 }, { "name": "video_fisheye_cam3", "l2": 5.571025, "mean_abs": 0.115811, "relative": 0.932801 }, { "name": "video_stereo_left", "l2": 5.355722, "mean_abs": 0.124471, "relative": 0.870315 }, { "name": "video_stereo_right", "l2": 6.397024, "mean_abs": 0.151268, "relative": 0.901846 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.096678, "mean_abs": 0.085515, "relative": 0.75037 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 744, "start_frame": 3720, "end_frame": 3739, "center_frame": 3729, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.381231, "mean_abs": 0.238692, "relative": 0.875701 }, { "name": "hand_right_joints", "l2": 12.14375, "mean_abs": 0.313531, "relative": 0.957252 }, { "name": "body_joints", "l2": 5.752798, "mean_abs": 0.086877, "relative": 0.777026 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.08836, "mean_abs": 0.012099, "relative": 0.208518 }, { "name": "camera_rotation_matrix", "l2": 3.011676, "mean_abs": 0.207162, "relative": 0.924837 }, { "name": "imu_accel_gyro", "l2": 17.536125, "mean_abs": 1.283074, "relative": 0.946134 }, { "name": "depth_confidence", "l2": 10.329476, "mean_abs": 0.19537, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.1507, "mean_abs": 0.108651, "relative": 0.932682 }, { "name": "video_fisheye_cam1", "l2": 5.200013, "mean_abs": 0.114102, "relative": 0.859258 }, { "name": "video_fisheye_cam2", "l2": 5.401458, "mean_abs": 0.112689, "relative": 0.862262 }, { "name": "video_fisheye_cam3", "l2": 5.491336, "mean_abs": 0.113417, "relative": 0.919458 }, { "name": "video_stereo_left", "l2": 5.345921, "mean_abs": 0.122613, "relative": 0.868723 }, { "name": "video_stereo_right", "l2": 6.316987, "mean_abs": 0.147349, "relative": 0.890563 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.17801, "mean_abs": 0.090752, "relative": 0.779477 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 745, "start_frame": 3725, "end_frame": 3744, "center_frame": 3734, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.37403, "mean_abs": 0.24188, "relative": 0.875094 }, { "name": "hand_right_joints", "l2": 11.681777, "mean_abs": 0.315625, "relative": 0.920837 }, { "name": "body_joints", "l2": 5.78171, "mean_abs": 0.092791, "relative": 0.780931 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.037428, "mean_abs": 0.004993, "relative": 0.088325 }, { "name": "camera_rotation_matrix", "l2": 3.001037, "mean_abs": 0.198335, "relative": 0.92157 }, { "name": "imu_accel_gyro", "l2": 17.36488, "mean_abs": 1.238322, "relative": 0.936895 }, { "name": "depth_confidence", "l2": 10.263309, "mean_abs": 0.189378, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.146909, "mean_abs": 0.107442, "relative": 0.931995 }, { "name": "video_fisheye_cam1", "l2": 5.165935, "mean_abs": 0.110822, "relative": 0.853627 }, { "name": "video_fisheye_cam2", "l2": 5.446995, "mean_abs": 0.111261, "relative": 0.869532 }, { "name": "video_fisheye_cam3", "l2": 5.459209, "mean_abs": 0.111086, "relative": 0.914079 }, { "name": "video_stereo_left", "l2": 5.352938, "mean_abs": 0.11828, "relative": 0.869863 }, { "name": "video_stereo_right", "l2": 6.318153, "mean_abs": 0.142096, "relative": 0.890727 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.211251, "mean_abs": 0.090842, "relative": 0.791373 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 746, "start_frame": 3730, "end_frame": 3749, "center_frame": 3739, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.396846, "mean_abs": 0.245802, "relative": 0.877018 }, { "name": "hand_right_joints", "l2": 11.50566, "mean_abs": 0.315312, "relative": 0.906954 }, { "name": "body_joints", "l2": 5.81876, "mean_abs": 0.095669, "relative": 0.785935 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.072929, "mean_abs": 0.010693, "relative": 0.172101 }, { "name": "camera_rotation_matrix", "l2": 3.011156, "mean_abs": 0.207492, "relative": 0.924677 }, { "name": "imu_accel_gyro", "l2": 17.467783, "mean_abs": 1.291195, "relative": 0.942447 }, { "name": "depth_confidence", "l2": 10.309524, "mean_abs": 0.194081, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.150753, "mean_abs": 0.108227, "relative": 0.932691 }, { "name": "video_fisheye_cam1", "l2": 5.191841, "mean_abs": 0.11339, "relative": 0.857908 }, { "name": "video_fisheye_cam2", "l2": 5.481021, "mean_abs": 0.113154, "relative": 0.874963 }, { "name": "video_fisheye_cam3", "l2": 5.478591, "mean_abs": 0.112696, "relative": 0.917324 }, { "name": "video_stereo_left", "l2": 5.389708, "mean_abs": 0.122485, "relative": 0.875838 }, { "name": "video_stereo_right", "l2": 6.332908, "mean_abs": 0.14749, "relative": 0.892807 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.126514, "mean_abs": 0.084052, "relative": 0.761047 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 747, "start_frame": 3735, "end_frame": 3754, "center_frame": 3744, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.422965, "mean_abs": 0.249545, "relative": 0.879222 }, { "name": "hand_right_joints", "l2": 11.295923, "mean_abs": 0.307195, "relative": 0.890421 }, { "name": "body_joints", "l2": 5.795469, "mean_abs": 0.094812, "relative": 0.782789 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.097995, "mean_abs": 0.014882, "relative": 0.231255 }, { "name": "camera_rotation_matrix", "l2": 3.019642, "mean_abs": 0.213145, "relative": 0.927283 }, { "name": "imu_accel_gyro", "l2": 17.491014, "mean_abs": 1.278904, "relative": 0.9437 }, { "name": "depth_confidence", "l2": 10.312042, "mean_abs": 0.196617, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.151759, "mean_abs": 0.108905, "relative": 0.932873 }, { "name": "video_fisheye_cam1", "l2": 5.213833, "mean_abs": 0.114894, "relative": 0.861542 }, { "name": "video_fisheye_cam2", "l2": 5.483866, "mean_abs": 0.114353, "relative": 0.875418 }, { "name": "video_fisheye_cam3", "l2": 5.51975, "mean_abs": 0.114659, "relative": 0.924216 }, { "name": "video_stereo_left", "l2": 5.410974, "mean_abs": 0.124592, "relative": 0.879294 }, { "name": "video_stereo_right", "l2": 6.35926, "mean_abs": 0.149555, "relative": 0.896522 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.147985, "mean_abs": 0.085508, "relative": 0.768731 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 748, "start_frame": 3740, "end_frame": 3759, "center_frame": 3749, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.660979, "mean_abs": 0.259493, "relative": 0.899299 }, { "name": "hand_right_joints", "l2": 10.767551, "mean_abs": 0.278772, "relative": 0.848771 }, { "name": "body_joints", "l2": 5.659069, "mean_abs": 0.088592, "relative": 0.764366 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.090206, "mean_abs": 0.01341, "relative": 0.212874 }, { "name": "camera_rotation_matrix", "l2": 3.018127, "mean_abs": 0.21142, "relative": 0.926818 }, { "name": "imu_accel_gyro", "l2": 17.083715, "mean_abs": 1.157041, "relative": 0.921725 }, { "name": "depth_confidence", "l2": 10.250061, "mean_abs": 0.196386, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.146796, "mean_abs": 0.108757, "relative": 0.931975 }, { "name": "video_fisheye_cam1", "l2": 5.218382, "mean_abs": 0.114757, "relative": 0.862294 }, { "name": "video_fisheye_cam2", "l2": 5.454739, "mean_abs": 0.114083, "relative": 0.870768 }, { "name": "video_fisheye_cam3", "l2": 5.550626, "mean_abs": 0.115412, "relative": 0.929385 }, { "name": "video_stereo_left", "l2": 5.384124, "mean_abs": 0.123623, "relative": 0.874931 }, { "name": "video_stereo_right", "l2": 6.339257, "mean_abs": 0.148038, "relative": 0.893702 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.142385, "mean_abs": 0.082238, "relative": 0.766727 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010557, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 749, "start_frame": 3745, "end_frame": 3764, "center_frame": 3754, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.977989, "mean_abs": 0.274388, "relative": 0.92604 }, { "name": "hand_right_joints", "l2": 10.411656, "mean_abs": 0.260914, "relative": 0.820717 }, { "name": "body_joints", "l2": 5.699935, "mean_abs": 0.090649, "relative": 0.769886 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.103522, "mean_abs": 0.016727, "relative": 0.244296 }, { "name": "camera_rotation_matrix", "l2": 3.008738, "mean_abs": 0.204588, "relative": 0.923935 }, { "name": "imu_accel_gyro", "l2": 17.161642, "mean_abs": 1.140035, "relative": 0.92593 }, { "name": "depth_confidence", "l2": 10.094995, "mean_abs": 0.193942, "relative": 0.000335 }, { "name": "video_fisheye_cam0", "l2": 5.155931, "mean_abs": 0.108565, "relative": 0.933629 }, { "name": "video_fisheye_cam1", "l2": 5.225638, "mean_abs": 0.113682, "relative": 0.863493 }, { "name": "video_fisheye_cam2", "l2": 5.421741, "mean_abs": 0.112734, "relative": 0.8655 }, { "name": "video_fisheye_cam3", "l2": 5.605139, "mean_abs": 0.116592, "relative": 0.938513 }, { "name": "video_stereo_left", "l2": 5.325299, "mean_abs": 0.121216, "relative": 0.865372 }, { "name": "video_stereo_right", "l2": 6.3218, "mean_abs": 0.146607, "relative": 0.891241 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.16283, "mean_abs": 0.084688, "relative": 0.774044 }, { "name": "caption_objects_interaction_text", "l2": 1.786199, "mean_abs": 0.011902, "relative": 0.820179 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 750, "start_frame": 3750, "end_frame": 3769, "center_frame": 3759, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.154022, "mean_abs": 0.279918, "relative": 0.940889 }, { "name": "hand_right_joints", "l2": 10.459897, "mean_abs": 0.259726, "relative": 0.82452 }, { "name": "body_joints", "l2": 5.795433, "mean_abs": 0.092753, "relative": 0.782784 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.12693, "mean_abs": 0.019928, "relative": 0.299537 }, { "name": "camera_rotation_matrix", "l2": 3.006564, "mean_abs": 0.200664, "relative": 0.923267 }, { "name": "imu_accel_gyro", "l2": 17.159664, "mean_abs": 1.130128, "relative": 0.925823 }, { "name": "depth_confidence", "l2": 9.954576, "mean_abs": 0.191243, "relative": 0.00033 }, { "name": "video_fisheye_cam0", "l2": 5.154789, "mean_abs": 0.108141, "relative": 0.933422 }, { "name": "video_fisheye_cam1", "l2": 5.197777, "mean_abs": 0.112382, "relative": 0.858889 }, { "name": "video_fisheye_cam2", "l2": 5.396832, "mean_abs": 0.111892, "relative": 0.861524 }, { "name": "video_fisheye_cam3", "l2": 5.617622, "mean_abs": 0.116189, "relative": 0.940603 }, { "name": "video_stereo_left", "l2": 5.297948, "mean_abs": 0.120323, "relative": 0.860927 }, { "name": "video_stereo_right", "l2": 6.224048, "mean_abs": 0.143711, "relative": 0.87746 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.191462, "mean_abs": 0.084936, "relative": 0.784291 }, { "name": "caption_objects_interaction_text", "l2": 1.786199, "mean_abs": 0.01192, "relative": 0.820179 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 751, "start_frame": 3755, "end_frame": 3774, "center_frame": 3764, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.286197, "mean_abs": 0.278112, "relative": 0.952039 }, { "name": "hand_right_joints", "l2": 10.501907, "mean_abs": 0.260145, "relative": 0.827831 }, { "name": "body_joints", "l2": 5.881794, "mean_abs": 0.09343, "relative": 0.794449 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.101557, "mean_abs": 0.014968, "relative": 0.23966 }, { "name": "camera_rotation_matrix", "l2": 3.005494, "mean_abs": 0.199556, "relative": 0.922938 }, { "name": "imu_accel_gyro", "l2": 17.14164, "mean_abs": 1.107547, "relative": 0.92485 }, { "name": "depth_confidence", "l2": 9.821972, "mean_abs": 0.186812, "relative": 0.000326 }, { "name": "video_fisheye_cam0", "l2": 5.149598, "mean_abs": 0.107772, "relative": 0.932482 }, { "name": "video_fisheye_cam1", "l2": 5.190868, "mean_abs": 0.112325, "relative": 0.857747 }, { "name": "video_fisheye_cam2", "l2": 5.375145, "mean_abs": 0.11175, "relative": 0.858062 }, { "name": "video_fisheye_cam3", "l2": 5.620751, "mean_abs": 0.115433, "relative": 0.941127 }, { "name": "video_stereo_left", "l2": 5.281485, "mean_abs": 0.119961, "relative": 0.858252 }, { "name": "video_stereo_right", "l2": 6.146756, "mean_abs": 0.140854, "relative": 0.866564 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.269155, "mean_abs": 0.090845, "relative": 0.812096 }, { "name": "caption_objects_interaction_text", "l2": 1.786199, "mean_abs": 0.011817, "relative": 0.82018 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 752, "start_frame": 3760, "end_frame": 3779, "center_frame": 3769, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.310119, "mean_abs": 0.265522, "relative": 0.954057 }, { "name": "hand_right_joints", "l2": 10.519519, "mean_abs": 0.269799, "relative": 0.82922 }, { "name": "body_joints", "l2": 5.98697, "mean_abs": 0.093971, "relative": 0.808655 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.045897, "mean_abs": 0.007244, "relative": 0.10831 }, { "name": "camera_rotation_matrix", "l2": 3.004727, "mean_abs": 0.197493, "relative": 0.922703 }, { "name": "imu_accel_gyro", "l2": 17.189157, "mean_abs": 1.082603, "relative": 0.927414 }, { "name": "depth_confidence", "l2": 9.741432, "mean_abs": 0.184386, "relative": 0.000323 }, { "name": "video_fisheye_cam0", "l2": 5.14732, "mean_abs": 0.107098, "relative": 0.93207 }, { "name": "video_fisheye_cam1", "l2": 5.211238, "mean_abs": 0.113001, "relative": 0.861113 }, { "name": "video_fisheye_cam2", "l2": 5.374676, "mean_abs": 0.111549, "relative": 0.857987 }, { "name": "video_fisheye_cam3", "l2": 5.6428, "mean_abs": 0.116195, "relative": 0.944819 }, { "name": "video_stereo_left", "l2": 5.27958, "mean_abs": 0.119394, "relative": 0.857942 }, { "name": "video_stereo_right", "l2": 6.120796, "mean_abs": 0.139755, "relative": 0.862904 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.15777, "mean_abs": 0.083779, "relative": 0.772233 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 753, "start_frame": 3765, "end_frame": 3784, "center_frame": 3774, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.127008, "mean_abs": 0.264793, "relative": 0.938611 }, { "name": "hand_right_joints", "l2": 10.55461, "mean_abs": 0.275321, "relative": 0.831986 }, { "name": "body_joints", "l2": 6.093782, "mean_abs": 0.09829, "relative": 0.823082 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.111775, "mean_abs": 0.016123, "relative": 0.263774 }, { "name": "camera_rotation_matrix", "l2": 3.003916, "mean_abs": 0.195678, "relative": 0.922454 }, { "name": "imu_accel_gyro", "l2": 17.182592, "mean_abs": 1.112905, "relative": 0.92706 }, { "name": "depth_confidence", "l2": 9.915051, "mean_abs": 0.188915, "relative": 0.000329 }, { "name": "video_fisheye_cam0", "l2": 5.151172, "mean_abs": 0.106847, "relative": 0.932767 }, { "name": "video_fisheye_cam1", "l2": 5.263178, "mean_abs": 0.113649, "relative": 0.869696 }, { "name": "video_fisheye_cam2", "l2": 5.438204, "mean_abs": 0.11351, "relative": 0.868128 }, { "name": "video_fisheye_cam3", "l2": 5.637402, "mean_abs": 0.116702, "relative": 0.943915 }, { "name": "video_stereo_left", "l2": 5.317286, "mean_abs": 0.121798, "relative": 0.864069 }, { "name": "video_stereo_right", "l2": 6.181716, "mean_abs": 0.142459, "relative": 0.871492 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.190135, "mean_abs": 0.085665, "relative": 0.783817 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 754, "start_frame": 3770, "end_frame": 3789, "center_frame": 3779, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.029937, "mean_abs": 0.267697, "relative": 0.930422 }, { "name": "hand_right_joints", "l2": 10.619025, "mean_abs": 0.278926, "relative": 0.837063 }, { "name": "body_joints", "l2": 6.152343, "mean_abs": 0.098729, "relative": 0.830992 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.158747, "mean_abs": 0.021353, "relative": 0.37462 }, { "name": "camera_rotation_matrix", "l2": 3.005604, "mean_abs": 0.203259, "relative": 0.922972 }, { "name": "imu_accel_gyro", "l2": 17.230564, "mean_abs": 1.119461, "relative": 0.929648 }, { "name": "depth_confidence", "l2": 10.13435, "mean_abs": 0.194192, "relative": 0.000336 }, { "name": "video_fisheye_cam0", "l2": 5.1392, "mean_abs": 0.107488, "relative": 0.930599 }, { "name": "video_fisheye_cam1", "l2": 5.327713, "mean_abs": 0.115891, "relative": 0.88036 }, { "name": "video_fisheye_cam2", "l2": 5.516752, "mean_abs": 0.116242, "relative": 0.880667 }, { "name": "video_fisheye_cam3", "l2": 5.616982, "mean_abs": 0.116528, "relative": 0.940496 }, { "name": "video_stereo_left", "l2": 5.398491, "mean_abs": 0.12493, "relative": 0.877266 }, { "name": "video_stereo_right", "l2": 6.294457, "mean_abs": 0.146389, "relative": 0.887386 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.190981, "mean_abs": 0.085699, "relative": 0.784119 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 755, "start_frame": 3775, "end_frame": 3794, "center_frame": 3784, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.802862, "mean_abs": 0.262165, "relative": 0.911268 }, { "name": "hand_right_joints", "l2": 10.683284, "mean_abs": 0.281097, "relative": 0.842129 }, { "name": "body_joints", "l2": 6.173239, "mean_abs": 0.094947, "relative": 0.833814 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.159963, "mean_abs": 0.021065, "relative": 0.377491 }, { "name": "camera_rotation_matrix", "l2": 3.009618, "mean_abs": 0.207902, "relative": 0.924205 }, { "name": "imu_accel_gyro", "l2": 17.162561, "mean_abs": 1.103061, "relative": 0.925979 }, { "name": "depth_confidence", "l2": 10.263742, "mean_abs": 0.195414, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.138583, "mean_abs": 0.108106, "relative": 0.930488 }, { "name": "video_fisheye_cam1", "l2": 5.380881, "mean_abs": 0.116887, "relative": 0.889145 }, { "name": "video_fisheye_cam2", "l2": 5.56849, "mean_abs": 0.117517, "relative": 0.888927 }, { "name": "video_fisheye_cam3", "l2": 5.584211, "mean_abs": 0.115762, "relative": 0.935009 }, { "name": "video_stereo_left", "l2": 5.445694, "mean_abs": 0.125763, "relative": 0.884936 }, { "name": "video_stereo_right", "l2": 6.364339, "mean_abs": 0.147726, "relative": 0.897238 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.097137, "mean_abs": 0.083363, "relative": 0.750534 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 756, "start_frame": 3780, "end_frame": 3799, "center_frame": 3789, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.573833, "mean_abs": 0.256149, "relative": 0.891948 }, { "name": "hand_right_joints", "l2": 10.795251, "mean_abs": 0.285446, "relative": 0.850955 }, { "name": "body_joints", "l2": 6.198926, "mean_abs": 0.090996, "relative": 0.837284 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.129232, "mean_abs": 0.017441, "relative": 0.304969 }, { "name": "camera_rotation_matrix", "l2": 3.014458, "mean_abs": 0.211278, "relative": 0.925691 }, { "name": "imu_accel_gyro", "l2": 16.996426, "mean_abs": 1.054133, "relative": 0.917016 }, { "name": "depth_confidence", "l2": 10.385238, "mean_abs": 0.195751, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.140708, "mean_abs": 0.108883, "relative": 0.930872 }, { "name": "video_fisheye_cam1", "l2": 5.416626, "mean_abs": 0.117978, "relative": 0.895052 }, { "name": "video_fisheye_cam2", "l2": 5.624624, "mean_abs": 0.118909, "relative": 0.897888 }, { "name": "video_fisheye_cam3", "l2": 5.557401, "mean_abs": 0.115056, "relative": 0.93052 }, { "name": "video_stereo_left", "l2": 5.567417, "mean_abs": 0.128039, "relative": 0.904716 }, { "name": "video_stereo_right", "l2": 6.490351, "mean_abs": 0.150183, "relative": 0.915003 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.066004, "mean_abs": 0.080007, "relative": 0.739392 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 757, "start_frame": 3785, "end_frame": 3804, "center_frame": 3794, "action": "Pick up white bottle", "subtask": "Set down kettle and retrieve white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.661272, "mean_abs": 0.260159, "relative": 0.899324 }, { "name": "hand_right_joints", "l2": 10.895627, "mean_abs": 0.28795, "relative": 0.858867 }, { "name": "body_joints", "l2": 6.202602, "mean_abs": 0.090794, "relative": 0.83778 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.090222, "mean_abs": 0.012821, "relative": 0.212911 }, { "name": "camera_rotation_matrix", "l2": 3.011909, "mean_abs": 0.212991, "relative": 0.924908 }, { "name": "imu_accel_gyro", "l2": 17.114305, "mean_abs": 1.077953, "relative": 0.923376 }, { "name": "depth_confidence", "l2": 10.393967, "mean_abs": 0.196693, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.130299, "mean_abs": 0.108759, "relative": 0.928987 }, { "name": "video_fisheye_cam1", "l2": 5.422843, "mean_abs": 0.118177, "relative": 0.896079 }, { "name": "video_fisheye_cam2", "l2": 5.661543, "mean_abs": 0.119554, "relative": 0.903781 }, { "name": "video_fisheye_cam3", "l2": 5.576049, "mean_abs": 0.116171, "relative": 0.933642 }, { "name": "video_stereo_left", "l2": 5.668278, "mean_abs": 0.129698, "relative": 0.921106 }, { "name": "video_stereo_right", "l2": 6.621226, "mean_abs": 0.152772, "relative": 0.933454 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.069465, "mean_abs": 0.080597, "relative": 0.74063 }, { "name": "caption_objects_interaction_text", "l2": 1.90567, "mean_abs": 0.016325, "relative": 0.875038 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 758, "start_frame": 3790, "end_frame": 3809, "center_frame": 3799, "action": "", "subtask": "", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.736894, "mean_abs": 0.261007, "relative": 0.905703 }, { "name": "hand_right_joints", "l2": 10.972657, "mean_abs": 0.287726, "relative": 0.864939 }, { "name": "body_joints", "l2": 6.188114, "mean_abs": 0.089653, "relative": 0.835824 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.066245, "mean_abs": 0.01022, "relative": 0.156329 }, { "name": "camera_rotation_matrix", "l2": 3.006088, "mean_abs": 0.211681, "relative": 0.923121 }, { "name": "imu_accel_gyro", "l2": 17.076536, "mean_abs": 1.081813, "relative": 0.921338 }, { "name": "depth_confidence", "l2": 10.348453, "mean_abs": 0.195284, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.114994, "mean_abs": 0.108167, "relative": 0.926216 }, { "name": "video_fisheye_cam1", "l2": 5.42126, "mean_abs": 0.117851, "relative": 0.895817 }, { "name": "video_fisheye_cam2", "l2": 5.65339, "mean_abs": 0.118612, "relative": 0.90248 }, { "name": "video_fisheye_cam3", "l2": 5.595981, "mean_abs": 0.116283, "relative": 0.936979 }, { "name": "video_stereo_left", "l2": 5.68458, "mean_abs": 0.128456, "relative": 0.923756 }, { "name": "video_stereo_right", "l2": 6.639968, "mean_abs": 0.151498, "relative": 0.936096 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.018735, "mean_abs": 0.080568, "relative": 0.722475 }, { "name": "caption_objects_interaction_text", "l2": 1.90567, "mean_abs": 0.01666, "relative": 0.875038 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 759, "start_frame": 3795, "end_frame": 3814, "center_frame": 3804, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "gooseneck kettle", "coffee dripper", "scale", "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.797323, "mean_abs": 0.261114, "relative": 0.9108 }, { "name": "hand_right_joints", "l2": 11.046974, "mean_abs": 0.286432, "relative": 0.870797 }, { "name": "body_joints", "l2": 6.163881, "mean_abs": 0.088113, "relative": 0.83255 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.047138, "mean_abs": 0.007113, "relative": 0.111239 }, { "name": "camera_rotation_matrix", "l2": 3.002311, "mean_abs": 0.208036, "relative": 0.921961 }, { "name": "imu_accel_gyro", "l2": 17.013016, "mean_abs": 1.074586, "relative": 0.917911 }, { "name": "depth_confidence", "l2": 10.306316, "mean_abs": 0.192463, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.110544, "mean_abs": 0.107194, "relative": 0.92541 }, { "name": "video_fisheye_cam1", "l2": 5.419513, "mean_abs": 0.11663, "relative": 0.895529 }, { "name": "video_fisheye_cam2", "l2": 5.631693, "mean_abs": 0.117135, "relative": 0.899016 }, { "name": "video_fisheye_cam3", "l2": 5.623136, "mean_abs": 0.115621, "relative": 0.941526 }, { "name": "video_stereo_left", "l2": 5.701807, "mean_abs": 0.126334, "relative": 0.926555 }, { "name": "video_stereo_right", "l2": 6.667311, "mean_abs": 0.149061, "relative": 0.939951 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.011147, "mean_abs": 0.079948, "relative": 0.719759 }, { "name": "caption_objects_interaction_text", "l2": 1.90567, "mean_abs": 0.016548, "relative": 0.875038 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 760, "start_frame": 3800, "end_frame": 3819, "center_frame": 3809, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.852489, "mean_abs": 0.261383, "relative": 0.915454 }, { "name": "hand_right_joints", "l2": 11.09288, "mean_abs": 0.288239, "relative": 0.874416 }, { "name": "body_joints", "l2": 6.153326, "mean_abs": 0.087422, "relative": 0.831125 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081672, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.026066, "mean_abs": 0.00361, "relative": 0.061511 }, { "name": "camera_rotation_matrix", "l2": 2.999929, "mean_abs": 0.202918, "relative": 0.92123 }, { "name": "imu_accel_gyro", "l2": 17.079294, "mean_abs": 1.064291, "relative": 0.921487 }, { "name": "depth_confidence", "l2": 10.242262, "mean_abs": 0.189201, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.139173, "mean_abs": 0.107738, "relative": 0.930594 }, { "name": "video_fisheye_cam1", "l2": 5.406563, "mean_abs": 0.114418, "relative": 0.893389 }, { "name": "video_fisheye_cam2", "l2": 5.615827, "mean_abs": 0.115453, "relative": 0.896483 }, { "name": "video_fisheye_cam3", "l2": 5.638287, "mean_abs": 0.114591, "relative": 0.944063 }, { "name": "video_stereo_left", "l2": 5.71439, "mean_abs": 0.124116, "relative": 0.9286 }, { "name": "video_stereo_right", "l2": 6.684092, "mean_abs": 0.146794, "relative": 0.942317 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.074351, "mean_abs": 0.080159, "relative": 0.742379 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 761, "start_frame": 3805, "end_frame": 3824, "center_frame": 3814, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.85823, "mean_abs": 0.262022, "relative": 0.915938 }, { "name": "hand_right_joints", "l2": 11.095389, "mean_abs": 0.288931, "relative": 0.874614 }, { "name": "body_joints", "l2": 6.148373, "mean_abs": 0.086967, "relative": 0.830456 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08167, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01439, "mean_abs": 0.002101, "relative": 0.033958 }, { "name": "camera_rotation_matrix", "l2": 3.000381, "mean_abs": 0.201362, "relative": 0.921368 }, { "name": "imu_accel_gyro", "l2": 17.046394, "mean_abs": 1.043662, "relative": 0.919712 }, { "name": "depth_confidence", "l2": 10.243818, "mean_abs": 0.187491, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.16646, "mean_abs": 0.10848, "relative": 0.935535 }, { "name": "video_fisheye_cam1", "l2": 5.412448, "mean_abs": 0.114226, "relative": 0.894361 }, { "name": "video_fisheye_cam2", "l2": 5.607751, "mean_abs": 0.114805, "relative": 0.895194 }, { "name": "video_fisheye_cam3", "l2": 5.636364, "mean_abs": 0.114349, "relative": 0.943741 }, { "name": "video_stereo_left", "l2": 5.718811, "mean_abs": 0.123377, "relative": 0.929318 }, { "name": "video_stereo_right", "l2": 6.689847, "mean_abs": 0.146106, "relative": 0.943128 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.143259, "mean_abs": 0.084873, "relative": 0.76704 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 762, "start_frame": 3810, "end_frame": 3829, "center_frame": 3819, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.86599, "mean_abs": 0.261758, "relative": 0.916593 }, { "name": "hand_right_joints", "l2": 11.100713, "mean_abs": 0.288289, "relative": 0.875033 }, { "name": "body_joints", "l2": 6.142483, "mean_abs": 0.086885, "relative": 0.82966 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081669, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009637, "mean_abs": 0.001477, "relative": 0.022742 }, { "name": "camera_rotation_matrix", "l2": 3.000242, "mean_abs": 0.200119, "relative": 0.921326 }, { "name": "imu_accel_gyro", "l2": 17.048201, "mean_abs": 1.041476, "relative": 0.919809 }, { "name": "depth_confidence", "l2": 10.258235, "mean_abs": 0.186916, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.179682, "mean_abs": 0.108573, "relative": 0.93793 }, { "name": "video_fisheye_cam1", "l2": 5.412157, "mean_abs": 0.113967, "relative": 0.894313 }, { "name": "video_fisheye_cam2", "l2": 5.598799, "mean_abs": 0.114429, "relative": 0.893765 }, { "name": "video_fisheye_cam3", "l2": 5.637986, "mean_abs": 0.114231, "relative": 0.944013 }, { "name": "video_stereo_left", "l2": 5.711298, "mean_abs": 0.122719, "relative": 0.928097 }, { "name": "video_stereo_right", "l2": 6.684151, "mean_abs": 0.145249, "relative": 0.942325 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.048953, "mean_abs": 0.080271, "relative": 0.73329 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 763, "start_frame": 3815, "end_frame": 3834, "center_frame": 3824, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.861101, "mean_abs": 0.26176, "relative": 0.91618 }, { "name": "hand_right_joints", "l2": 11.095174, "mean_abs": 0.289203, "relative": 0.874597 }, { "name": "body_joints", "l2": 6.136273, "mean_abs": 0.086776, "relative": 0.828821 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081669, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004947, "mean_abs": 0.000628, "relative": 0.011674 }, { "name": "camera_rotation_matrix", "l2": 2.999984, "mean_abs": 0.199725, "relative": 0.921246 }, { "name": "imu_accel_gyro", "l2": 17.040789, "mean_abs": 1.043537, "relative": 0.919409 }, { "name": "depth_confidence", "l2": 10.214362, "mean_abs": 0.185398, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.207755, "mean_abs": 0.1085, "relative": 0.943013 }, { "name": "video_fisheye_cam1", "l2": 5.412743, "mean_abs": 0.11385, "relative": 0.89441 }, { "name": "video_fisheye_cam2", "l2": 5.593618, "mean_abs": 0.114307, "relative": 0.892938 }, { "name": "video_fisheye_cam3", "l2": 5.630154, "mean_abs": 0.113979, "relative": 0.942701 }, { "name": "video_stereo_left", "l2": 5.711106, "mean_abs": 0.122814, "relative": 0.928066 }, { "name": "video_stereo_right", "l2": 6.682755, "mean_abs": 0.145166, "relative": 0.942128 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.052785, "mean_abs": 0.08029, "relative": 0.734661 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 764, "start_frame": 3820, "end_frame": 3839, "center_frame": 3829, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.854842, "mean_abs": 0.260687, "relative": 0.915652 }, { "name": "hand_right_joints", "l2": 11.076161, "mean_abs": 0.287662, "relative": 0.873098 }, { "name": "body_joints", "l2": 6.136346, "mean_abs": 0.08637, "relative": 0.828831 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081669, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00535, "mean_abs": 0.000808, "relative": 0.012625 }, { "name": "camera_rotation_matrix", "l2": 3.000083, "mean_abs": 0.199321, "relative": 0.921277 }, { "name": "imu_accel_gyro", "l2": 17.024162, "mean_abs": 1.043298, "relative": 0.918512 }, { "name": "depth_confidence", "l2": 10.176015, "mean_abs": 0.185292, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.222248, "mean_abs": 0.108803, "relative": 0.945637 }, { "name": "video_fisheye_cam1", "l2": 5.413588, "mean_abs": 0.114175, "relative": 0.89455 }, { "name": "video_fisheye_cam2", "l2": 5.590231, "mean_abs": 0.114352, "relative": 0.892397 }, { "name": "video_fisheye_cam3", "l2": 5.636435, "mean_abs": 0.114846, "relative": 0.943753 }, { "name": "video_stereo_left", "l2": 5.703341, "mean_abs": 0.122901, "relative": 0.926804 }, { "name": "video_stereo_right", "l2": 6.677481, "mean_abs": 0.14561, "relative": 0.941385 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.137032, "mean_abs": 0.084397, "relative": 0.764812 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 765, "start_frame": 3825, "end_frame": 3844, "center_frame": 3834, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.85672, "mean_abs": 0.260498, "relative": 0.915811 }, { "name": "hand_right_joints", "l2": 11.076063, "mean_abs": 0.287804, "relative": 0.87309 }, { "name": "body_joints", "l2": 6.130654, "mean_abs": 0.08597, "relative": 0.828062 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081668, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007965, "mean_abs": 0.00112, "relative": 0.018796 }, { "name": "camera_rotation_matrix", "l2": 3.00027, "mean_abs": 0.199225, "relative": 0.921334 }, { "name": "imu_accel_gyro", "l2": 17.019104, "mean_abs": 1.0463, "relative": 0.918239 }, { "name": "depth_confidence", "l2": 10.150352, "mean_abs": 0.18454, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.247427, "mean_abs": 0.108434, "relative": 0.950197 }, { "name": "video_fisheye_cam1", "l2": 5.436625, "mean_abs": 0.113993, "relative": 0.898356 }, { "name": "video_fisheye_cam2", "l2": 5.592769, "mean_abs": 0.114245, "relative": 0.892802 }, { "name": "video_fisheye_cam3", "l2": 5.679603, "mean_abs": 0.114447, "relative": 0.950981 }, { "name": "video_stereo_left", "l2": 5.709226, "mean_abs": 0.122771, "relative": 0.927761 }, { "name": "video_stereo_right", "l2": 6.697603, "mean_abs": 0.145161, "relative": 0.944222 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.08442, "mean_abs": 0.081539, "relative": 0.745983 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 766, "start_frame": 3830, "end_frame": 3849, "center_frame": 3839, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.855169, "mean_abs": 0.260972, "relative": 0.91568 }, { "name": "hand_right_joints", "l2": 11.068201, "mean_abs": 0.288909, "relative": 0.87247 }, { "name": "body_joints", "l2": 6.145185, "mean_abs": 0.086511, "relative": 0.830025 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081669, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009204, "mean_abs": 0.001211, "relative": 0.021719 }, { "name": "camera_rotation_matrix", "l2": 3.000199, "mean_abs": 0.198982, "relative": 0.921312 }, { "name": "imu_accel_gyro", "l2": 17.040682, "mean_abs": 1.046388, "relative": 0.919403 }, { "name": "depth_confidence", "l2": 10.091206, "mean_abs": 0.183414, "relative": 0.000335 }, { "name": "video_fisheye_cam0", "l2": 5.250582, "mean_abs": 0.108394, "relative": 0.950768 }, { "name": "video_fisheye_cam1", "l2": 5.441381, "mean_abs": 0.114102, "relative": 0.899142 }, { "name": "video_fisheye_cam2", "l2": 5.589307, "mean_abs": 0.114088, "relative": 0.89225 }, { "name": "video_fisheye_cam3", "l2": 5.679728, "mean_abs": 0.114658, "relative": 0.951002 }, { "name": "video_stereo_left", "l2": 5.706499, "mean_abs": 0.122369, "relative": 0.927317 }, { "name": "video_stereo_right", "l2": 6.699737, "mean_abs": 0.144832, "relative": 0.944522 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.15806, "mean_abs": 0.084867, "relative": 0.772337 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 767, "start_frame": 3835, "end_frame": 3854, "center_frame": 3844, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.8601, "mean_abs": 0.260958, "relative": 0.916096 }, { "name": "hand_right_joints", "l2": 11.05911, "mean_abs": 0.288768, "relative": 0.871754 }, { "name": "body_joints", "l2": 6.146337, "mean_abs": 0.086579, "relative": 0.830181 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08167, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010814, "mean_abs": 0.001437, "relative": 0.025521 }, { "name": "camera_rotation_matrix", "l2": 3.00017, "mean_abs": 0.198673, "relative": 0.921303 }, { "name": "imu_accel_gyro", "l2": 17.039192, "mean_abs": 1.046889, "relative": 0.919323 }, { "name": "depth_confidence", "l2": 10.050042, "mean_abs": 0.182982, "relative": 0.000334 }, { "name": "video_fisheye_cam0", "l2": 5.254504, "mean_abs": 0.108496, "relative": 0.951478 }, { "name": "video_fisheye_cam1", "l2": 5.44207, "mean_abs": 0.11425, "relative": 0.899256 }, { "name": "video_fisheye_cam2", "l2": 5.585704, "mean_abs": 0.114025, "relative": 0.891675 }, { "name": "video_fisheye_cam3", "l2": 5.678723, "mean_abs": 0.114699, "relative": 0.950834 }, { "name": "video_stereo_left", "l2": 5.697872, "mean_abs": 0.122203, "relative": 0.925915 }, { "name": "video_stereo_right", "l2": 6.691464, "mean_abs": 0.144843, "relative": 0.943356 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.181473, "mean_abs": 0.083159, "relative": 0.780717 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 768, "start_frame": 3840, "end_frame": 3859, "center_frame": 3849, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.847587, "mean_abs": 0.261201, "relative": 0.91504 }, { "name": "hand_right_joints", "l2": 11.050515, "mean_abs": 0.289868, "relative": 0.871076 }, { "name": "body_joints", "l2": 6.151436, "mean_abs": 0.087383, "relative": 0.830869 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081671, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021603, "mean_abs": 0.002679, "relative": 0.05098 }, { "name": "camera_rotation_matrix", "l2": 3.000463, "mean_abs": 0.199522, "relative": 0.921393 }, { "name": "imu_accel_gyro", "l2": 17.038694, "mean_abs": 1.057124, "relative": 0.919296 }, { "name": "depth_confidence", "l2": 10.025464, "mean_abs": 0.183732, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.261384, "mean_abs": 0.108849, "relative": 0.952724 }, { "name": "video_fisheye_cam1", "l2": 5.438299, "mean_abs": 0.11506, "relative": 0.898633 }, { "name": "video_fisheye_cam2", "l2": 5.579317, "mean_abs": 0.114351, "relative": 0.890655 }, { "name": "video_fisheye_cam3", "l2": 5.675667, "mean_abs": 0.114777, "relative": 0.950322 }, { "name": "video_stereo_left", "l2": 5.678427, "mean_abs": 0.12286, "relative": 0.922756 }, { "name": "video_stereo_right", "l2": 6.654371, "mean_abs": 0.145676, "relative": 0.938127 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.183654, "mean_abs": 0.085067, "relative": 0.781497 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 769, "start_frame": 3845, "end_frame": 3864, "center_frame": 3854, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.917188, "mean_abs": 0.264569, "relative": 0.920911 }, { "name": "hand_right_joints", "l2": 10.99696, "mean_abs": 0.288649, "relative": 0.866855 }, { "name": "body_joints", "l2": 6.183467, "mean_abs": 0.091507, "relative": 0.835196 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.063509, "mean_abs": 0.007813, "relative": 0.149872 }, { "name": "camera_rotation_matrix", "l2": 3.003622, "mean_abs": 0.202835, "relative": 0.922364 }, { "name": "imu_accel_gyro", "l2": 16.966661, "mean_abs": 1.079222, "relative": 0.91541 }, { "name": "depth_confidence", "l2": 9.978696, "mean_abs": 0.18556, "relative": 0.000331 }, { "name": "video_fisheye_cam0", "l2": 5.270391, "mean_abs": 0.110019, "relative": 0.954355 }, { "name": "video_fisheye_cam1", "l2": 5.426304, "mean_abs": 0.116838, "relative": 0.896651 }, { "name": "video_fisheye_cam2", "l2": 5.563132, "mean_abs": 0.116048, "relative": 0.888071 }, { "name": "video_fisheye_cam3", "l2": 5.658239, "mean_abs": 0.115698, "relative": 0.947404 }, { "name": "video_stereo_left", "l2": 5.632246, "mean_abs": 0.12545, "relative": 0.915251 }, { "name": "video_stereo_right", "l2": 6.587837, "mean_abs": 0.148366, "relative": 0.928747 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.129512, "mean_abs": 0.083183, "relative": 0.76212 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 770, "start_frame": 3850, "end_frame": 3869, "center_frame": 3859, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.988522, "mean_abs": 0.266475, "relative": 0.926929 }, { "name": "hand_right_joints", "l2": 10.866723, "mean_abs": 0.287784, "relative": 0.856588 }, { "name": "body_joints", "l2": 6.210015, "mean_abs": 0.096693, "relative": 0.838782 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.123234, "mean_abs": 0.015935, "relative": 0.290816 }, { "name": "camera_rotation_matrix", "l2": 3.012205, "mean_abs": 0.20696, "relative": 0.924999 }, { "name": "imu_accel_gyro", "l2": 16.997765, "mean_abs": 1.093888, "relative": 0.917088 }, { "name": "depth_confidence", "l2": 9.843699, "mean_abs": 0.186165, "relative": 0.000327 }, { "name": "video_fisheye_cam0", "l2": 5.281352, "mean_abs": 0.111391, "relative": 0.95634 }, { "name": "video_fisheye_cam1", "l2": 5.401293, "mean_abs": 0.117872, "relative": 0.892518 }, { "name": "video_fisheye_cam2", "l2": 5.533674, "mean_abs": 0.117636, "relative": 0.883369 }, { "name": "video_fisheye_cam3", "l2": 5.652141, "mean_abs": 0.117095, "relative": 0.946383 }, { "name": "video_stereo_left", "l2": 5.575177, "mean_abs": 0.12736, "relative": 0.905977 }, { "name": "video_stereo_right", "l2": 6.496227, "mean_abs": 0.151025, "relative": 0.915832 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.092371, "mean_abs": 0.080738, "relative": 0.748828 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 771, "start_frame": 3855, "end_frame": 3874, "center_frame": 3864, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.074705, "mean_abs": 0.267241, "relative": 0.934199 }, { "name": "hand_right_joints", "l2": 10.713218, "mean_abs": 0.285245, "relative": 0.844488 }, { "name": "body_joints", "l2": 6.221567, "mean_abs": 0.099552, "relative": 0.840342 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.171943, "mean_abs": 0.023287, "relative": 0.405762 }, { "name": "camera_rotation_matrix", "l2": 3.020388, "mean_abs": 0.208063, "relative": 0.927512 }, { "name": "imu_accel_gyro", "l2": 17.031706, "mean_abs": 1.118241, "relative": 0.918919 }, { "name": "depth_confidence", "l2": 9.692208, "mean_abs": 0.185907, "relative": 0.000322 }, { "name": "video_fisheye_cam0", "l2": 5.294603, "mean_abs": 0.112256, "relative": 0.958739 }, { "name": "video_fisheye_cam1", "l2": 5.367083, "mean_abs": 0.117899, "relative": 0.886865 }, { "name": "video_fisheye_cam2", "l2": 5.524237, "mean_abs": 0.118504, "relative": 0.881862 }, { "name": "video_fisheye_cam3", "l2": 5.65362, "mean_abs": 0.1183, "relative": 0.94663 }, { "name": "video_stereo_left", "l2": 5.534415, "mean_abs": 0.128179, "relative": 0.899353 }, { "name": "video_stereo_right", "l2": 6.43619, "mean_abs": 0.151278, "relative": 0.907368 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.003119, "mean_abs": 0.079223, "relative": 0.716886 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 772, "start_frame": 3860, "end_frame": 3879, "center_frame": 3869, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.130922, "mean_abs": 0.264312, "relative": 0.938941 }, { "name": "hand_right_joints", "l2": 10.614873, "mean_abs": 0.276993, "relative": 0.836736 }, { "name": "body_joints", "l2": 6.215887, "mean_abs": 0.09944, "relative": 0.839575 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.185709, "mean_abs": 0.025962, "relative": 0.438246 }, { "name": "camera_rotation_matrix", "l2": 3.020261, "mean_abs": 0.204933, "relative": 0.927473 }, { "name": "imu_accel_gyro", "l2": 17.151976, "mean_abs": 1.148921, "relative": 0.925408 }, { "name": "depth_confidence", "l2": 9.530793, "mean_abs": 0.18366, "relative": 0.000316 }, { "name": "video_fisheye_cam0", "l2": 5.302562, "mean_abs": 0.112135, "relative": 0.960181 }, { "name": "video_fisheye_cam1", "l2": 5.335963, "mean_abs": 0.117355, "relative": 0.881723 }, { "name": "video_fisheye_cam2", "l2": 5.483135, "mean_abs": 0.117681, "relative": 0.875301 }, { "name": "video_fisheye_cam3", "l2": 5.662961, "mean_abs": 0.118441, "relative": 0.948195 }, { "name": "video_stereo_left", "l2": 5.457917, "mean_abs": 0.125979, "relative": 0.886922 }, { "name": "video_stereo_right", "l2": 6.355658, "mean_abs": 0.149082, "relative": 0.896014 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.10459, "mean_abs": 0.082064, "relative": 0.753201 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011384, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 773, "start_frame": 3865, "end_frame": 3884, "center_frame": 3874, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.206966, "mean_abs": 0.255828, "relative": 0.945355 }, { "name": "hand_right_joints", "l2": 10.456517, "mean_abs": 0.264096, "relative": 0.824253 }, { "name": "body_joints", "l2": 6.152519, "mean_abs": 0.095046, "relative": 0.831016 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.145076, "mean_abs": 0.020666, "relative": 0.34236 }, { "name": "camera_rotation_matrix", "l2": 3.009782, "mean_abs": 0.195966, "relative": 0.924255 }, { "name": "imu_accel_gyro", "l2": 17.140121, "mean_abs": 1.130318, "relative": 0.924769 }, { "name": "depth_confidence", "l2": 9.332117, "mean_abs": 0.178609, "relative": 0.00031 }, { "name": "video_fisheye_cam0", "l2": 5.289035, "mean_abs": 0.110922, "relative": 0.957731 }, { "name": "video_fisheye_cam1", "l2": 5.277178, "mean_abs": 0.114995, "relative": 0.872009 }, { "name": "video_fisheye_cam2", "l2": 5.411058, "mean_abs": 0.114322, "relative": 0.863795 }, { "name": "video_fisheye_cam3", "l2": 5.652216, "mean_abs": 0.116913, "relative": 0.946395 }, { "name": "video_stereo_left", "l2": 5.333061, "mean_abs": 0.120489, "relative": 0.866633 }, { "name": "video_stereo_right", "l2": 6.215316, "mean_abs": 0.143671, "relative": 0.876229 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.12556, "mean_abs": 0.085269, "relative": 0.760706 }, { "name": "caption_objects_interaction_text", "l2": 1.882361, "mean_abs": 0.015489, "relative": 0.864335 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 774, "start_frame": 3870, "end_frame": 3889, "center_frame": 3879, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.273526, "mean_abs": 0.253671, "relative": 0.95097 }, { "name": "hand_right_joints", "l2": 10.346931, "mean_abs": 0.258839, "relative": 0.815615 }, { "name": "body_joints", "l2": 6.113342, "mean_abs": 0.091699, "relative": 0.825724 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.084151, "mean_abs": 0.01193, "relative": 0.198585 }, { "name": "camera_rotation_matrix", "l2": 3.002545, "mean_abs": 0.18585, "relative": 0.922033 }, { "name": "imu_accel_gyro", "l2": 17.160236, "mean_abs": 1.093208, "relative": 0.925854 }, { "name": "depth_confidence", "l2": 9.152363, "mean_abs": 0.173242, "relative": 0.000304 }, { "name": "video_fisheye_cam0", "l2": 5.278145, "mean_abs": 0.10957, "relative": 0.955759 }, { "name": "video_fisheye_cam1", "l2": 5.221712, "mean_abs": 0.111537, "relative": 0.862844 }, { "name": "video_fisheye_cam2", "l2": 5.35954, "mean_abs": 0.110324, "relative": 0.855571 }, { "name": "video_fisheye_cam3", "l2": 5.64156, "mean_abs": 0.115457, "relative": 0.944611 }, { "name": "video_stereo_left", "l2": 5.259184, "mean_abs": 0.115813, "relative": 0.854628 }, { "name": "video_stereo_right", "l2": 6.095047, "mean_abs": 0.137348, "relative": 0.859274 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.276017, "mean_abs": 0.090584, "relative": 0.814552 }, { "name": "caption_objects_interaction_text", "l2": 1.882361, "mean_abs": 0.015523, "relative": 0.864335 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 775, "start_frame": 3875, "end_frame": 3894, "center_frame": 3884, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.308758, "mean_abs": 0.256526, "relative": 0.953942 }, { "name": "hand_right_joints", "l2": 10.373837, "mean_abs": 0.259378, "relative": 0.817736 }, { "name": "body_joints", "l2": 6.127039, "mean_abs": 0.092435, "relative": 0.827574 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.031535, "mean_abs": 0.004358, "relative": 0.074419 }, { "name": "camera_rotation_matrix", "l2": 3.000492, "mean_abs": 0.17747, "relative": 0.921402 }, { "name": "imu_accel_gyro", "l2": 17.189615, "mean_abs": 1.053151, "relative": 0.927439 }, { "name": "depth_confidence", "l2": 9.032493, "mean_abs": 0.168317, "relative": 0.0003 }, { "name": "video_fisheye_cam0", "l2": 5.271921, "mean_abs": 0.108689, "relative": 0.954632 }, { "name": "video_fisheye_cam1", "l2": 5.187951, "mean_abs": 0.109013, "relative": 0.857265 }, { "name": "video_fisheye_cam2", "l2": 5.374999, "mean_abs": 0.109482, "relative": 0.858039 }, { "name": "video_fisheye_cam3", "l2": 5.603117, "mean_abs": 0.114112, "relative": 0.938174 }, { "name": "video_stereo_left", "l2": 5.283641, "mean_abs": 0.114589, "relative": 0.858602 }, { "name": "video_stereo_right", "l2": 6.074392, "mean_abs": 0.133602, "relative": 0.856362 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.255397, "mean_abs": 0.088872, "relative": 0.807173 }, { "name": "caption_objects_interaction_text", "l2": 1.882361, "mean_abs": 0.015174, "relative": 0.864335 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 776, "start_frame": 3880, "end_frame": 3899, "center_frame": 3889, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.308805, "mean_abs": 0.258246, "relative": 0.953946 }, { "name": "hand_right_joints", "l2": 10.386641, "mean_abs": 0.262653, "relative": 0.818745 }, { "name": "body_joints", "l2": 6.168238, "mean_abs": 0.095065, "relative": 0.833139 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.024488, "mean_abs": 0.003617, "relative": 0.057789 }, { "name": "camera_rotation_matrix", "l2": 3.000537, "mean_abs": 0.178142, "relative": 0.921416 }, { "name": "imu_accel_gyro", "l2": 17.176619, "mean_abs": 1.039133, "relative": 0.926738 }, { "name": "depth_confidence", "l2": 9.038524, "mean_abs": 0.169686, "relative": 0.0003 }, { "name": "video_fisheye_cam0", "l2": 5.269342, "mean_abs": 0.109037, "relative": 0.954165 }, { "name": "video_fisheye_cam1", "l2": 5.196373, "mean_abs": 0.109438, "relative": 0.858657 }, { "name": "video_fisheye_cam2", "l2": 5.422194, "mean_abs": 0.111023, "relative": 0.865573 }, { "name": "video_fisheye_cam3", "l2": 5.57556, "mean_abs": 0.113813, "relative": 0.93356 }, { "name": "video_stereo_left", "l2": 5.336158, "mean_abs": 0.116605, "relative": 0.867136 }, { "name": "video_stereo_right", "l2": 6.119889, "mean_abs": 0.135278, "relative": 0.862776 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.289612, "mean_abs": 0.092283, "relative": 0.819418 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009498, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 777, "start_frame": 3885, "end_frame": 3904, "center_frame": 3894, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.981956, "mean_abs": 0.268894, "relative": 0.926375 }, { "name": "hand_right_joints", "l2": 10.413396, "mean_abs": 0.263949, "relative": 0.820854 }, { "name": "body_joints", "l2": 6.176763, "mean_abs": 0.095404, "relative": 0.83429 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.071118, "mean_abs": 0.010162, "relative": 0.167829 }, { "name": "camera_rotation_matrix", "l2": 3.003117, "mean_abs": 0.182278, "relative": 0.922209 }, { "name": "imu_accel_gyro", "l2": 17.16485, "mean_abs": 1.061536, "relative": 0.926103 }, { "name": "depth_confidence", "l2": 9.240227, "mean_abs": 0.176401, "relative": 0.000307 }, { "name": "video_fisheye_cam0", "l2": 5.255799, "mean_abs": 0.109053, "relative": 0.951713 }, { "name": "video_fisheye_cam1", "l2": 5.249482, "mean_abs": 0.111844, "relative": 0.867433 }, { "name": "video_fisheye_cam2", "l2": 5.465718, "mean_abs": 0.112729, "relative": 0.872521 }, { "name": "video_fisheye_cam3", "l2": 5.515582, "mean_abs": 0.113188, "relative": 0.923518 }, { "name": "video_stereo_left", "l2": 5.41879, "mean_abs": 0.12074, "relative": 0.880564 }, { "name": "video_stereo_right", "l2": 6.201177, "mean_abs": 0.14069, "relative": 0.874236 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.309788, "mean_abs": 0.091508, "relative": 0.826638 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009498, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 778, "start_frame": 3890, "end_frame": 3909, "center_frame": 3899, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.866567, "mean_abs": 0.277059, "relative": 0.916641 }, { "name": "hand_right_joints", "l2": 10.421992, "mean_abs": 0.263157, "relative": 0.821532 }, { "name": "body_joints", "l2": 6.147725, "mean_abs": 0.098386, "relative": 0.830368 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.132698, "mean_abs": 0.019185, "relative": 0.313149 }, { "name": "camera_rotation_matrix", "l2": 3.00756, "mean_abs": 0.191152, "relative": 0.923573 }, { "name": "imu_accel_gyro", "l2": 17.053452, "mean_abs": 1.110945, "relative": 0.920092 }, { "name": "depth_confidence", "l2": 9.620718, "mean_abs": 0.187013, "relative": 0.000319 }, { "name": "video_fisheye_cam0", "l2": 5.251329, "mean_abs": 0.110094, "relative": 0.950903 }, { "name": "video_fisheye_cam1", "l2": 5.317559, "mean_abs": 0.114959, "relative": 0.878682 }, { "name": "video_fisheye_cam2", "l2": 5.556991, "mean_abs": 0.115873, "relative": 0.887091 }, { "name": "video_fisheye_cam3", "l2": 5.468445, "mean_abs": 0.114088, "relative": 0.915625 }, { "name": "video_stereo_left", "l2": 5.548657, "mean_abs": 0.126804, "relative": 0.901668 }, { "name": "video_stereo_right", "l2": 6.347859, "mean_abs": 0.148123, "relative": 0.894915 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.128204, "mean_abs": 0.082774, "relative": 0.761652 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009498, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 779, "start_frame": 3895, "end_frame": 3914, "center_frame": 3904, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.748192, "mean_abs": 0.275052, "relative": 0.906656 }, { "name": "hand_right_joints", "l2": 10.411426, "mean_abs": 0.264347, "relative": 0.820699 }, { "name": "body_joints", "l2": 6.081028, "mean_abs": 0.100169, "relative": 0.821359 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.173396, "mean_abs": 0.025425, "relative": 0.40919 }, { "name": "camera_rotation_matrix", "l2": 3.01173, "mean_abs": 0.200852, "relative": 0.924853 }, { "name": "imu_accel_gyro", "l2": 17.208519, "mean_abs": 1.165388, "relative": 0.928459 }, { "name": "depth_confidence", "l2": 10.005807, "mean_abs": 0.195605, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.239254, "mean_abs": 0.110457, "relative": 0.948717 }, { "name": "video_fisheye_cam1", "l2": 5.387654, "mean_abs": 0.117192, "relative": 0.890264 }, { "name": "video_fisheye_cam2", "l2": 5.654173, "mean_abs": 0.118812, "relative": 0.902605 }, { "name": "video_fisheye_cam3", "l2": 5.391496, "mean_abs": 0.113032, "relative": 0.902741 }, { "name": "video_stereo_left", "l2": 5.636768, "mean_abs": 0.131187, "relative": 0.915986 }, { "name": "video_stereo_right", "l2": 6.519702, "mean_abs": 0.153941, "relative": 0.919141 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.208061, "mean_abs": 0.08574, "relative": 0.790232 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009498, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 780, "start_frame": 3900, "end_frame": 3919, "center_frame": 3909, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.623715, "mean_abs": 0.270113, "relative": 0.896156 }, { "name": "hand_right_joints", "l2": 10.503911, "mean_abs": 0.268447, "relative": 0.827989 }, { "name": "body_joints", "l2": 6.015939, "mean_abs": 0.100971, "relative": 0.812568 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.173186, "mean_abs": 0.026055, "relative": 0.408695 }, { "name": "camera_rotation_matrix", "l2": 3.008404, "mean_abs": 0.202413, "relative": 0.923832 }, { "name": "imu_accel_gyro", "l2": 17.36294, "mean_abs": 1.215631, "relative": 0.93679 }, { "name": "depth_confidence", "l2": 10.190211, "mean_abs": 0.197828, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.225826, "mean_abs": 0.109964, "relative": 0.946285 }, { "name": "video_fisheye_cam1", "l2": 5.436513, "mean_abs": 0.117921, "relative": 0.898338 }, { "name": "video_fisheye_cam2", "l2": 5.703711, "mean_abs": 0.119605, "relative": 0.910513 }, { "name": "video_fisheye_cam3", "l2": 5.349931, "mean_abs": 0.111729, "relative": 0.895781 }, { "name": "video_stereo_left", "l2": 5.649835, "mean_abs": 0.13133, "relative": 0.918109 }, { "name": "video_stereo_right", "l2": 6.581368, "mean_abs": 0.155166, "relative": 0.927835 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.168008, "mean_abs": 0.085152, "relative": 0.775897 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009498, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 781, "start_frame": 3905, "end_frame": 3924, "center_frame": 3914, "action": "Pour liquid from white bottle", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.226246, "mean_abs": 0.250877, "relative": 0.862627 }, { "name": "hand_right_joints", "l2": 11.182709, "mean_abs": 0.297066, "relative": 0.881497 }, { "name": "body_joints", "l2": 5.841607, "mean_abs": 0.097435, "relative": 0.789021 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.125128, "mean_abs": 0.018861, "relative": 0.295285 }, { "name": "camera_rotation_matrix", "l2": 3.003943, "mean_abs": 0.202766, "relative": 0.922462 }, { "name": "imu_accel_gyro", "l2": 17.185419, "mean_abs": 1.1616, "relative": 0.927213 }, { "name": "depth_confidence", "l2": 10.400134, "mean_abs": 0.19782, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.216019, "mean_abs": 0.109108, "relative": 0.94451 }, { "name": "video_fisheye_cam1", "l2": 5.489182, "mean_abs": 0.118142, "relative": 0.907041 }, { "name": "video_fisheye_cam2", "l2": 5.727435, "mean_abs": 0.119008, "relative": 0.9143 }, { "name": "video_fisheye_cam3", "l2": 5.342062, "mean_abs": 0.110606, "relative": 0.894464 }, { "name": "video_stereo_left", "l2": 5.597552, "mean_abs": 0.128065, "relative": 0.909613 }, { "name": "video_stereo_right", "l2": 6.602558, "mean_abs": 0.153374, "relative": 0.930822 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.069816, "mean_abs": 0.082083, "relative": 0.740756 }, { "name": "caption_objects_interaction_text", "l2": 1.9778, "mean_abs": 0.014979, "relative": 0.908158 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 782, "start_frame": 3910, "end_frame": 3929, "center_frame": 3919, "action": "", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.32924, "mean_abs": 0.253622, "relative": 0.871315 }, { "name": "hand_right_joints", "l2": 11.65929, "mean_abs": 0.314986, "relative": 0.919064 }, { "name": "body_joints", "l2": 5.764647, "mean_abs": 0.097718, "relative": 0.778626 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.054017, "mean_abs": 0.007873, "relative": 0.127474 }, { "name": "camera_rotation_matrix", "l2": 3.005195, "mean_abs": 0.209331, "relative": 0.922846 }, { "name": "imu_accel_gyro", "l2": 17.100266, "mean_abs": 1.120599, "relative": 0.922618 }, { "name": "depth_confidence", "l2": 10.615458, "mean_abs": 0.199681, "relative": 0.000352 }, { "name": "video_fisheye_cam0", "l2": 5.20906, "mean_abs": 0.109446, "relative": 0.943249 }, { "name": "video_fisheye_cam1", "l2": 5.509089, "mean_abs": 0.119277, "relative": 0.91033 }, { "name": "video_fisheye_cam2", "l2": 5.793663, "mean_abs": 0.120437, "relative": 0.924872 }, { "name": "video_fisheye_cam3", "l2": 5.315881, "mean_abs": 0.10968, "relative": 0.89008 }, { "name": "video_stereo_left", "l2": 5.598358, "mean_abs": 0.12725, "relative": 0.909744 }, { "name": "video_stereo_right", "l2": 6.646817, "mean_abs": 0.153785, "relative": 0.937062 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.156103, "mean_abs": 0.083523, "relative": 0.771637 }, { "name": "caption_objects_interaction_text", "l2": 1.9778, "mean_abs": 0.015076, "relative": 0.908158 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 783, "start_frame": 3915, "end_frame": 3934, "center_frame": 3924, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.368997, "mean_abs": 0.248537, "relative": 0.874669 }, { "name": "hand_right_joints", "l2": 11.827179, "mean_abs": 0.315858, "relative": 0.932298 }, { "name": "body_joints", "l2": 5.665493, "mean_abs": 0.095285, "relative": 0.765234 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.041956, "mean_abs": 0.005819, "relative": 0.09901 }, { "name": "camera_rotation_matrix", "l2": 3.006777, "mean_abs": 0.213469, "relative": 0.923332 }, { "name": "imu_accel_gyro", "l2": 17.282597, "mean_abs": 1.158254, "relative": 0.932456 }, { "name": "depth_confidence", "l2": 10.693026, "mean_abs": 0.201312, "relative": 0.000355 }, { "name": "video_fisheye_cam0", "l2": 5.214262, "mean_abs": 0.109982, "relative": 0.944191 }, { "name": "video_fisheye_cam1", "l2": 5.520692, "mean_abs": 0.119987, "relative": 0.912248 }, { "name": "video_fisheye_cam2", "l2": 5.859921, "mean_abs": 0.122083, "relative": 0.935449 }, { "name": "video_fisheye_cam3", "l2": 5.395479, "mean_abs": 0.112788, "relative": 0.903408 }, { "name": "video_stereo_left", "l2": 5.617561, "mean_abs": 0.128463, "relative": 0.912865 }, { "name": "video_stereo_right", "l2": 6.70687, "mean_abs": 0.155497, "relative": 0.945528 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.068091, "mean_abs": 0.079224, "relative": 0.740139 }, { "name": "caption_objects_interaction_text", "l2": 1.9778, "mean_abs": 0.014685, "relative": 0.908158 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 784, "start_frame": 3920, "end_frame": 3939, "center_frame": 3929, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.379043, "mean_abs": 0.234904, "relative": 0.875517 }, { "name": "hand_right_joints", "l2": 11.822565, "mean_abs": 0.298511, "relative": 0.931934 }, { "name": "body_joints", "l2": 5.536903, "mean_abs": 0.089227, "relative": 0.747865 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.107329, "mean_abs": 0.013687, "relative": 0.253282 }, { "name": "camera_rotation_matrix", "l2": 3.024466, "mean_abs": 0.22254, "relative": 0.928764 }, { "name": "imu_accel_gyro", "l2": 17.228497, "mean_abs": 1.178789, "relative": 0.929537 }, { "name": "depth_confidence", "l2": 10.703037, "mean_abs": 0.20374, "relative": 0.000355 }, { "name": "video_fisheye_cam0", "l2": 5.23449, "mean_abs": 0.111544, "relative": 0.947854 }, { "name": "video_fisheye_cam1", "l2": 5.521764, "mean_abs": 0.122822, "relative": 0.912425 }, { "name": "video_fisheye_cam2", "l2": 5.905833, "mean_abs": 0.124677, "relative": 0.942778 }, { "name": "video_fisheye_cam3", "l2": 5.466198, "mean_abs": 0.11544, "relative": 0.915249 }, { "name": "video_stereo_left", "l2": 5.687129, "mean_abs": 0.132629, "relative": 0.92417 }, { "name": "video_stereo_right", "l2": 6.786293, "mean_abs": 0.159279, "relative": 0.956725 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.034321, "mean_abs": 0.076395, "relative": 0.728053 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007732, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 785, "start_frame": 3925, "end_frame": 3944, "center_frame": 3934, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.477533, "mean_abs": 0.235667, "relative": 0.883825 }, { "name": "hand_right_joints", "l2": 11.618198, "mean_abs": 0.315091, "relative": 0.915825 }, { "name": "body_joints", "l2": 5.675951, "mean_abs": 0.093858, "relative": 0.766646 }, { "name": "body_contacts", "l2": 3.463892, "mean_abs": 0.081704, "relative": 0.999564 }, { "name": "camera_translation", "l2": 0.187881, "mean_abs": 0.0245, "relative": 0.443373 }, { "name": "camera_rotation_matrix", "l2": 3.033584, "mean_abs": 0.226964, "relative": 0.931565 }, { "name": "imu_accel_gyro", "l2": 17.224213, "mean_abs": 1.151204, "relative": 0.929306 }, { "name": "depth_confidence", "l2": 10.812217, "mean_abs": 0.205245, "relative": 0.000359 }, { "name": "video_fisheye_cam0", "l2": 5.246753, "mean_abs": 0.112267, "relative": 0.950075 }, { "name": "video_fisheye_cam1", "l2": 5.571242, "mean_abs": 0.125706, "relative": 0.920601 }, { "name": "video_fisheye_cam2", "l2": 5.934201, "mean_abs": 0.126398, "relative": 0.947307 }, { "name": "video_fisheye_cam3", "l2": 5.543972, "mean_abs": 0.117421, "relative": 0.928271 }, { "name": "video_stereo_left", "l2": 5.754893, "mean_abs": 0.134382, "relative": 0.935182 }, { "name": "video_stereo_right", "l2": 6.822782, "mean_abs": 0.16127, "relative": 0.961869 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.034525, "mean_abs": 0.077587, "relative": 0.728126 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007732, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 786, "start_frame": 3930, "end_frame": 3949, "center_frame": 3939, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.532058, "mean_abs": 0.242662, "relative": 0.888424 }, { "name": "hand_right_joints", "l2": 11.429364, "mean_abs": 0.318718, "relative": 0.90094 }, { "name": "body_joints", "l2": 5.760594, "mean_abs": 0.098102, "relative": 0.778079 }, { "name": "body_contacts", "l2": 3.463891, "mean_abs": 0.081691, "relative": 0.999564 }, { "name": "camera_translation", "l2": 0.256426, "mean_abs": 0.03342, "relative": 0.60513 }, { "name": "camera_rotation_matrix", "l2": 3.036873, "mean_abs": 0.233028, "relative": 0.932575 }, { "name": "imu_accel_gyro", "l2": 17.366339, "mean_abs": 1.195738, "relative": 0.936974 }, { "name": "depth_confidence", "l2": 10.894564, "mean_abs": 0.207403, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.230914, "mean_abs": 0.112131, "relative": 0.947207 }, { "name": "video_fisheye_cam1", "l2": 5.605687, "mean_abs": 0.126802, "relative": 0.926292 }, { "name": "video_fisheye_cam2", "l2": 5.951735, "mean_abs": 0.127554, "relative": 0.950106 }, { "name": "video_fisheye_cam3", "l2": 5.638199, "mean_abs": 0.119299, "relative": 0.944048 }, { "name": "video_stereo_left", "l2": 5.852437, "mean_abs": 0.13703, "relative": 0.951033 }, { "name": "video_stereo_right", "l2": 6.931504, "mean_abs": 0.164253, "relative": 0.977197 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.519607, "mean_abs": 0.104994, "relative": 0.901729 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007732, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 787, "start_frame": 3935, "end_frame": 3954, "center_frame": 3944, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.50926, "mean_abs": 0.241221, "relative": 0.886501 }, { "name": "hand_right_joints", "l2": 10.832058, "mean_abs": 0.295506, "relative": 0.853856 }, { "name": "body_joints", "l2": 5.654044, "mean_abs": 0.09617, "relative": 0.763687 }, { "name": "body_contacts", "l2": 3.463891, "mean_abs": 0.081688, "relative": 0.999564 }, { "name": "camera_translation", "l2": 0.275643, "mean_abs": 0.036088, "relative": 0.650478 }, { "name": "camera_rotation_matrix", "l2": 3.042866, "mean_abs": 0.238055, "relative": 0.934415 }, { "name": "imu_accel_gyro", "l2": 17.394781, "mean_abs": 1.204615, "relative": 0.938508 }, { "name": "depth_confidence", "l2": 10.952022, "mean_abs": 0.209124, "relative": 0.000363 }, { "name": "video_fisheye_cam0", "l2": 5.217925, "mean_abs": 0.111524, "relative": 0.944855 }, { "name": "video_fisheye_cam1", "l2": 5.596169, "mean_abs": 0.126603, "relative": 0.92472 }, { "name": "video_fisheye_cam2", "l2": 5.927716, "mean_abs": 0.127238, "relative": 0.946272 }, { "name": "video_fisheye_cam3", "l2": 5.694466, "mean_abs": 0.121156, "relative": 0.95347 }, { "name": "video_stereo_left", "l2": 5.887321, "mean_abs": 0.13712, "relative": 0.956701 }, { "name": "video_stereo_right", "l2": 6.996419, "mean_abs": 0.164795, "relative": 0.986348 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.358631, "mean_abs": 0.096544, "relative": 0.844119 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007732, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 788, "start_frame": 3940, "end_frame": 3959, "center_frame": 3949, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.483863, "mean_abs": 0.239587, "relative": 0.884359 }, { "name": "hand_right_joints", "l2": 9.832249, "mean_abs": 0.249999, "relative": 0.775044 }, { "name": "body_joints", "l2": 5.541266, "mean_abs": 0.090948, "relative": 0.748454 }, { "name": "body_contacts", "l2": 3.463891, "mean_abs": 0.081705, "relative": 0.999564 }, { "name": "camera_translation", "l2": 0.216993, "mean_abs": 0.028473, "relative": 0.512073 }, { "name": "camera_rotation_matrix", "l2": 3.02973, "mean_abs": 0.234283, "relative": 0.930381 }, { "name": "imu_accel_gyro", "l2": 17.361176, "mean_abs": 1.177889, "relative": 0.936695 }, { "name": "depth_confidence", "l2": 10.967484, "mean_abs": 0.207117, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.195302, "mean_abs": 0.110206, "relative": 0.940758 }, { "name": "video_fisheye_cam1", "l2": 5.605603, "mean_abs": 0.124229, "relative": 0.926279 }, { "name": "video_fisheye_cam2", "l2": 5.887683, "mean_abs": 0.12556, "relative": 0.939881 }, { "name": "video_fisheye_cam3", "l2": 5.753574, "mean_abs": 0.121813, "relative": 0.963367 }, { "name": "video_stereo_left", "l2": 5.846019, "mean_abs": 0.134655, "relative": 0.94999 }, { "name": "video_stereo_right", "l2": 7.038703, "mean_abs": 0.164717, "relative": 0.992309 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.576378, "mean_abs": 0.117139, "relative": 0.922047 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007732, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 789, "start_frame": 3945, "end_frame": 3964, "center_frame": 3954, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.449207, "mean_abs": 0.240249, "relative": 0.881435 }, { "name": "hand_right_joints", "l2": 9.790425, "mean_abs": 0.244509, "relative": 0.771747 }, { "name": "body_joints", "l2": 5.53447, "mean_abs": 0.088647, "relative": 0.747536 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.122537, "mean_abs": 0.016891, "relative": 0.28917 }, { "name": "camera_rotation_matrix", "l2": 3.020807, "mean_abs": 0.230193, "relative": 0.927641 }, { "name": "imu_accel_gyro", "l2": 17.326323, "mean_abs": 1.208768, "relative": 0.934815 }, { "name": "depth_confidence", "l2": 11.094066, "mean_abs": 0.209569, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.225842, "mean_abs": 0.111312, "relative": 0.946288 }, { "name": "video_fisheye_cam1", "l2": 5.617068, "mean_abs": 0.123591, "relative": 0.928173 }, { "name": "video_fisheye_cam2", "l2": 5.750217, "mean_abs": 0.12164, "relative": 0.917937 }, { "name": "video_fisheye_cam3", "l2": 5.871312, "mean_abs": 0.124316, "relative": 0.98308 }, { "name": "video_stereo_left", "l2": 5.725185, "mean_abs": 0.133053, "relative": 0.930354 }, { "name": "video_stereo_right", "l2": 6.998739, "mean_abs": 0.163207, "relative": 0.986675 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.574229, "mean_abs": 0.118, "relative": 0.921278 }, { "name": "caption_objects_interaction_text", "l2": 1.923994, "mean_abs": 0.012137, "relative": 0.883452 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 790, "start_frame": 3950, "end_frame": 3969, "center_frame": 3959, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.389596, "mean_abs": 0.237989, "relative": 0.876407 }, { "name": "hand_right_joints", "l2": 9.803851, "mean_abs": 0.241779, "relative": 0.772806 }, { "name": "body_joints", "l2": 5.510033, "mean_abs": 0.08813, "relative": 0.744236 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.067251, "mean_abs": 0.009955, "relative": 0.158702 }, { "name": "camera_rotation_matrix", "l2": 3.030106, "mean_abs": 0.242231, "relative": 0.930496 }, { "name": "imu_accel_gyro", "l2": 17.314245, "mean_abs": 1.208345, "relative": 0.934163 }, { "name": "depth_confidence", "l2": 11.266393, "mean_abs": 0.214831, "relative": 0.000374 }, { "name": "video_fisheye_cam0", "l2": 5.260066, "mean_abs": 0.114088, "relative": 0.952485 }, { "name": "video_fisheye_cam1", "l2": 5.68844, "mean_abs": 0.12827, "relative": 0.939967 }, { "name": "video_fisheye_cam2", "l2": 5.671644, "mean_abs": 0.123042, "relative": 0.905394 }, { "name": "video_fisheye_cam3", "l2": 5.886882, "mean_abs": 0.126153, "relative": 0.985687 }, { "name": "video_stereo_left", "l2": 5.680241, "mean_abs": 0.134382, "relative": 0.92305 }, { "name": "video_stereo_right", "l2": 6.983957, "mean_abs": 0.167825, "relative": 0.984591 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.327096, "mean_abs": 0.105658, "relative": 0.832833 }, { "name": "caption_objects_interaction_text", "l2": 1.923994, "mean_abs": 0.01231, "relative": 0.883452 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 791, "start_frame": 3955, "end_frame": 3974, "center_frame": 3964, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.332854, "mean_abs": 0.231696, "relative": 0.87162 }, { "name": "hand_right_joints", "l2": 9.778502, "mean_abs": 0.235866, "relative": 0.770808 }, { "name": "body_joints", "l2": 5.513612, "mean_abs": 0.088486, "relative": 0.744719 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.093857, "mean_abs": 0.011697, "relative": 0.22149 }, { "name": "camera_rotation_matrix", "l2": 3.036818, "mean_abs": 0.24482, "relative": 0.932557 }, { "name": "imu_accel_gyro", "l2": 17.262272, "mean_abs": 1.17188, "relative": 0.931359 }, { "name": "depth_confidence", "l2": 11.303561, "mean_abs": 0.215257, "relative": 0.000375 }, { "name": "video_fisheye_cam0", "l2": 5.266389, "mean_abs": 0.114387, "relative": 0.95363 }, { "name": "video_fisheye_cam1", "l2": 5.682937, "mean_abs": 0.128718, "relative": 0.939057 }, { "name": "video_fisheye_cam2", "l2": 5.63528, "mean_abs": 0.122759, "relative": 0.899589 }, { "name": "video_fisheye_cam3", "l2": 5.874059, "mean_abs": 0.126341, "relative": 0.98354 }, { "name": "video_stereo_left", "l2": 5.667346, "mean_abs": 0.134795, "relative": 0.920955 }, { "name": "video_stereo_right", "l2": 6.916101, "mean_abs": 0.166991, "relative": 0.975025 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.328691, "mean_abs": 0.104599, "relative": 0.833403 }, { "name": "caption_objects_interaction_text", "l2": 1.923994, "mean_abs": 0.012137, "relative": 0.883452 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 792, "start_frame": 3960, "end_frame": 3979, "center_frame": 3969, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.283092, "mean_abs": 0.227392, "relative": 0.867423 }, { "name": "hand_right_joints", "l2": 9.769372, "mean_abs": 0.231933, "relative": 0.770088 }, { "name": "body_joints", "l2": 5.540116, "mean_abs": 0.088369, "relative": 0.748299 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.108193, "mean_abs": 0.014186, "relative": 0.255321 }, { "name": "camera_rotation_matrix", "l2": 3.036493, "mean_abs": 0.241909, "relative": 0.932458 }, { "name": "imu_accel_gyro", "l2": 17.302258, "mean_abs": 1.17988, "relative": 0.933516 }, { "name": "depth_confidence", "l2": 11.313897, "mean_abs": 0.213584, "relative": 0.000375 }, { "name": "video_fisheye_cam0", "l2": 5.28812, "mean_abs": 0.114239, "relative": 0.957565 }, { "name": "video_fisheye_cam1", "l2": 5.671047, "mean_abs": 0.127565, "relative": 0.937093 }, { "name": "video_fisheye_cam2", "l2": 5.596243, "mean_abs": 0.121399, "relative": 0.893357 }, { "name": "video_fisheye_cam3", "l2": 5.855725, "mean_abs": 0.125505, "relative": 0.980471 }, { "name": "video_stereo_left", "l2": 5.624502, "mean_abs": 0.133317, "relative": 0.913993 }, { "name": "video_stereo_right", "l2": 6.854743, "mean_abs": 0.164565, "relative": 0.966375 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.187043, "mean_abs": 0.094734, "relative": 0.78271 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007732, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 793, "start_frame": 3965, "end_frame": 3984, "center_frame": 3974, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.189837, "mean_abs": 0.212852, "relative": 0.859556 }, { "name": "hand_right_joints", "l2": 9.69109, "mean_abs": 0.22151, "relative": 0.763917 }, { "name": "body_joints", "l2": 5.557948, "mean_abs": 0.088548, "relative": 0.750707 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.058148, "mean_abs": 0.008071, "relative": 0.137221 }, { "name": "camera_rotation_matrix", "l2": 3.005137, "mean_abs": 0.223748, "relative": 0.922829 }, { "name": "imu_accel_gyro", "l2": 17.284721, "mean_abs": 1.164729, "relative": 0.93257 }, { "name": "depth_confidence", "l2": 11.423343, "mean_abs": 0.206168, "relative": 0.000379 }, { "name": "video_fisheye_cam0", "l2": 5.285921, "mean_abs": 0.110788, "relative": 0.957167 }, { "name": "video_fisheye_cam1", "l2": 5.602963, "mean_abs": 0.121134, "relative": 0.925842 }, { "name": "video_fisheye_cam2", "l2": 5.433387, "mean_abs": 0.113846, "relative": 0.867359 }, { "name": "video_fisheye_cam3", "l2": 5.75963, "mean_abs": 0.119446, "relative": 0.964381 }, { "name": "video_stereo_left", "l2": 5.328788, "mean_abs": 0.121286, "relative": 0.865939 }, { "name": "video_stereo_right", "l2": 6.539986, "mean_abs": 0.149109, "relative": 0.922001 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.200836, "mean_abs": 0.09359, "relative": 0.787646 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007732, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 794, "start_frame": 3970, "end_frame": 3989, "center_frame": 3979, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.197114, "mean_abs": 0.215715, "relative": 0.86017 }, { "name": "hand_right_joints", "l2": 9.702951, "mean_abs": 0.223152, "relative": 0.764852 }, { "name": "body_joints", "l2": 5.569072, "mean_abs": 0.088957, "relative": 0.75221 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014475, "mean_abs": 0.002177, "relative": 0.034159 }, { "name": "camera_rotation_matrix", "l2": 3.00065, "mean_abs": 0.21793, "relative": 0.921451 }, { "name": "imu_accel_gyro", "l2": 17.104614, "mean_abs": 1.076266, "relative": 0.922853 }, { "name": "depth_confidence", "l2": 11.370912, "mean_abs": 0.204254, "relative": 0.000377 }, { "name": "video_fisheye_cam0", "l2": 5.287085, "mean_abs": 0.10986, "relative": 0.957378 }, { "name": "video_fisheye_cam1", "l2": 5.5559, "mean_abs": 0.117808, "relative": 0.918065 }, { "name": "video_fisheye_cam2", "l2": 5.408546, "mean_abs": 0.111263, "relative": 0.863394 }, { "name": "video_fisheye_cam3", "l2": 5.712194, "mean_abs": 0.117508, "relative": 0.956438 }, { "name": "video_stereo_left", "l2": 5.266107, "mean_abs": 0.117066, "relative": 0.855753 }, { "name": "video_stereo_right", "l2": 6.436256, "mean_abs": 0.142872, "relative": 0.907377 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.029479, "mean_abs": 0.078856, "relative": 0.72632 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007732, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 795, "start_frame": 3975, "end_frame": 3994, "center_frame": 3984, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.388409, "mean_abs": 0.224365, "relative": 0.876307 }, { "name": "hand_right_joints", "l2": 9.856668, "mean_abs": 0.228908, "relative": 0.776969 }, { "name": "body_joints", "l2": 5.583822, "mean_abs": 0.088899, "relative": 0.754202 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.035712, "mean_abs": 0.004156, "relative": 0.084276 }, { "name": "camera_rotation_matrix", "l2": 3.005871, "mean_abs": 0.223934, "relative": 0.923054 }, { "name": "imu_accel_gyro", "l2": 17.056421, "mean_abs": 1.075849, "relative": 0.920253 }, { "name": "depth_confidence", "l2": 11.36599, "mean_abs": 0.208333, "relative": 0.000377 }, { "name": "video_fisheye_cam0", "l2": 5.304678, "mean_abs": 0.111451, "relative": 0.960564 }, { "name": "video_fisheye_cam1", "l2": 5.510483, "mean_abs": 0.119155, "relative": 0.910561 }, { "name": "video_fisheye_cam2", "l2": 5.460052, "mean_abs": 0.114141, "relative": 0.871616 }, { "name": "video_fisheye_cam3", "l2": 5.712339, "mean_abs": 0.118708, "relative": 0.956462 }, { "name": "video_stereo_left", "l2": 5.369936, "mean_abs": 0.12248, "relative": 0.872625 }, { "name": "video_stereo_right", "l2": 6.540681, "mean_abs": 0.150593, "relative": 0.922099 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.083555, "mean_abs": 0.080489, "relative": 0.745673 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007732, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 796, "start_frame": 3980, "end_frame": 3999, "center_frame": 3989, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.66562, "mean_abs": 0.233565, "relative": 0.899691 }, { "name": "hand_right_joints", "l2": 10.049615, "mean_abs": 0.236442, "relative": 0.792179 }, { "name": "body_joints", "l2": 5.63224, "mean_abs": 0.091665, "relative": 0.760742 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.072928, "mean_abs": 0.009468, "relative": 0.172099 }, { "name": "camera_rotation_matrix", "l2": 3.007047, "mean_abs": 0.225439, "relative": 0.923415 }, { "name": "imu_accel_gyro", "l2": 17.004065, "mean_abs": 1.079795, "relative": 0.917428 }, { "name": "depth_confidence", "l2": 11.272314, "mean_abs": 0.209069, "relative": 0.000374 }, { "name": "video_fisheye_cam0", "l2": 5.319998, "mean_abs": 0.112149, "relative": 0.963338 }, { "name": "video_fisheye_cam1", "l2": 5.453544, "mean_abs": 0.119742, "relative": 0.901152 }, { "name": "video_fisheye_cam2", "l2": 5.48776, "mean_abs": 0.115441, "relative": 0.876039 }, { "name": "video_fisheye_cam3", "l2": 5.721007, "mean_abs": 0.120218, "relative": 0.957914 }, { "name": "video_stereo_left", "l2": 5.394476, "mean_abs": 0.123861, "relative": 0.876613 }, { "name": "video_stereo_right", "l2": 6.544586, "mean_abs": 0.15375, "relative": 0.922649 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.165803, "mean_abs": 0.085586, "relative": 0.775109 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.007732, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 797, "start_frame": 3985, "end_frame": 4004, "center_frame": 3994, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.784859, "mean_abs": 0.236796, "relative": 0.909749 }, { "name": "hand_right_joints", "l2": 10.237277, "mean_abs": 0.239729, "relative": 0.806971 }, { "name": "body_joints", "l2": 5.666719, "mean_abs": 0.093954, "relative": 0.765399 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.129801, "mean_abs": 0.01855, "relative": 0.306312 }, { "name": "camera_rotation_matrix", "l2": 3.004472, "mean_abs": 0.223618, "relative": 0.922625 }, { "name": "imu_accel_gyro", "l2": 17.063562, "mean_abs": 1.103029, "relative": 0.920638 }, { "name": "depth_confidence", "l2": 11.185803, "mean_abs": 0.20896, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.32843, "mean_abs": 0.11208, "relative": 0.964865 }, { "name": "video_fisheye_cam1", "l2": 5.409343, "mean_abs": 0.119338, "relative": 0.893848 }, { "name": "video_fisheye_cam2", "l2": 5.495332, "mean_abs": 0.115683, "relative": 0.877248 }, { "name": "video_fisheye_cam3", "l2": 5.731921, "mean_abs": 0.121024, "relative": 0.959741 }, { "name": "video_stereo_left", "l2": 5.419598, "mean_abs": 0.124212, "relative": 0.880695 }, { "name": "video_stereo_right", "l2": 6.52927, "mean_abs": 0.153824, "relative": 0.92049 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.409915, "mean_abs": 0.100286, "relative": 0.862472 }, { "name": "caption_objects_interaction_text", "l2": 2.00171, "mean_abs": 0.014095, "relative": 0.919137 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 798, "start_frame": 3990, "end_frame": 4009, "center_frame": 3999, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.931581, "mean_abs": 0.24048, "relative": 0.922125 }, { "name": "hand_right_joints", "l2": 10.458396, "mean_abs": 0.247883, "relative": 0.824401 }, { "name": "body_joints", "l2": 5.772514, "mean_abs": 0.095957, "relative": 0.779689 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.176791, "mean_abs": 0.026348, "relative": 0.417201 }, { "name": "camera_rotation_matrix", "l2": 3.00534, "mean_abs": 0.221103, "relative": 0.922891 }, { "name": "imu_accel_gyro", "l2": 17.081991, "mean_abs": 1.117766, "relative": 0.921632 }, { "name": "depth_confidence", "l2": 11.061253, "mean_abs": 0.208882, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.345859, "mean_abs": 0.112043, "relative": 0.968021 }, { "name": "video_fisheye_cam1", "l2": 5.334278, "mean_abs": 0.118344, "relative": 0.881444 }, { "name": "video_fisheye_cam2", "l2": 5.508873, "mean_abs": 0.116166, "relative": 0.87941 }, { "name": "video_fisheye_cam3", "l2": 5.723697, "mean_abs": 0.121173, "relative": 0.958364 }, { "name": "video_stereo_left", "l2": 5.48824, "mean_abs": 0.125198, "relative": 0.89185 }, { "name": "video_stereo_right", "l2": 6.528382, "mean_abs": 0.152705, "relative": 0.920365 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.215351, "mean_abs": 0.090669, "relative": 0.792841 }, { "name": "caption_objects_interaction_text", "l2": 2.00171, "mean_abs": 0.014325, "relative": 0.919137 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 799, "start_frame": 3995, "end_frame": 4014, "center_frame": 4004, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.127092, "mean_abs": 0.238147, "relative": 0.938618 }, { "name": "hand_right_joints", "l2": 10.853699, "mean_abs": 0.260562, "relative": 0.855562 }, { "name": "body_joints", "l2": 5.790402, "mean_abs": 0.092295, "relative": 0.782105 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.194474, "mean_abs": 0.02999, "relative": 0.45893 }, { "name": "camera_rotation_matrix", "l2": 3.008673, "mean_abs": 0.215204, "relative": 0.923915 }, { "name": "imu_accel_gyro", "l2": 17.207102, "mean_abs": 1.120755, "relative": 0.928382 }, { "name": "depth_confidence", "l2": 10.80829, "mean_abs": 0.205042, "relative": 0.000359 }, { "name": "video_fisheye_cam0", "l2": 5.375003, "mean_abs": 0.112537, "relative": 0.973298 }, { "name": "video_fisheye_cam1", "l2": 5.245997, "mean_abs": 0.114421, "relative": 0.866857 }, { "name": "video_fisheye_cam2", "l2": 5.51793, "mean_abs": 0.115058, "relative": 0.880855 }, { "name": "video_fisheye_cam3", "l2": 5.710959, "mean_abs": 0.119636, "relative": 0.956231 }, { "name": "video_stereo_left", "l2": 5.460088, "mean_abs": 0.123162, "relative": 0.887275 }, { "name": "video_stereo_right", "l2": 6.448806, "mean_abs": 0.147259, "relative": 0.909146 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.212838, "mean_abs": 0.087705, "relative": 0.791941 }, { "name": "caption_objects_interaction_text", "l2": 2.00171, "mean_abs": 0.014374, "relative": 0.919137 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 800, "start_frame": 4000, "end_frame": 4019, "center_frame": 4009, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.049132, "mean_abs": 0.239607, "relative": 0.932041 }, { "name": "hand_right_joints", "l2": 11.182856, "mean_abs": 0.268485, "relative": 0.881508 }, { "name": "body_joints", "l2": 5.810286, "mean_abs": 0.091739, "relative": 0.784791 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.176619, "mean_abs": 0.027571, "relative": 0.416795 }, { "name": "camera_rotation_matrix", "l2": 3.006133, "mean_abs": 0.213429, "relative": 0.923135 }, { "name": "imu_accel_gyro", "l2": 17.15855, "mean_abs": 1.111063, "relative": 0.925763 }, { "name": "depth_confidence", "l2": 10.629565, "mean_abs": 0.203077, "relative": 0.000353 }, { "name": "video_fisheye_cam0", "l2": 5.377932, "mean_abs": 0.112872, "relative": 0.973828 }, { "name": "video_fisheye_cam1", "l2": 5.19352, "mean_abs": 0.11308, "relative": 0.858185 }, { "name": "video_fisheye_cam2", "l2": 5.505539, "mean_abs": 0.114866, "relative": 0.878877 }, { "name": "video_fisheye_cam3", "l2": 5.682854, "mean_abs": 0.119254, "relative": 0.951525 }, { "name": "video_stereo_left", "l2": 5.425525, "mean_abs": 0.121522, "relative": 0.881659 }, { "name": "video_stereo_right", "l2": 6.378807, "mean_abs": 0.145013, "relative": 0.899278 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.163871, "mean_abs": 0.086138, "relative": 0.774417 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 801, "start_frame": 4005, "end_frame": 4024, "center_frame": 4014, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.935331, "mean_abs": 0.236019, "relative": 0.922442 }, { "name": "hand_right_joints", "l2": 11.373071, "mean_abs": 0.272112, "relative": 0.896502 }, { "name": "body_joints", "l2": 5.885396, "mean_abs": 0.094171, "relative": 0.794936 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.121315, "mean_abs": 0.019216, "relative": 0.286285 }, { "name": "camera_rotation_matrix", "l2": 3.006117, "mean_abs": 0.212365, "relative": 0.92313 }, { "name": "imu_accel_gyro", "l2": 17.246279, "mean_abs": 1.1168, "relative": 0.930496 }, { "name": "depth_confidence", "l2": 10.518517, "mean_abs": 0.201958, "relative": 0.000349 }, { "name": "video_fisheye_cam0", "l2": 5.382133, "mean_abs": 0.112905, "relative": 0.974589 }, { "name": "video_fisheye_cam1", "l2": 5.186543, "mean_abs": 0.113184, "relative": 0.857032 }, { "name": "video_fisheye_cam2", "l2": 5.531132, "mean_abs": 0.115, "relative": 0.882963 }, { "name": "video_fisheye_cam3", "l2": 5.672, "mean_abs": 0.118983, "relative": 0.949708 }, { "name": "video_stereo_left", "l2": 5.471694, "mean_abs": 0.122893, "relative": 0.889161 }, { "name": "video_stereo_right", "l2": 6.415904, "mean_abs": 0.146569, "relative": 0.904508 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.005398, "mean_abs": 0.079292, "relative": 0.717702 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 802, "start_frame": 4010, "end_frame": 4029, "center_frame": 4019, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.798339, "mean_abs": 0.230676, "relative": 0.910886 }, { "name": "hand_right_joints", "l2": 11.5193, "mean_abs": 0.274591, "relative": 0.908029 }, { "name": "body_joints", "l2": 5.931112, "mean_abs": 0.093011, "relative": 0.80111 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.078491, "mean_abs": 0.011264, "relative": 0.185228 }, { "name": "camera_rotation_matrix", "l2": 3.007652, "mean_abs": 0.21326, "relative": 0.923601 }, { "name": "imu_accel_gyro", "l2": 17.380455, "mean_abs": 1.15234, "relative": 0.937735 }, { "name": "depth_confidence", "l2": 10.341683, "mean_abs": 0.198621, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.381575, "mean_abs": 0.113246, "relative": 0.974488 }, { "name": "video_fisheye_cam1", "l2": 5.188729, "mean_abs": 0.113822, "relative": 0.857394 }, { "name": "video_fisheye_cam2", "l2": 5.546607, "mean_abs": 0.115833, "relative": 0.885433 }, { "name": "video_fisheye_cam3", "l2": 5.660044, "mean_abs": 0.117986, "relative": 0.947706 }, { "name": "video_stereo_left", "l2": 5.489056, "mean_abs": 0.123123, "relative": 0.891982 }, { "name": "video_stereo_right", "l2": 6.436946, "mean_abs": 0.146907, "relative": 0.907474 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.928249, "mean_abs": 0.073738, "relative": 0.690091 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 803, "start_frame": 4015, "end_frame": 4034, "center_frame": 4024, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.665799, "mean_abs": 0.226246, "relative": 0.899706 }, { "name": "hand_right_joints", "l2": 11.37043, "mean_abs": 0.28389, "relative": 0.896294 }, { "name": "body_joints", "l2": 5.916212, "mean_abs": 0.08969, "relative": 0.799098 }, { "name": "body_contacts", "l2": 3.464111, "mean_abs": 0.081714, "relative": 0.999627 }, { "name": "camera_translation", "l2": 0.112067, "mean_abs": 0.015009, "relative": 0.264463 }, { "name": "camera_rotation_matrix", "l2": 3.008277, "mean_abs": 0.210694, "relative": 0.923793 }, { "name": "imu_accel_gyro", "l2": 17.29287, "mean_abs": 1.123732, "relative": 0.93301 }, { "name": "depth_confidence", "l2": 10.281856, "mean_abs": 0.195578, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.36311, "mean_abs": 0.112752, "relative": 0.971145 }, { "name": "video_fisheye_cam1", "l2": 5.238605, "mean_abs": 0.11589, "relative": 0.865635 }, { "name": "video_fisheye_cam2", "l2": 5.531469, "mean_abs": 0.115845, "relative": 0.883017 }, { "name": "video_fisheye_cam3", "l2": 5.633103, "mean_abs": 0.117001, "relative": 0.943195 }, { "name": "video_stereo_left", "l2": 5.451599, "mean_abs": 0.122285, "relative": 0.885896 }, { "name": "video_stereo_right", "l2": 6.385776, "mean_abs": 0.14661, "relative": 0.90026 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.980501, "mean_abs": 0.077427, "relative": 0.708792 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 804, "start_frame": 4020, "end_frame": 4039, "center_frame": 4029, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.720324, "mean_abs": 0.231187, "relative": 0.904305 }, { "name": "hand_right_joints", "l2": 11.244705, "mean_abs": 0.293561, "relative": 0.886384 }, { "name": "body_joints", "l2": 6.040574, "mean_abs": 0.098456, "relative": 0.815895 }, { "name": "body_contacts", "l2": 3.46411, "mean_abs": 0.081697, "relative": 0.999627 }, { "name": "camera_translation", "l2": 0.192463, "mean_abs": 0.023083, "relative": 0.454185 }, { "name": "camera_rotation_matrix", "l2": 3.008008, "mean_abs": 0.208581, "relative": 0.92371 }, { "name": "imu_accel_gyro", "l2": 17.188028, "mean_abs": 1.105055, "relative": 0.927353 }, { "name": "depth_confidence", "l2": 10.258665, "mean_abs": 0.197346, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.349219, "mean_abs": 0.112821, "relative": 0.968629 }, { "name": "video_fisheye_cam1", "l2": 5.261896, "mean_abs": 0.117286, "relative": 0.869484 }, { "name": "video_fisheye_cam2", "l2": 5.516411, "mean_abs": 0.116271, "relative": 0.880613 }, { "name": "video_fisheye_cam3", "l2": 5.629555, "mean_abs": 0.117693, "relative": 0.942601 }, { "name": "video_stereo_left", "l2": 5.39753, "mean_abs": 0.123327, "relative": 0.877109 }, { "name": "video_stereo_right", "l2": 6.400542, "mean_abs": 0.148748, "relative": 0.902342 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.9816, "mean_abs": 0.076084, "relative": 0.709185 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 805, "start_frame": 4025, "end_frame": 4044, "center_frame": 4034, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.884764, "mean_abs": 0.239747, "relative": 0.918176 }, { "name": "hand_right_joints", "l2": 11.096506, "mean_abs": 0.295854, "relative": 0.874702 }, { "name": "body_joints", "l2": 6.115995, "mean_abs": 0.103429, "relative": 0.826082 }, { "name": "body_contacts", "l2": 3.46411, "mean_abs": 0.081698, "relative": 0.999627 }, { "name": "camera_translation", "l2": 0.217766, "mean_abs": 0.026255, "relative": 0.513897 }, { "name": "camera_rotation_matrix", "l2": 3.00829, "mean_abs": 0.209075, "relative": 0.923797 }, { "name": "imu_accel_gyro", "l2": 17.167549, "mean_abs": 1.117668, "relative": 0.926248 }, { "name": "depth_confidence", "l2": 10.236515, "mean_abs": 0.19989, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.323146, "mean_abs": 0.112129, "relative": 0.963908 }, { "name": "video_fisheye_cam1", "l2": 5.272017, "mean_abs": 0.118095, "relative": 0.871156 }, { "name": "video_fisheye_cam2", "l2": 5.484206, "mean_abs": 0.116479, "relative": 0.875472 }, { "name": "video_fisheye_cam3", "l2": 5.610458, "mean_abs": 0.116969, "relative": 0.939404 }, { "name": "video_stereo_left", "l2": 5.366417, "mean_abs": 0.124648, "relative": 0.872053 }, { "name": "video_stereo_right", "l2": 6.437698, "mean_abs": 0.15248, "relative": 0.90758 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.995843, "mean_abs": 0.075069, "relative": 0.714282 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 806, "start_frame": 4030, "end_frame": 4049, "center_frame": 4039, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.007415, "mean_abs": 0.2465, "relative": 0.928522 }, { "name": "hand_right_joints", "l2": 10.742352, "mean_abs": 0.282701, "relative": 0.846785 }, { "name": "body_joints", "l2": 6.012321, "mean_abs": 0.099488, "relative": 0.812079 }, { "name": "body_contacts", "l2": 3.46411, "mean_abs": 0.081696, "relative": 0.999627 }, { "name": "camera_translation", "l2": 0.175099, "mean_abs": 0.021715, "relative": 0.413209 }, { "name": "camera_rotation_matrix", "l2": 3.006243, "mean_abs": 0.203538, "relative": 0.923168 }, { "name": "imu_accel_gyro", "l2": 17.085629, "mean_abs": 1.100651, "relative": 0.921829 }, { "name": "depth_confidence", "l2": 10.103834, "mean_abs": 0.196414, "relative": 0.000335 }, { "name": "video_fisheye_cam0", "l2": 5.30886, "mean_abs": 0.112059, "relative": 0.961321 }, { "name": "video_fisheye_cam1", "l2": 5.226712, "mean_abs": 0.114939, "relative": 0.86367 }, { "name": "video_fisheye_cam2", "l2": 5.402647, "mean_abs": 0.114288, "relative": 0.862452 }, { "name": "video_fisheye_cam3", "l2": 5.612916, "mean_abs": 0.117792, "relative": 0.939815 }, { "name": "video_stereo_left", "l2": 5.361498, "mean_abs": 0.124683, "relative": 0.871254 }, { "name": "video_stereo_right", "l2": 6.36993, "mean_abs": 0.150997, "relative": 0.898026 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.04161, "mean_abs": 0.076973, "relative": 0.730662 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 807, "start_frame": 4035, "end_frame": 4054, "center_frame": 4044, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.1415, "mean_abs": 0.250522, "relative": 0.939833 }, { "name": "hand_right_joints", "l2": 10.515497, "mean_abs": 0.264567, "relative": 0.828902 }, { "name": "body_joints", "l2": 5.864493, "mean_abs": 0.089998, "relative": 0.792112 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.102336, "mean_abs": 0.015125, "relative": 0.241498 }, { "name": "camera_rotation_matrix", "l2": 3.011242, "mean_abs": 0.209187, "relative": 0.924704 }, { "name": "imu_accel_gyro", "l2": 17.16976, "mean_abs": 1.175288, "relative": 0.926368 }, { "name": "depth_confidence", "l2": 9.902898, "mean_abs": 0.19256, "relative": 0.000329 }, { "name": "video_fisheye_cam0", "l2": 5.296736, "mean_abs": 0.112247, "relative": 0.959126 }, { "name": "video_fisheye_cam1", "l2": 5.229632, "mean_abs": 0.114481, "relative": 0.864153 }, { "name": "video_fisheye_cam2", "l2": 5.350837, "mean_abs": 0.11355, "relative": 0.854182 }, { "name": "video_fisheye_cam3", "l2": 5.640432, "mean_abs": 0.119051, "relative": 0.944422 }, { "name": "video_stereo_left", "l2": 5.360812, "mean_abs": 0.123007, "relative": 0.871143 }, { "name": "video_stereo_right", "l2": 6.290777, "mean_abs": 0.146977, "relative": 0.886868 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.167829, "mean_abs": 0.084932, "relative": 0.775834 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 808, "start_frame": 4040, "end_frame": 4059, "center_frame": 4049, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.382859, "mean_abs": 0.254741, "relative": 0.960193 }, { "name": "hand_right_joints", "l2": 10.683808, "mean_abs": 0.268566, "relative": 0.84217 }, { "name": "body_joints", "l2": 5.909683, "mean_abs": 0.091648, "relative": 0.798216 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.084108, "mean_abs": 0.011036, "relative": 0.198484 }, { "name": "camera_rotation_matrix", "l2": 3.016288, "mean_abs": 0.213152, "relative": 0.926253 }, { "name": "imu_accel_gyro", "l2": 17.20388, "mean_abs": 1.195538, "relative": 0.928209 }, { "name": "depth_confidence", "l2": 9.765123, "mean_abs": 0.189366, "relative": 0.000324 }, { "name": "video_fisheye_cam0", "l2": 5.30096, "mean_abs": 0.112479, "relative": 0.95989 }, { "name": "video_fisheye_cam1", "l2": 5.246286, "mean_abs": 0.115548, "relative": 0.866904 }, { "name": "video_fisheye_cam2", "l2": 5.353693, "mean_abs": 0.11368, "relative": 0.854637 }, { "name": "video_fisheye_cam3", "l2": 5.642964, "mean_abs": 0.119021, "relative": 0.944846 }, { "name": "video_stereo_left", "l2": 5.357334, "mean_abs": 0.12211, "relative": 0.870577 }, { "name": "video_stereo_right", "l2": 6.25569, "mean_abs": 0.144931, "relative": 0.881921 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.017699, "mean_abs": 0.07867, "relative": 0.722104 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 809, "start_frame": 4045, "end_frame": 4064, "center_frame": 4054, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.377158, "mean_abs": 0.259534, "relative": 0.959712 }, { "name": "hand_right_joints", "l2": 10.81256, "mean_abs": 0.274911, "relative": 0.852319 }, { "name": "body_joints", "l2": 5.974437, "mean_abs": 0.092723, "relative": 0.806962 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.097072, "mean_abs": 0.014049, "relative": 0.229077 }, { "name": "camera_rotation_matrix", "l2": 3.01055, "mean_abs": 0.210395, "relative": 0.924491 }, { "name": "imu_accel_gyro", "l2": 17.143387, "mean_abs": 1.185655, "relative": 0.924945 }, { "name": "depth_confidence", "l2": 9.579651, "mean_abs": 0.183879, "relative": 0.000318 }, { "name": "video_fisheye_cam0", "l2": 5.285543, "mean_abs": 0.111244, "relative": 0.957099 }, { "name": "video_fisheye_cam1", "l2": 5.235734, "mean_abs": 0.115242, "relative": 0.865161 }, { "name": "video_fisheye_cam2", "l2": 5.320634, "mean_abs": 0.111934, "relative": 0.84936 }, { "name": "video_fisheye_cam3", "l2": 5.606114, "mean_abs": 0.117139, "relative": 0.938676 }, { "name": "video_stereo_left", "l2": 5.338667, "mean_abs": 0.120353, "relative": 0.867544 }, { "name": "video_stereo_right", "l2": 6.238998, "mean_abs": 0.142448, "relative": 0.879568 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.118121, "mean_abs": 0.086749, "relative": 0.758044 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 810, "start_frame": 4050, "end_frame": 4069, "center_frame": 4059, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.349115, "mean_abs": 0.261511, "relative": 0.957346 }, { "name": "hand_right_joints", "l2": 10.938685, "mean_abs": 0.279743, "relative": 0.862261 }, { "name": "body_joints", "l2": 6.000726, "mean_abs": 0.090813, "relative": 0.810513 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.084649, "mean_abs": 0.012477, "relative": 0.199761 }, { "name": "camera_rotation_matrix", "l2": 3.003442, "mean_abs": 0.205239, "relative": 0.922308 }, { "name": "imu_accel_gyro", "l2": 17.196854, "mean_abs": 1.173789, "relative": 0.927829 }, { "name": "depth_confidence", "l2": 9.371161, "mean_abs": 0.179325, "relative": 0.000311 }, { "name": "video_fisheye_cam0", "l2": 5.274112, "mean_abs": 0.110047, "relative": 0.955029 }, { "name": "video_fisheye_cam1", "l2": 5.219566, "mean_abs": 0.113573, "relative": 0.862489 }, { "name": "video_fisheye_cam2", "l2": 5.306264, "mean_abs": 0.110568, "relative": 0.847066 }, { "name": "video_fisheye_cam3", "l2": 5.603145, "mean_abs": 0.116024, "relative": 0.938179 }, { "name": "video_stereo_left", "l2": 5.274808, "mean_abs": 0.117258, "relative": 0.857167 }, { "name": "video_stereo_right", "l2": 6.196517, "mean_abs": 0.140011, "relative": 0.873579 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.148432, "mean_abs": 0.088957, "relative": 0.768891 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 811, "start_frame": 4055, "end_frame": 4074, "center_frame": 4064, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.335739, "mean_abs": 0.260142, "relative": 0.956218 }, { "name": "hand_right_joints", "l2": 11.131083, "mean_abs": 0.282181, "relative": 0.877427 }, { "name": "body_joints", "l2": 6.027937, "mean_abs": 0.08849, "relative": 0.814189 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.071443, "mean_abs": 0.010731, "relative": 0.168596 }, { "name": "camera_rotation_matrix", "l2": 3.003254, "mean_abs": 0.204199, "relative": 0.922251 }, { "name": "imu_accel_gyro", "l2": 17.044287, "mean_abs": 1.140639, "relative": 0.919598 }, { "name": "depth_confidence", "l2": 9.235234, "mean_abs": 0.1751, "relative": 0.000307 }, { "name": "video_fisheye_cam0", "l2": 5.275416, "mean_abs": 0.109419, "relative": 0.955265 }, { "name": "video_fisheye_cam1", "l2": 5.240833, "mean_abs": 0.113157, "relative": 0.866003 }, { "name": "video_fisheye_cam2", "l2": 5.317775, "mean_abs": 0.11036, "relative": 0.848904 }, { "name": "video_fisheye_cam3", "l2": 5.598926, "mean_abs": 0.115049, "relative": 0.937473 }, { "name": "video_stereo_left", "l2": 5.263485, "mean_abs": 0.115428, "relative": 0.855327 }, { "name": "video_stereo_right", "l2": 6.193674, "mean_abs": 0.1379, "relative": 0.873178 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.091767, "mean_abs": 0.086375, "relative": 0.748612 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 812, "start_frame": 4060, "end_frame": 4079, "center_frame": 4069, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.324466, "mean_abs": 0.258879, "relative": 0.955267 }, { "name": "hand_right_joints", "l2": 11.213057, "mean_abs": 0.282714, "relative": 0.883889 }, { "name": "body_joints", "l2": 6.012936, "mean_abs": 0.088451, "relative": 0.812162 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.047573, "mean_abs": 0.006611, "relative": 0.112265 }, { "name": "camera_rotation_matrix", "l2": 3.001589, "mean_abs": 0.201577, "relative": 0.921739 }, { "name": "imu_accel_gyro", "l2": 17.036774, "mean_abs": 1.153202, "relative": 0.919193 }, { "name": "depth_confidence", "l2": 9.16679, "mean_abs": 0.17237, "relative": 0.000304 }, { "name": "video_fisheye_cam0", "l2": 5.276905, "mean_abs": 0.109114, "relative": 0.955535 }, { "name": "video_fisheye_cam1", "l2": 5.24663, "mean_abs": 0.111863, "relative": 0.866961 }, { "name": "video_fisheye_cam2", "l2": 5.368429, "mean_abs": 0.110464, "relative": 0.85699 }, { "name": "video_fisheye_cam3", "l2": 5.571531, "mean_abs": 0.113679, "relative": 0.932886 }, { "name": "video_stereo_left", "l2": 5.28509, "mean_abs": 0.115651, "relative": 0.858838 }, { "name": "video_stereo_right", "l2": 6.197618, "mean_abs": 0.13744, "relative": 0.873734 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.164553, "mean_abs": 0.092443, "relative": 0.774661 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": null, "next_action": null, "contact_prediction": null, "object_relevance": null } }, { "window_index": 813, "start_frame": 4065, "end_frame": 4084, "center_frame": 4074, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.346816, "mean_abs": 0.259533, "relative": 0.957152 }, { "name": "hand_right_joints", "l2": 11.22286, "mean_abs": 0.281427, "relative": 0.884662 }, { "name": "body_joints", "l2": 5.992306, "mean_abs": 0.088045, "relative": 0.809376 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030456, "mean_abs": 0.004515, "relative": 0.071873 }, { "name": "camera_rotation_matrix", "l2": 3.001274, "mean_abs": 0.202583, "relative": 0.921643 }, { "name": "imu_accel_gyro", "l2": 17.058809, "mean_abs": 1.135212, "relative": 0.920381 }, { "name": "depth_confidence", "l2": 9.102178, "mean_abs": 0.170334, "relative": 0.000302 }, { "name": "video_fisheye_cam0", "l2": 5.279932, "mean_abs": 0.109363, "relative": 0.956083 }, { "name": "video_fisheye_cam1", "l2": 5.254374, "mean_abs": 0.112264, "relative": 0.868241 }, { "name": "video_fisheye_cam2", "l2": 5.395824, "mean_abs": 0.110694, "relative": 0.861363 }, { "name": "video_fisheye_cam3", "l2": 5.575292, "mean_abs": 0.113846, "relative": 0.933515 }, { "name": "video_stereo_left", "l2": 5.2935, "mean_abs": 0.115043, "relative": 0.860204 }, { "name": "video_stereo_right", "l2": 6.180744, "mean_abs": 0.136977, "relative": 0.871355 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.148417, "mean_abs": 0.095004, "relative": 0.768886 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pour and close white bottle", "correct": 1, "confidence": "0.9981641173362732" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999991655349731" }, "next_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.985846221446991" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|mug|white bottle", "predicted": "coffee dripper|scale|mug|white bottle", "correct": 1, "confidence": "" } } }, { "window_index": 814, "start_frame": 4070, "end_frame": 4089, "center_frame": 4079, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.330984, "mean_abs": 0.258854, "relative": 0.955817 }, { "name": "hand_right_joints", "l2": 11.246032, "mean_abs": 0.276724, "relative": 0.886488 }, { "name": "body_joints", "l2": 5.97885, "mean_abs": 0.087596, "relative": 0.807558 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018432, "mean_abs": 0.00285, "relative": 0.043497 }, { "name": "camera_rotation_matrix", "l2": 3.000129, "mean_abs": 0.201071, "relative": 0.921291 }, { "name": "imu_accel_gyro", "l2": 17.046793, "mean_abs": 1.135102, "relative": 0.919733 }, { "name": "depth_confidence", "l2": 9.018756, "mean_abs": 0.166752, "relative": 0.000299 }, { "name": "video_fisheye_cam0", "l2": 5.288094, "mean_abs": 0.109224, "relative": 0.957561 }, { "name": "video_fisheye_cam1", "l2": 5.265172, "mean_abs": 0.111863, "relative": 0.870025 }, { "name": "video_fisheye_cam2", "l2": 5.400666, "mean_abs": 0.110343, "relative": 0.862136 }, { "name": "video_fisheye_cam3", "l2": 5.604566, "mean_abs": 0.114077, "relative": 0.938417 }, { "name": "video_stereo_left", "l2": 5.257659, "mean_abs": 0.113818, "relative": 0.85438 }, { "name": "video_stereo_right", "l2": 6.095864, "mean_abs": 0.134478, "relative": 0.859389 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.117616, "mean_abs": 0.091611, "relative": 0.757863 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9614766836166382" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pour and close white bottle", "correct": 1, "confidence": "0.9945576190948486" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9981683492660522" }, "next_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9508370757102966" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|mug|white bottle", "predicted": "coffee dripper|scale|digital scale|mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 815, "start_frame": 4075, "end_frame": 4094, "center_frame": 4084, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.281485, "mean_abs": 0.26032, "relative": 0.951641 }, { "name": "hand_right_joints", "l2": 11.107795, "mean_abs": 0.278873, "relative": 0.875591 }, { "name": "body_joints", "l2": 5.987145, "mean_abs": 0.088854, "relative": 0.808679 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.038092, "mean_abs": 0.005178, "relative": 0.089891 }, { "name": "camera_rotation_matrix", "l2": 3.002466, "mean_abs": 0.202551, "relative": 0.922009 }, { "name": "imu_accel_gyro", "l2": 17.163584, "mean_abs": 1.157671, "relative": 0.926034 }, { "name": "depth_confidence", "l2": 9.013953, "mean_abs": 0.167138, "relative": 0.000299 }, { "name": "video_fisheye_cam0", "l2": 5.29157, "mean_abs": 0.109673, "relative": 0.95819 }, { "name": "video_fisheye_cam1", "l2": 5.288358, "mean_abs": 0.113569, "relative": 0.873856 }, { "name": "video_fisheye_cam2", "l2": 5.404102, "mean_abs": 0.111207, "relative": 0.862684 }, { "name": "video_fisheye_cam3", "l2": 5.624794, "mean_abs": 0.114851, "relative": 0.941804 }, { "name": "video_stereo_left", "l2": 5.262946, "mean_abs": 0.114907, "relative": 0.855239 }, { "name": "video_stereo_right", "l2": 6.105752, "mean_abs": 0.136277, "relative": 0.860783 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.104944, "mean_abs": 0.091162, "relative": 0.753328 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9060729146003723" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pour and close white bottle", "correct": 1, "confidence": "0.9972338080406189" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.6682508587837219" }, "next_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.972966730594635" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|mug|white bottle", "predicted": "coffee dripper|gooseneck kettle|digital scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 816, "start_frame": 4080, "end_frame": 4099, "center_frame": 4089, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.12554, "mean_abs": 0.265854, "relative": 0.938487 }, { "name": "hand_right_joints", "l2": 10.952866, "mean_abs": 0.283684, "relative": 0.863379 }, { "name": "body_joints", "l2": 6.010051, "mean_abs": 0.090664, "relative": 0.811773 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.085263, "mean_abs": 0.011283, "relative": 0.201208 }, { "name": "camera_rotation_matrix", "l2": 3.007039, "mean_abs": 0.205211, "relative": 0.923413 }, { "name": "imu_accel_gyro", "l2": 17.239559, "mean_abs": 1.174178, "relative": 0.930134 }, { "name": "depth_confidence", "l2": 9.015797, "mean_abs": 0.168762, "relative": 0.000299 }, { "name": "video_fisheye_cam0", "l2": 5.301674, "mean_abs": 0.110595, "relative": 0.96002 }, { "name": "video_fisheye_cam1", "l2": 5.35285, "mean_abs": 0.116527, "relative": 0.884513 }, { "name": "video_fisheye_cam2", "l2": 5.42043, "mean_abs": 0.113687, "relative": 0.865291 }, { "name": "video_fisheye_cam3", "l2": 5.658895, "mean_abs": 0.117251, "relative": 0.947514 }, { "name": "video_stereo_left", "l2": 5.308252, "mean_abs": 0.117556, "relative": 0.862601 }, { "name": "video_stereo_right", "l2": 6.217695, "mean_abs": 0.141109, "relative": 0.876565 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.252852, "mean_abs": 0.09812, "relative": 0.806262 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9094582200050354" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pour and close white bottle", "correct": 1, "confidence": "0.996189534664154" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9955647587776184" }, "next_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9830050468444824" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|mug|white bottle", "predicted": "coffee dripper|gooseneck kettle|digital scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 817, "start_frame": 4085, "end_frame": 4104, "center_frame": 4094, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.008843, "mean_abs": 0.271643, "relative": 0.928643 }, { "name": "hand_right_joints", "l2": 10.850536, "mean_abs": 0.289376, "relative": 0.855313 }, { "name": "body_joints", "l2": 6.066105, "mean_abs": 0.093099, "relative": 0.819344 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.134665, "mean_abs": 0.018278, "relative": 0.31779 }, { "name": "camera_rotation_matrix", "l2": 3.011482, "mean_abs": 0.206712, "relative": 0.924777 }, { "name": "imu_accel_gyro", "l2": 17.187407, "mean_abs": 1.161901, "relative": 0.92732 }, { "name": "depth_confidence", "l2": 9.026792, "mean_abs": 0.171522, "relative": 0.0003 }, { "name": "video_fisheye_cam0", "l2": 5.30385, "mean_abs": 0.111347, "relative": 0.960414 }, { "name": "video_fisheye_cam1", "l2": 5.437165, "mean_abs": 0.119739, "relative": 0.898446 }, { "name": "video_fisheye_cam2", "l2": 5.471928, "mean_abs": 0.116649, "relative": 0.873512 }, { "name": "video_fisheye_cam3", "l2": 5.639133, "mean_abs": 0.118696, "relative": 0.944205 }, { "name": "video_stereo_left", "l2": 5.404737, "mean_abs": 0.121423, "relative": 0.878281 }, { "name": "video_stereo_right", "l2": 6.36545, "mean_abs": 0.146437, "relative": 0.897395 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.12993, "mean_abs": 0.086224, "relative": 0.76227 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9192006587982178" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pour and close white bottle", "correct": 1, "confidence": "0.993263304233551" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999998807907104" }, "next_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9894412755966187" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|mug|white bottle", "predicted": "coffee dripper|gooseneck kettle|digital scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 818, "start_frame": 4090, "end_frame": 4109, "center_frame": 4099, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.93535, "mean_abs": 0.27408, "relative": 0.922443 }, { "name": "hand_right_joints", "l2": 10.710489, "mean_abs": 0.288617, "relative": 0.844273 }, { "name": "body_joints", "l2": 6.128921, "mean_abs": 0.095418, "relative": 0.827828 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.173165, "mean_abs": 0.025134, "relative": 0.408646 }, { "name": "camera_rotation_matrix", "l2": 3.013525, "mean_abs": 0.206634, "relative": 0.925405 }, { "name": "imu_accel_gyro", "l2": 17.264364, "mean_abs": 1.200074, "relative": 0.931472 }, { "name": "depth_confidence", "l2": 9.031713, "mean_abs": 0.172717, "relative": 0.0003 }, { "name": "video_fisheye_cam0", "l2": 5.303514, "mean_abs": 0.111863, "relative": 0.960353 }, { "name": "video_fisheye_cam1", "l2": 5.504992, "mean_abs": 0.121537, "relative": 0.909654 }, { "name": "video_fisheye_cam2", "l2": 5.502157, "mean_abs": 0.117877, "relative": 0.878338 }, { "name": "video_fisheye_cam3", "l2": 5.619502, "mean_abs": 0.119496, "relative": 0.940918 }, { "name": "video_stereo_left", "l2": 5.483545, "mean_abs": 0.12474, "relative": 0.891087 }, { "name": "video_stereo_right", "l2": 6.488694, "mean_abs": 0.150564, "relative": 0.91477 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.170504, "mean_abs": 0.088256, "relative": 0.776791 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.4563400447368622" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pour and close white bottle", "correct": 1, "confidence": "0.994482159614563" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9793558716773987" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|mug|white bottle", "predicted": "coffee dripper|gooseneck kettle|digital scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 819, "start_frame": 4095, "end_frame": 4114, "center_frame": 4104, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.75204, "mean_abs": 0.269612, "relative": 0.90698 }, { "name": "hand_right_joints", "l2": 10.495479, "mean_abs": 0.28191, "relative": 0.827325 }, { "name": "body_joints", "l2": 6.154868, "mean_abs": 0.094431, "relative": 0.831333 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.177026, "mean_abs": 0.026673, "relative": 0.417756 }, { "name": "camera_rotation_matrix", "l2": 3.011448, "mean_abs": 0.201747, "relative": 0.924767 }, { "name": "imu_accel_gyro", "l2": 17.195488, "mean_abs": 1.188078, "relative": 0.927756 }, { "name": "depth_confidence", "l2": 9.041092, "mean_abs": 0.172738, "relative": 0.0003 }, { "name": "video_fisheye_cam0", "l2": 5.301311, "mean_abs": 0.111629, "relative": 0.959954 }, { "name": "video_fisheye_cam1", "l2": 5.559116, "mean_abs": 0.121746, "relative": 0.918597 }, { "name": "video_fisheye_cam2", "l2": 5.517035, "mean_abs": 0.117928, "relative": 0.880713 }, { "name": "video_fisheye_cam3", "l2": 5.585454, "mean_abs": 0.119593, "relative": 0.935217 }, { "name": "video_stereo_left", "l2": 5.559309, "mean_abs": 0.126585, "relative": 0.903399 }, { "name": "video_stereo_right", "l2": 6.644889, "mean_abs": 0.153032, "relative": 0.93679 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.081777, "mean_abs": 0.083314, "relative": 0.745037 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8612736463546753" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pour and close white bottle", "correct": 1, "confidence": "0.9911273121833801" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9131289720535278" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|mug|white bottle", "predicted": "coffee dripper|digital scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 820, "start_frame": 4100, "end_frame": 4119, "center_frame": 4109, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.597322, "mean_abs": 0.264668, "relative": 0.893929 }, { "name": "hand_right_joints", "l2": 10.369132, "mean_abs": 0.278939, "relative": 0.817365 }, { "name": "body_joints", "l2": 6.190743, "mean_abs": 0.093591, "relative": 0.836179 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.139351, "mean_abs": 0.021491, "relative": 0.32885 }, { "name": "camera_rotation_matrix", "l2": 3.00862, "mean_abs": 0.196103, "relative": 0.923898 }, { "name": "imu_accel_gyro", "l2": 17.121035, "mean_abs": 1.144762, "relative": 0.923739 }, { "name": "depth_confidence", "l2": 9.170792, "mean_abs": 0.172958, "relative": 0.000304 }, { "name": "video_fisheye_cam0", "l2": 5.286664, "mean_abs": 0.110615, "relative": 0.957302 }, { "name": "video_fisheye_cam1", "l2": 5.590685, "mean_abs": 0.119911, "relative": 0.923813 }, { "name": "video_fisheye_cam2", "l2": 5.538882, "mean_abs": 0.117043, "relative": 0.8842 }, { "name": "video_fisheye_cam3", "l2": 5.4946, "mean_abs": 0.116272, "relative": 0.920004 }, { "name": "video_stereo_left", "l2": 5.668775, "mean_abs": 0.127868, "relative": 0.921187 }, { "name": "video_stereo_right", "l2": 6.770641, "mean_abs": 0.15373, "relative": 0.954518 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.110657, "mean_abs": 0.088999, "relative": 0.755372 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.00951, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.995844304561615" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pour and close white bottle", "correct": 1, "confidence": "0.9878567457199097" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.5049135684967041" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|mug|white bottle", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 821, "start_frame": 4105, "end_frame": 4124, "center_frame": 4114, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.55541, "mean_abs": 0.26461, "relative": 0.890394 }, { "name": "hand_right_joints", "l2": 10.343202, "mean_abs": 0.279211, "relative": 0.815321 }, { "name": "body_joints", "l2": 6.229258, "mean_abs": 0.093365, "relative": 0.841381 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.1037, "mean_abs": 0.015817, "relative": 0.244717 }, { "name": "camera_rotation_matrix", "l2": 3.011544, "mean_abs": 0.195334, "relative": 0.924796 }, { "name": "imu_accel_gyro", "l2": 16.980179, "mean_abs": 1.102353, "relative": 0.916139 }, { "name": "depth_confidence", "l2": 9.471438, "mean_abs": 0.180083, "relative": 0.000314 }, { "name": "video_fisheye_cam0", "l2": 5.279957, "mean_abs": 0.11066, "relative": 0.956087 }, { "name": "video_fisheye_cam1", "l2": 5.645813, "mean_abs": 0.11993, "relative": 0.932923 }, { "name": "video_fisheye_cam2", "l2": 5.604624, "mean_abs": 0.117498, "relative": 0.894695 }, { "name": "video_fisheye_cam3", "l2": 5.430186, "mean_abs": 0.1134, "relative": 0.909219 }, { "name": "video_stereo_left", "l2": 5.771262, "mean_abs": 0.13092, "relative": 0.937842 }, { "name": "video_stereo_right", "l2": 6.910061, "mean_abs": 0.157437, "relative": 0.974174 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.07203, "mean_abs": 0.086036, "relative": 0.741548 }, { "name": "caption_objects_interaction_text", "l2": 1.752537, "mean_abs": 0.009941, "relative": 0.804723 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9990665316581726" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pour and close white bottle", "correct": 1, "confidence": "0.9757102131843567" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9998751878738403" }, "next_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7660852074623108" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|mug|white bottle", "predicted": "coffee dripper|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 822, "start_frame": 4110, "end_frame": 4129, "center_frame": 4119, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.55214, "mean_abs": 0.266287, "relative": 0.890118 }, { "name": "hand_right_joints", "l2": 10.340579, "mean_abs": 0.280005, "relative": 0.815114 }, { "name": "body_joints", "l2": 6.245344, "mean_abs": 0.094081, "relative": 0.843553 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.091731, "mean_abs": 0.013391, "relative": 0.216471 }, { "name": "camera_rotation_matrix", "l2": 3.012448, "mean_abs": 0.198201, "relative": 0.925074 }, { "name": "imu_accel_gyro", "l2": 17.117744, "mean_abs": 1.11041, "relative": 0.923561 }, { "name": "depth_confidence", "l2": 9.807372, "mean_abs": 0.188522, "relative": 0.000325 }, { "name": "video_fisheye_cam0", "l2": 5.268601, "mean_abs": 0.110669, "relative": 0.954031 }, { "name": "video_fisheye_cam1", "l2": 5.62574, "mean_abs": 0.120276, "relative": 0.929606 }, { "name": "video_fisheye_cam2", "l2": 5.65322, "mean_abs": 0.118669, "relative": 0.902453 }, { "name": "video_fisheye_cam3", "l2": 5.354285, "mean_abs": 0.111102, "relative": 0.89651 }, { "name": "video_stereo_left", "l2": 5.821263, "mean_abs": 0.133613, "relative": 0.945967 }, { "name": "video_stereo_right", "l2": 6.947029, "mean_abs": 0.159708, "relative": 0.979385 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.057919, "mean_abs": 0.082044, "relative": 0.736498 }, { "name": "caption_objects_interaction_text", "l2": 1.752537, "mean_abs": 0.00992, "relative": 0.804722 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9990559220314026" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pour and close white bottle", "correct": 1, "confidence": "0.9482909440994263" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9976678490638733" }, "next_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6877899169921875" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|mug|white bottle", "predicted": "coffee dripper|digital scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 823, "start_frame": 4115, "end_frame": 4134, "center_frame": 4124, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale", "mug" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.560363, "mean_abs": 0.267254, "relative": 0.890812 }, { "name": "hand_right_joints", "l2": 10.356044, "mean_abs": 0.279467, "relative": 0.816333 }, { "name": "body_joints", "l2": 6.252188, "mean_abs": 0.094244, "relative": 0.844478 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.088234, "mean_abs": 0.012255, "relative": 0.208221 }, { "name": "camera_rotation_matrix", "l2": 3.010422, "mean_abs": 0.201128, "relative": 0.924452 }, { "name": "imu_accel_gyro", "l2": 17.17519, "mean_abs": 1.138455, "relative": 0.926661 }, { "name": "depth_confidence", "l2": 10.054501, "mean_abs": 0.193891, "relative": 0.000334 }, { "name": "video_fisheye_cam0", "l2": 5.264468, "mean_abs": 0.110963, "relative": 0.953283 }, { "name": "video_fisheye_cam1", "l2": 5.621536, "mean_abs": 0.120506, "relative": 0.928911 }, { "name": "video_fisheye_cam2", "l2": 5.692379, "mean_abs": 0.119454, "relative": 0.908704 }, { "name": "video_fisheye_cam3", "l2": 5.3627, "mean_abs": 0.111978, "relative": 0.89792 }, { "name": "video_stereo_left", "l2": 5.844367, "mean_abs": 0.134498, "relative": 0.949721 }, { "name": "video_stereo_right", "l2": 7.006176, "mean_abs": 0.161066, "relative": 0.987724 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.124394, "mean_abs": 0.085717, "relative": 0.760289 }, { "name": "caption_objects_interaction_text", "l2": 1.752537, "mean_abs": 0.009848, "relative": 0.804722 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9975965619087219" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pour and close white bottle", "correct": 1, "confidence": "0.9146074056625366" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999756813049316" }, "next_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5502874851226807" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|mug|white bottle", "predicted": "coffee dripper|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 824, "start_frame": 4120, "end_frame": 4139, "center_frame": 4129, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.556339, "mean_abs": 0.2644, "relative": 0.890472 }, { "name": "hand_right_joints", "l2": 10.36485, "mean_abs": 0.27779, "relative": 0.817027 }, { "name": "body_joints", "l2": 6.25061, "mean_abs": 0.094603, "relative": 0.844265 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.075848, "mean_abs": 0.0103, "relative": 0.178991 }, { "name": "camera_rotation_matrix", "l2": 3.007607, "mean_abs": 0.202735, "relative": 0.923587 }, { "name": "imu_accel_gyro", "l2": 17.172083, "mean_abs": 1.130591, "relative": 0.926493 }, { "name": "depth_confidence", "l2": 10.314184, "mean_abs": 0.197036, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.258231, "mean_abs": 0.110543, "relative": 0.952153 }, { "name": "video_fisheye_cam1", "l2": 5.593607, "mean_abs": 0.119732, "relative": 0.924296 }, { "name": "video_fisheye_cam2", "l2": 5.713509, "mean_abs": 0.11957, "relative": 0.912077 }, { "name": "video_fisheye_cam3", "l2": 5.375736, "mean_abs": 0.112045, "relative": 0.900102 }, { "name": "video_stereo_left", "l2": 5.806382, "mean_abs": 0.133293, "relative": 0.943549 }, { "name": "video_stereo_right", "l2": 6.974843, "mean_abs": 0.159814, "relative": 0.983307 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.014397, "mean_abs": 0.082137, "relative": 0.720923 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008948, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9955922961235046" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pour and close white bottle", "correct": 1, "confidence": "0.6174887418746948" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999983310699463" }, "next_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5016466379165649" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|scale|white bottle", "correct": 1, "confidence": "" } } }, { "window_index": 825, "start_frame": 4125, "end_frame": 4144, "center_frame": 4134, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.542014, "mean_abs": 0.259264, "relative": 0.889264 }, { "name": "hand_right_joints", "l2": 10.363018, "mean_abs": 0.27521, "relative": 0.816883 }, { "name": "body_joints", "l2": 6.229575, "mean_abs": 0.094135, "relative": 0.841424 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.054615, "mean_abs": 0.007708, "relative": 0.128883 }, { "name": "camera_rotation_matrix", "l2": 3.004501, "mean_abs": 0.201587, "relative": 0.922633 }, { "name": "imu_accel_gyro", "l2": 17.182514, "mean_abs": 1.096955, "relative": 0.927056 }, { "name": "depth_confidence", "l2": 10.517119, "mean_abs": 0.196374, "relative": 0.000349 }, { "name": "video_fisheye_cam0", "l2": 5.242838, "mean_abs": 0.109484, "relative": 0.949366 }, { "name": "video_fisheye_cam1", "l2": 5.551177, "mean_abs": 0.117497, "relative": 0.917285 }, { "name": "video_fisheye_cam2", "l2": 5.730763, "mean_abs": 0.118376, "relative": 0.914831 }, { "name": "video_fisheye_cam3", "l2": 5.392553, "mean_abs": 0.111668, "relative": 0.902918 }, { "name": "video_stereo_left", "l2": 5.688829, "mean_abs": 0.128159, "relative": 0.924446 }, { "name": "video_stereo_right", "l2": 6.872818, "mean_abs": 0.154574, "relative": 0.968923 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.99129, "mean_abs": 0.080352, "relative": 0.712653 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008948, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9297770261764526" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.8555364012718201" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9998675584793091" }, "next_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.5952329039573669" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|scale|bottle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 826, "start_frame": 4130, "end_frame": 4149, "center_frame": 4139, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.53662, "mean_abs": 0.256549, "relative": 0.888809 }, { "name": "hand_right_joints", "l2": 10.35833, "mean_abs": 0.272881, "relative": 0.816513 }, { "name": "body_joints", "l2": 6.194186, "mean_abs": 0.093032, "relative": 0.836644 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.031348, "mean_abs": 0.004843, "relative": 0.073976 }, { "name": "camera_rotation_matrix", "l2": 3.001944, "mean_abs": 0.198777, "relative": 0.921848 }, { "name": "imu_accel_gyro", "l2": 17.048681, "mean_abs": 1.042963, "relative": 0.919835 }, { "name": "depth_confidence", "l2": 10.588, "mean_abs": 0.194616, "relative": 0.000351 }, { "name": "video_fisheye_cam0", "l2": 5.227119, "mean_abs": 0.108335, "relative": 0.946519 }, { "name": "video_fisheye_cam1", "l2": 5.544506, "mean_abs": 0.116234, "relative": 0.916183 }, { "name": "video_fisheye_cam2", "l2": 5.74491, "mean_abs": 0.117907, "relative": 0.917089 }, { "name": "video_fisheye_cam3", "l2": 5.467968, "mean_abs": 0.112356, "relative": 0.915545 }, { "name": "video_stereo_left", "l2": 5.639524, "mean_abs": 0.124456, "relative": 0.916434 }, { "name": "video_stereo_right", "l2": 6.86916, "mean_abs": 0.151927, "relative": 0.968407 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.990703, "mean_abs": 0.080199, "relative": 0.712443 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008948, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5903907418251038" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9208164811134338" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9984899759292603" }, "next_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.401456356048584" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|scale|bottle|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 827, "start_frame": 4135, "end_frame": 4154, "center_frame": 4144, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.541407, "mean_abs": 0.255398, "relative": 0.889213 }, { "name": "hand_right_joints", "l2": 10.354847, "mean_abs": 0.272721, "relative": 0.816239 }, { "name": "body_joints", "l2": 6.183129, "mean_abs": 0.092586, "relative": 0.83515 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020416, "mean_abs": 0.002871, "relative": 0.048178 }, { "name": "camera_rotation_matrix", "l2": 3.000206, "mean_abs": 0.197682, "relative": 0.921314 }, { "name": "imu_accel_gyro", "l2": 17.088169, "mean_abs": 1.040367, "relative": 0.921966 }, { "name": "depth_confidence", "l2": 10.650749, "mean_abs": 0.193577, "relative": 0.000353 }, { "name": "video_fisheye_cam0", "l2": 5.221791, "mean_abs": 0.107686, "relative": 0.945555 }, { "name": "video_fisheye_cam1", "l2": 5.539671, "mean_abs": 0.115582, "relative": 0.915384 }, { "name": "video_fisheye_cam2", "l2": 5.755138, "mean_abs": 0.117489, "relative": 0.918722 }, { "name": "video_fisheye_cam3", "l2": 5.506383, "mean_abs": 0.111162, "relative": 0.921978 }, { "name": "video_stereo_left", "l2": 5.626131, "mean_abs": 0.122029, "relative": 0.914257 }, { "name": "video_stereo_right", "l2": 6.886612, "mean_abs": 0.148603, "relative": 0.970868 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.01745, "mean_abs": 0.081992, "relative": 0.722015 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008948, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.5007221102714539" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9362494945526123" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.5771944522857666" }, "next_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.4788297712802887" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|scale|bottle|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 828, "start_frame": 4140, "end_frame": 4159, "center_frame": 4149, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.546544, "mean_abs": 0.256231, "relative": 0.889646 }, { "name": "hand_right_joints", "l2": 10.36547, "mean_abs": 0.272415, "relative": 0.817076 }, { "name": "body_joints", "l2": 6.17605, "mean_abs": 0.092662, "relative": 0.834194 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017902, "mean_abs": 0.002155, "relative": 0.042247 }, { "name": "camera_rotation_matrix", "l2": 3.000639, "mean_abs": 0.198548, "relative": 0.921448 }, { "name": "imu_accel_gyro", "l2": 17.070545, "mean_abs": 1.02321, "relative": 0.921015 }, { "name": "depth_confidence", "l2": 10.664892, "mean_abs": 0.194149, "relative": 0.000354 }, { "name": "video_fisheye_cam0", "l2": 5.216743, "mean_abs": 0.107772, "relative": 0.944641 }, { "name": "video_fisheye_cam1", "l2": 5.527896, "mean_abs": 0.115948, "relative": 0.913438 }, { "name": "video_fisheye_cam2", "l2": 5.754392, "mean_abs": 0.117675, "relative": 0.918603 }, { "name": "video_fisheye_cam3", "l2": 5.522768, "mean_abs": 0.111243, "relative": 0.924721 }, { "name": "video_stereo_left", "l2": 5.626396, "mean_abs": 0.122977, "relative": 0.914301 }, { "name": "video_stereo_right", "l2": 6.892865, "mean_abs": 0.149269, "relative": 0.971749 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.026423, "mean_abs": 0.082713, "relative": 0.725227 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008948, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.3735993802547455" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9306931495666504" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.7210298180580139" }, "next_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.3303944766521454" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|scale|bottle|gooseneck kettle|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 829, "start_frame": 4145, "end_frame": 4164, "center_frame": 4154, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.547962, "mean_abs": 0.256864, "relative": 0.889766 }, { "name": "hand_right_joints", "l2": 10.367913, "mean_abs": 0.273296, "relative": 0.817269 }, { "name": "body_joints", "l2": 6.166885, "mean_abs": 0.09237, "relative": 0.832956 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01159, "mean_abs": 0.001829, "relative": 0.02735 }, { "name": "camera_rotation_matrix", "l2": 2.999917, "mean_abs": 0.197617, "relative": 0.921226 }, { "name": "imu_accel_gyro", "l2": 17.069086, "mean_abs": 1.027339, "relative": 0.920936 }, { "name": "depth_confidence", "l2": 10.657366, "mean_abs": 0.193343, "relative": 0.000354 }, { "name": "video_fisheye_cam0", "l2": 5.213004, "mean_abs": 0.107554, "relative": 0.943964 }, { "name": "video_fisheye_cam1", "l2": 5.511768, "mean_abs": 0.115576, "relative": 0.910773 }, { "name": "video_fisheye_cam2", "l2": 5.749675, "mean_abs": 0.117051, "relative": 0.91785 }, { "name": "video_fisheye_cam3", "l2": 5.532322, "mean_abs": 0.111282, "relative": 0.926321 }, { "name": "video_stereo_left", "l2": 5.624167, "mean_abs": 0.122628, "relative": 0.913938 }, { "name": "video_stereo_right", "l2": 6.886197, "mean_abs": 0.148749, "relative": 0.970809 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.089296, "mean_abs": 0.084514, "relative": 0.747728 }, { "name": "caption_objects_interaction_text", "l2": 1.816072, "mean_abs": 0.012146, "relative": 0.833896 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7122005820274353" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.8956818580627441" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.7340475916862488" }, "next_action": { "true": "Close bottle cap", "predicted": "Hold coffee carafe", "correct": 0, "confidence": "0.21699179708957672" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|scale|bottle|gooseneck kettle|coffee filter|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 830, "start_frame": 4150, "end_frame": 4169, "center_frame": 4159, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.560312, "mean_abs": 0.257525, "relative": 0.890807 }, { "name": "hand_right_joints", "l2": 10.362153, "mean_abs": 0.272476, "relative": 0.816815 }, { "name": "body_joints", "l2": 6.162789, "mean_abs": 0.092206, "relative": 0.832403 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01103, "mean_abs": 0.001818, "relative": 0.026029 }, { "name": "camera_rotation_matrix", "l2": 3.000391, "mean_abs": 0.19718, "relative": 0.921371 }, { "name": "imu_accel_gyro", "l2": 17.064125, "mean_abs": 1.027715, "relative": 0.920668 }, { "name": "depth_confidence", "l2": 10.661658, "mean_abs": 0.193087, "relative": 0.000354 }, { "name": "video_fisheye_cam0", "l2": 5.207198, "mean_abs": 0.107302, "relative": 0.942912 }, { "name": "video_fisheye_cam1", "l2": 5.492819, "mean_abs": 0.115125, "relative": 0.907642 }, { "name": "video_fisheye_cam2", "l2": 5.738678, "mean_abs": 0.116513, "relative": 0.916095 }, { "name": "video_fisheye_cam3", "l2": 5.517481, "mean_abs": 0.111353, "relative": 0.923836 }, { "name": "video_stereo_left", "l2": 5.610978, "mean_abs": 0.122024, "relative": 0.911795 }, { "name": "video_stereo_right", "l2": 6.871476, "mean_abs": 0.148244, "relative": 0.968734 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.144722, "mean_abs": 0.089853, "relative": 0.767564 }, { "name": "caption_objects_interaction_text", "l2": 1.816071, "mean_abs": 0.012341, "relative": 0.833896 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7186502814292908" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.8819759488105774" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.9655812978744507" }, "next_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.2309247851371765" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|bottle|gooseneck kettle|digital scale|coffee filter|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 831, "start_frame": 4155, "end_frame": 4174, "center_frame": 4164, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.567039, "mean_abs": 0.257561, "relative": 0.891375 }, { "name": "hand_right_joints", "l2": 10.36358, "mean_abs": 0.273224, "relative": 0.816927 }, { "name": "body_joints", "l2": 6.152116, "mean_abs": 0.092564, "relative": 0.830961 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010611, "mean_abs": 0.001319, "relative": 0.025041 }, { "name": "camera_rotation_matrix", "l2": 3.000502, "mean_abs": 0.197648, "relative": 0.921406 }, { "name": "imu_accel_gyro", "l2": 17.064243, "mean_abs": 1.034336, "relative": 0.920675 }, { "name": "depth_confidence", "l2": 10.650648, "mean_abs": 0.192716, "relative": 0.000353 }, { "name": "video_fisheye_cam0", "l2": 5.20455, "mean_abs": 0.107146, "relative": 0.942433 }, { "name": "video_fisheye_cam1", "l2": 5.478523, "mean_abs": 0.114356, "relative": 0.90528 }, { "name": "video_fisheye_cam2", "l2": 5.735628, "mean_abs": 0.116283, "relative": 0.915608 }, { "name": "video_fisheye_cam3", "l2": 5.513717, "mean_abs": 0.11092, "relative": 0.923205 }, { "name": "video_stereo_left", "l2": 5.603027, "mean_abs": 0.121379, "relative": 0.910503 }, { "name": "video_stereo_right", "l2": 6.869415, "mean_abs": 0.147639, "relative": 0.968443 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.098488, "mean_abs": 0.084165, "relative": 0.751017 }, { "name": "caption_objects_interaction_text", "l2": 1.816071, "mean_abs": 0.012319, "relative": 0.833896 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6736355423927307" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.8453457951545715" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9105895757675171" }, "next_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.2073867917060852" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|bottle|gooseneck kettle|digital scale|coffee filter|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 832, "start_frame": 4160, "end_frame": 4179, "center_frame": 4169, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.563662, "mean_abs": 0.256915, "relative": 0.89109 }, { "name": "hand_right_joints", "l2": 10.34407, "mean_abs": 0.272699, "relative": 0.815389 }, { "name": "body_joints", "l2": 6.148794, "mean_abs": 0.092779, "relative": 0.830513 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015506, "mean_abs": 0.002149, "relative": 0.036592 }, { "name": "camera_rotation_matrix", "l2": 3.00026, "mean_abs": 0.198968, "relative": 0.921331 }, { "name": "imu_accel_gyro", "l2": 17.081198, "mean_abs": 1.0407, "relative": 0.921589 }, { "name": "depth_confidence", "l2": 10.654692, "mean_abs": 0.193483, "relative": 0.000354 }, { "name": "video_fisheye_cam0", "l2": 5.198123, "mean_abs": 0.107414, "relative": 0.941269 }, { "name": "video_fisheye_cam1", "l2": 5.453759, "mean_abs": 0.114155, "relative": 0.901188 }, { "name": "video_fisheye_cam2", "l2": 5.727551, "mean_abs": 0.116519, "relative": 0.914318 }, { "name": "video_fisheye_cam3", "l2": 5.485658, "mean_abs": 0.110992, "relative": 0.918507 }, { "name": "video_stereo_left", "l2": 5.579762, "mean_abs": 0.121929, "relative": 0.906722 }, { "name": "video_stereo_right", "l2": 6.83981, "mean_abs": 0.148125, "relative": 0.96427 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.335819, "mean_abs": 0.095566, "relative": 0.835954 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5904861688613892" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.5268679261207581" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9507200717926025" }, "next_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.2594967484474182" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|bottle|gooseneck kettle|digital scale|coffee filter|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 833, "start_frame": 4165, "end_frame": 4184, "center_frame": 4174, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.556828, "mean_abs": 0.256869, "relative": 0.890513 }, { "name": "hand_right_joints", "l2": 10.343952, "mean_abs": 0.273168, "relative": 0.81538 }, { "name": "body_joints", "l2": 6.137893, "mean_abs": 0.092583, "relative": 0.82904 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014333, "mean_abs": 0.002114, "relative": 0.033825 }, { "name": "camera_rotation_matrix", "l2": 2.99987, "mean_abs": 0.198992, "relative": 0.921211 }, { "name": "imu_accel_gyro", "l2": 17.047222, "mean_abs": 1.041908, "relative": 0.919756 }, { "name": "depth_confidence", "l2": 10.662775, "mean_abs": 0.193897, "relative": 0.000354 }, { "name": "video_fisheye_cam0", "l2": 5.195642, "mean_abs": 0.107332, "relative": 0.94082 }, { "name": "video_fisheye_cam1", "l2": 5.436947, "mean_abs": 0.113778, "relative": 0.89841 }, { "name": "video_fisheye_cam2", "l2": 5.724536, "mean_abs": 0.116365, "relative": 0.913837 }, { "name": "video_fisheye_cam3", "l2": 5.474015, "mean_abs": 0.110832, "relative": 0.916558 }, { "name": "video_stereo_left", "l2": 5.565953, "mean_abs": 0.121975, "relative": 0.904478 }, { "name": "video_stereo_right", "l2": 6.81161, "mean_abs": 0.14843, "relative": 0.960294 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.219394, "mean_abs": 0.092101, "relative": 0.794288 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7091046571731567" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.509355366230011" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9898743629455566" }, "next_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.2391209602355957" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 834, "start_frame": 4170, "end_frame": 4189, "center_frame": 4179, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.572764, "mean_abs": 0.257287, "relative": 0.891858 }, { "name": "hand_right_joints", "l2": 10.366531, "mean_abs": 0.27466, "relative": 0.81716 }, { "name": "body_joints", "l2": 6.109059, "mean_abs": 0.092443, "relative": 0.825146 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015249, "mean_abs": 0.002252, "relative": 0.035985 }, { "name": "camera_rotation_matrix", "l2": 2.999384, "mean_abs": 0.19936, "relative": 0.921062 }, { "name": "imu_accel_gyro", "l2": 17.050869, "mean_abs": 1.039598, "relative": 0.919953 }, { "name": "depth_confidence", "l2": 10.688875, "mean_abs": 0.194999, "relative": 0.000355 }, { "name": "video_fisheye_cam0", "l2": 5.194396, "mean_abs": 0.107302, "relative": 0.940594 }, { "name": "video_fisheye_cam1", "l2": 5.427195, "mean_abs": 0.113401, "relative": 0.896798 }, { "name": "video_fisheye_cam2", "l2": 5.730403, "mean_abs": 0.116412, "relative": 0.914774 }, { "name": "video_fisheye_cam3", "l2": 5.46198, "mean_abs": 0.110435, "relative": 0.914543 }, { "name": "video_stereo_left", "l2": 5.555732, "mean_abs": 0.121408, "relative": 0.902817 }, { "name": "video_stereo_right", "l2": 6.78407, "mean_abs": 0.147791, "relative": 0.956411 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.45007, "mean_abs": 0.104861, "relative": 0.876843 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6469100117683411" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.5135206580162048" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.964396595954895" }, "next_action": { "true": "Close bottle cap", "predicted": "Hold coffee carafe", "correct": 0, "confidence": "0.24980054795742035" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|coffee filter|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 835, "start_frame": 4175, "end_frame": 4194, "center_frame": 4184, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.562834, "mean_abs": 0.256935, "relative": 0.89102 }, { "name": "hand_right_joints", "l2": 10.322125, "mean_abs": 0.272741, "relative": 0.81366 }, { "name": "body_joints", "l2": 6.05423, "mean_abs": 0.092461, "relative": 0.81774 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013275, "mean_abs": 0.001968, "relative": 0.031328 }, { "name": "camera_rotation_matrix", "l2": 2.999208, "mean_abs": 0.199486, "relative": 0.921008 }, { "name": "imu_accel_gyro", "l2": 17.057632, "mean_abs": 1.03825, "relative": 0.920318 }, { "name": "depth_confidence", "l2": 10.690269, "mean_abs": 0.19442, "relative": 0.000355 }, { "name": "video_fisheye_cam0", "l2": 5.191318, "mean_abs": 0.107127, "relative": 0.940037 }, { "name": "video_fisheye_cam1", "l2": 5.415555, "mean_abs": 0.113108, "relative": 0.894875 }, { "name": "video_fisheye_cam2", "l2": 5.731268, "mean_abs": 0.116358, "relative": 0.914912 }, { "name": "video_fisheye_cam3", "l2": 5.438511, "mean_abs": 0.109786, "relative": 0.910613 }, { "name": "video_stereo_left", "l2": 5.542345, "mean_abs": 0.120395, "relative": 0.900642 }, { "name": "video_stereo_right", "l2": 6.756955, "mean_abs": 0.146053, "relative": 0.952589 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.504895, "mean_abs": 0.11077, "relative": 0.896464 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7691445350646973" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.5844142436981201" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9429330825805664" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.29315507411956787" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|coffee filter|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 836, "start_frame": 4180, "end_frame": 4199, "center_frame": 4189, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.557538, "mean_abs": 0.254135, "relative": 0.890573 }, { "name": "hand_right_joints", "l2": 10.31321, "mean_abs": 0.269482, "relative": 0.812957 }, { "name": "body_joints", "l2": 6.022619, "mean_abs": 0.091641, "relative": 0.81347 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013682, "mean_abs": 0.00211, "relative": 0.032289 }, { "name": "camera_rotation_matrix", "l2": 3.001999, "mean_abs": 0.204186, "relative": 0.921865 }, { "name": "imu_accel_gyro", "l2": 17.071852, "mean_abs": 1.083893, "relative": 0.921085 }, { "name": "depth_confidence", "l2": 10.745451, "mean_abs": 0.197413, "relative": 0.000357 }, { "name": "video_fisheye_cam0", "l2": 5.195372, "mean_abs": 0.108126, "relative": 0.940771 }, { "name": "video_fisheye_cam1", "l2": 5.409544, "mean_abs": 0.113789, "relative": 0.893881 }, { "name": "video_fisheye_cam2", "l2": 5.765425, "mean_abs": 0.118351, "relative": 0.920364 }, { "name": "video_fisheye_cam3", "l2": 5.41939, "mean_abs": 0.110013, "relative": 0.907411 }, { "name": "video_stereo_left", "l2": 5.569979, "mean_abs": 0.12421, "relative": 0.905133 }, { "name": "video_stereo_right", "l2": 6.766709, "mean_abs": 0.149718, "relative": 0.953964 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.423864, "mean_abs": 0.112492, "relative": 0.867464 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.4380119740962982" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9496958255767822" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999499320983887" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.7440385222434998" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|white mug", "correct": 0, "confidence": "" } } }, { "window_index": 837, "start_frame": 4185, "end_frame": 4204, "center_frame": 4194, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.609956, "mean_abs": 0.257119, "relative": 0.894995 }, { "name": "hand_right_joints", "l2": 10.329334, "mean_abs": 0.270079, "relative": 0.814228 }, { "name": "body_joints", "l2": 5.988678, "mean_abs": 0.089776, "relative": 0.808886 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021596, "mean_abs": 0.003107, "relative": 0.050963 }, { "name": "camera_rotation_matrix", "l2": 3.005815, "mean_abs": 0.210698, "relative": 0.923037 }, { "name": "imu_accel_gyro", "l2": 17.157749, "mean_abs": 1.142448, "relative": 0.92572 }, { "name": "depth_confidence", "l2": 10.979686, "mean_abs": 0.204247, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.205733, "mean_abs": 0.10982, "relative": 0.942647 }, { "name": "video_fisheye_cam1", "l2": 5.422005, "mean_abs": 0.116285, "relative": 0.895941 }, { "name": "video_fisheye_cam2", "l2": 5.814398, "mean_abs": 0.121066, "relative": 0.928182 }, { "name": "video_fisheye_cam3", "l2": 5.419411, "mean_abs": 0.111225, "relative": 0.907415 }, { "name": "video_stereo_left", "l2": 5.6544, "mean_abs": 0.129965, "relative": 0.918851 }, { "name": "video_stereo_right", "l2": 6.829704, "mean_abs": 0.156409, "relative": 0.962845 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.309916, "mean_abs": 0.105355, "relative": 0.826684 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.7379541993141174" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9995589852333069" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.8875564336776733" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|bottle", "correct": 0, "confidence": "" } } }, { "window_index": 838, "start_frame": 4190, "end_frame": 4209, "center_frame": 4199, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.622873, "mean_abs": 0.256885, "relative": 0.896085 }, { "name": "hand_right_joints", "l2": 10.325912, "mean_abs": 0.269979, "relative": 0.813958 }, { "name": "body_joints", "l2": 5.961634, "mean_abs": 0.088381, "relative": 0.805233 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.02286, "mean_abs": 0.003204, "relative": 0.053946 }, { "name": "camera_rotation_matrix", "l2": 3.00446, "mean_abs": 0.210566, "relative": 0.922621 }, { "name": "imu_accel_gyro", "l2": 17.123758, "mean_abs": 1.142066, "relative": 0.923886 }, { "name": "depth_confidence", "l2": 11.014634, "mean_abs": 0.204213, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.204586, "mean_abs": 0.109687, "relative": 0.942439 }, { "name": "video_fisheye_cam1", "l2": 5.423243, "mean_abs": 0.116186, "relative": 0.896145 }, { "name": "video_fisheye_cam2", "l2": 5.819675, "mean_abs": 0.121103, "relative": 0.929025 }, { "name": "video_fisheye_cam3", "l2": 5.417471, "mean_abs": 0.111438, "relative": 0.90709 }, { "name": "video_stereo_left", "l2": 5.645061, "mean_abs": 0.129553, "relative": 0.917334 }, { "name": "video_stereo_right", "l2": 6.827666, "mean_abs": 0.156268, "relative": 0.962558 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.438892, "mean_abs": 0.110082, "relative": 0.872843 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.8976097106933594" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9999856948852539" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.970397412776947" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|bottle|digital scale|white mug", "correct": 0, "confidence": "" } } }, { "window_index": 839, "start_frame": 4195, "end_frame": 4214, "center_frame": 4204, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.644394, "mean_abs": 0.257727, "relative": 0.8979 }, { "name": "hand_right_joints", "l2": 10.328558, "mean_abs": 0.269865, "relative": 0.814167 }, { "name": "body_joints", "l2": 5.952583, "mean_abs": 0.087643, "relative": 0.804011 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.026064, "mean_abs": 0.003671, "relative": 0.061507 }, { "name": "camera_rotation_matrix", "l2": 3.003153, "mean_abs": 0.210146, "relative": 0.92222 }, { "name": "imu_accel_gyro", "l2": 17.104443, "mean_abs": 1.145303, "relative": 0.922844 }, { "name": "depth_confidence", "l2": 11.066998, "mean_abs": 0.20463, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.199351, "mean_abs": 0.109117, "relative": 0.941491 }, { "name": "video_fisheye_cam1", "l2": 5.406269, "mean_abs": 0.115145, "relative": 0.89334 }, { "name": "video_fisheye_cam2", "l2": 5.81517, "mean_abs": 0.120521, "relative": 0.928305 }, { "name": "video_fisheye_cam3", "l2": 5.389151, "mean_abs": 0.109848, "relative": 0.902348 }, { "name": "video_stereo_left", "l2": 5.617095, "mean_abs": 0.127538, "relative": 0.912789 }, { "name": "video_stereo_right", "l2": 6.794806, "mean_abs": 0.153764, "relative": 0.957925 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.276884, "mean_abs": 0.100888, "relative": 0.814863 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9307690262794495" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9999332427978516" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9794694185256958" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|bottle|digital scale|white mug", "correct": 0, "confidence": "" } } }, { "window_index": 840, "start_frame": 4200, "end_frame": 4219, "center_frame": 4209, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.666551, "mean_abs": 0.257211, "relative": 0.899769 }, { "name": "hand_right_joints", "l2": 10.336662, "mean_abs": 0.269217, "relative": 0.814806 }, { "name": "body_joints", "l2": 5.94473, "mean_abs": 0.087937, "relative": 0.80295 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016179, "mean_abs": 0.002554, "relative": 0.03818 }, { "name": "camera_rotation_matrix", "l2": 3.000756, "mean_abs": 0.208542, "relative": 0.921483 }, { "name": "imu_accel_gyro", "l2": 17.168892, "mean_abs": 1.150694, "relative": 0.926321 }, { "name": "depth_confidence", "l2": 11.174448, "mean_abs": 0.201975, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.197959, "mean_abs": 0.107913, "relative": 0.941239 }, { "name": "video_fisheye_cam1", "l2": 5.415374, "mean_abs": 0.113691, "relative": 0.894845 }, { "name": "video_fisheye_cam2", "l2": 5.829088, "mean_abs": 0.119297, "relative": 0.930527 }, { "name": "video_fisheye_cam3", "l2": 5.380507, "mean_abs": 0.108223, "relative": 0.900901 }, { "name": "video_stereo_left", "l2": 5.584678, "mean_abs": 0.121988, "relative": 0.907521 }, { "name": "video_stereo_right", "l2": 6.784697, "mean_abs": 0.147972, "relative": 0.9565 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.236481, "mean_abs": 0.095885, "relative": 0.800403 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9344055652618408" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9996405839920044" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9605967402458191" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|bottle|gooseneck kettle|digital scale|white mug", "correct": 0, "confidence": "" } } }, { "window_index": 841, "start_frame": 4205, "end_frame": 4224, "center_frame": 4214, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.662348, "mean_abs": 0.257464, "relative": 0.899415 }, { "name": "hand_right_joints", "l2": 10.33667, "mean_abs": 0.26998, "relative": 0.814806 }, { "name": "body_joints", "l2": 5.939758, "mean_abs": 0.087857, "relative": 0.802278 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012091, "mean_abs": 0.001649, "relative": 0.028533 }, { "name": "camera_rotation_matrix", "l2": 3.001093, "mean_abs": 0.209498, "relative": 0.921587 }, { "name": "imu_accel_gyro", "l2": 17.105658, "mean_abs": 1.115148, "relative": 0.922909 }, { "name": "depth_confidence", "l2": 11.169699, "mean_abs": 0.202165, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.195561, "mean_abs": 0.107985, "relative": 0.940805 }, { "name": "video_fisheye_cam1", "l2": 5.433384, "mean_abs": 0.114338, "relative": 0.897821 }, { "name": "video_fisheye_cam2", "l2": 5.803345, "mean_abs": 0.119077, "relative": 0.926418 }, { "name": "video_fisheye_cam3", "l2": 5.365439, "mean_abs": 0.108079, "relative": 0.898378 }, { "name": "video_stereo_left", "l2": 5.539398, "mean_abs": 0.121586, "relative": 0.900163 }, { "name": "video_stereo_right", "l2": 6.798957, "mean_abs": 0.149048, "relative": 0.95851 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.219504, "mean_abs": 0.093494, "relative": 0.794327 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.7654641270637512" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9553611874580383" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999659061431885" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9391448497772217" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|digital scale", "correct": 0, "confidence": "" } } }, { "window_index": 842, "start_frame": 4210, "end_frame": 4229, "center_frame": 4219, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.647676, "mean_abs": 0.255436, "relative": 0.898177 }, { "name": "hand_right_joints", "l2": 10.330379, "mean_abs": 0.268647, "relative": 0.81431 }, { "name": "body_joints", "l2": 5.92607, "mean_abs": 0.087825, "relative": 0.800429 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010407, "mean_abs": 0.001352, "relative": 0.02456 }, { "name": "camera_rotation_matrix", "l2": 3.000802, "mean_abs": 0.207707, "relative": 0.921498 }, { "name": "imu_accel_gyro", "l2": 17.043055, "mean_abs": 1.081121, "relative": 0.919531 }, { "name": "depth_confidence", "l2": 11.141468, "mean_abs": 0.201347, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.201331, "mean_abs": 0.107888, "relative": 0.94185 }, { "name": "video_fisheye_cam1", "l2": 5.457409, "mean_abs": 0.114449, "relative": 0.901791 }, { "name": "video_fisheye_cam2", "l2": 5.756848, "mean_abs": 0.117377, "relative": 0.918995 }, { "name": "video_fisheye_cam3", "l2": 5.410982, "mean_abs": 0.10934, "relative": 0.906004 }, { "name": "video_stereo_left", "l2": 5.481494, "mean_abs": 0.1197, "relative": 0.890754 }, { "name": "video_stereo_right", "l2": 6.82618, "mean_abs": 0.148386, "relative": 0.962348 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.160594, "mean_abs": 0.089237, "relative": 0.773244 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.7821319699287415" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.8388680219650269" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9997296929359436" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9104214906692505" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|digital scale|coffee filter", "correct": 0, "confidence": "" } } }, { "window_index": 843, "start_frame": 4215, "end_frame": 4234, "center_frame": 4224, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.617847, "mean_abs": 0.255799, "relative": 0.895661 }, { "name": "hand_right_joints", "l2": 10.343128, "mean_abs": 0.26979, "relative": 0.815315 }, { "name": "body_joints", "l2": 5.891473, "mean_abs": 0.088172, "relative": 0.795756 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.025379, "mean_abs": 0.003223, "relative": 0.059891 }, { "name": "camera_rotation_matrix", "l2": 3.004917, "mean_abs": 0.215632, "relative": 0.922761 }, { "name": "imu_accel_gyro", "l2": 17.218546, "mean_abs": 1.192335, "relative": 0.929 }, { "name": "depth_confidence", "l2": 11.250519, "mean_abs": 0.208142, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.205856, "mean_abs": 0.108729, "relative": 0.942669 }, { "name": "video_fisheye_cam1", "l2": 5.486627, "mean_abs": 0.116981, "relative": 0.906619 }, { "name": "video_fisheye_cam2", "l2": 5.75103, "mean_abs": 0.119011, "relative": 0.918066 }, { "name": "video_fisheye_cam3", "l2": 5.444158, "mean_abs": 0.110918, "relative": 0.911559 }, { "name": "video_stereo_left", "l2": 5.48174, "mean_abs": 0.12404, "relative": 0.890794 }, { "name": "video_stereo_right", "l2": 6.838649, "mean_abs": 0.153552, "relative": 0.964106 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.278434, "mean_abs": 0.095345, "relative": 0.815417 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9651442170143127" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9918482303619385" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9975402355194092" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9878210425376892" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle", "correct": 0, "confidence": "" } } }, { "window_index": 844, "start_frame": 4220, "end_frame": 4239, "center_frame": 4229, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.584239, "mean_abs": 0.258821, "relative": 0.892826 }, { "name": "hand_right_joints", "l2": 10.350744, "mean_abs": 0.275083, "relative": 0.815916 }, { "name": "body_joints", "l2": 5.828615, "mean_abs": 0.089322, "relative": 0.787266 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.051453, "mean_abs": 0.00599, "relative": 0.121421 }, { "name": "camera_rotation_matrix", "l2": 3.04983, "mean_abs": 0.237059, "relative": 0.936553 }, { "name": "imu_accel_gyro", "l2": 17.476496, "mean_abs": 1.302671, "relative": 0.942917 }, { "name": "depth_confidence", "l2": 11.577638, "mean_abs": 0.223399, "relative": 0.000384 }, { "name": "video_fisheye_cam0", "l2": 5.205985, "mean_abs": 0.111615, "relative": 0.942692 }, { "name": "video_fisheye_cam1", "l2": 5.59776, "mean_abs": 0.124058, "relative": 0.924983 }, { "name": "video_fisheye_cam2", "l2": 5.716945, "mean_abs": 0.122405, "relative": 0.912625 }, { "name": "video_fisheye_cam3", "l2": 5.489296, "mean_abs": 0.11487, "relative": 0.919116 }, { "name": "video_stereo_left", "l2": 5.518218, "mean_abs": 0.130977, "relative": 0.896721 }, { "name": "video_stereo_right", "l2": 6.834883, "mean_abs": 0.16427, "relative": 0.963575 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.148711, "mean_abs": 0.086622, "relative": 0.768991 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9930558800697327" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9999485015869141" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999996423721313" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9974431991577148" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|bottle|gooseneck kettle", "correct": 0, "confidence": "" } } }, { "window_index": 845, "start_frame": 4225, "end_frame": 4244, "center_frame": 4234, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.508518, "mean_abs": 0.259715, "relative": 0.886438 }, { "name": "hand_right_joints", "l2": 10.370075, "mean_abs": 0.274718, "relative": 0.817439 }, { "name": "body_joints", "l2": 5.784206, "mean_abs": 0.090564, "relative": 0.781268 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.066334, "mean_abs": 0.007939, "relative": 0.156538 }, { "name": "camera_rotation_matrix", "l2": 3.079743, "mean_abs": 0.248654, "relative": 0.945739 }, { "name": "imu_accel_gyro", "l2": 18.316507, "mean_abs": 1.380817, "relative": 0.988239 }, { "name": "depth_confidence", "l2": 11.83495, "mean_abs": 0.231026, "relative": 0.000393 }, { "name": "video_fisheye_cam0", "l2": 5.184753, "mean_abs": 0.112794, "relative": 0.938848 }, { "name": "video_fisheye_cam1", "l2": 5.72891, "mean_abs": 0.129916, "relative": 0.946654 }, { "name": "video_fisheye_cam2", "l2": 5.676966, "mean_abs": 0.122979, "relative": 0.906243 }, { "name": "video_fisheye_cam3", "l2": 5.561234, "mean_abs": 0.118246, "relative": 0.931162 }, { "name": "video_stereo_left", "l2": 5.486671, "mean_abs": 0.131343, "relative": 0.891595 }, { "name": "video_stereo_right", "l2": 6.779217, "mean_abs": 0.165198, "relative": 0.955727 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.243234, "mean_abs": 0.091384, "relative": 0.80282 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.5737177133560181" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.999962329864502" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.8879128098487854" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 846, "start_frame": 4230, "end_frame": 4249, "center_frame": 4239, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.405437, "mean_abs": 0.244114, "relative": 0.877743 }, { "name": "hand_right_joints", "l2": 10.312925, "mean_abs": 0.263752, "relative": 0.812934 }, { "name": "body_joints", "l2": 5.754168, "mean_abs": 0.089826, "relative": 0.777211 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.069452, "mean_abs": 0.010131, "relative": 0.163896 }, { "name": "camera_rotation_matrix", "l2": 3.051849, "mean_abs": 0.243803, "relative": 0.937173 }, { "name": "imu_accel_gyro", "l2": 18.013632, "mean_abs": 1.351701, "relative": 0.971897 }, { "name": "depth_confidence", "l2": 11.847945, "mean_abs": 0.228443, "relative": 0.000393 }, { "name": "video_fisheye_cam0", "l2": 5.1728, "mean_abs": 0.111318, "relative": 0.936683 }, { "name": "video_fisheye_cam1", "l2": 5.723353, "mean_abs": 0.128221, "relative": 0.945736 }, { "name": "video_fisheye_cam2", "l2": 5.667109, "mean_abs": 0.122656, "relative": 0.90467 }, { "name": "video_fisheye_cam3", "l2": 5.596273, "mean_abs": 0.118498, "relative": 0.937028 }, { "name": "video_stereo_left", "l2": 5.504786, "mean_abs": 0.132782, "relative": 0.894539 }, { "name": "video_stereo_right", "l2": 6.782817, "mean_abs": 0.164837, "relative": 0.956235 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.214062, "mean_abs": 0.090233, "relative": 0.792379 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9956004619598389" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9999809265136719" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.7693018913269043" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|bottle|digital scale|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 847, "start_frame": 4235, "end_frame": 4254, "center_frame": 4244, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.328405, "mean_abs": 0.228173, "relative": 0.871245 }, { "name": "hand_right_joints", "l2": 10.261957, "mean_abs": 0.24508, "relative": 0.808917 }, { "name": "body_joints", "l2": 5.71894, "mean_abs": 0.087974, "relative": 0.772453 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.064275, "mean_abs": 0.010117, "relative": 0.151679 }, { "name": "camera_rotation_matrix", "l2": 3.018627, "mean_abs": 0.233322, "relative": 0.926971 }, { "name": "imu_accel_gyro", "l2": 17.522614, "mean_abs": 1.270026, "relative": 0.945405 }, { "name": "depth_confidence", "l2": 11.868726, "mean_abs": 0.224362, "relative": 0.000394 }, { "name": "video_fisheye_cam0", "l2": 5.188404, "mean_abs": 0.110549, "relative": 0.939509 }, { "name": "video_fisheye_cam1", "l2": 5.711249, "mean_abs": 0.126185, "relative": 0.943736 }, { "name": "video_fisheye_cam2", "l2": 5.653332, "mean_abs": 0.120388, "relative": 0.90247 }, { "name": "video_fisheye_cam3", "l2": 5.599603, "mean_abs": 0.116929, "relative": 0.937586 }, { "name": "video_stereo_left", "l2": 5.504581, "mean_abs": 0.131046, "relative": 0.894505 }, { "name": "video_stereo_right", "l2": 6.595817, "mean_abs": 0.15644, "relative": 0.929872 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.230245, "mean_abs": 0.091826, "relative": 0.798171 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9999674558639526" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9976537823677063" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9993638396263123" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|bottle|digital scale|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 848, "start_frame": 4240, "end_frame": 4259, "center_frame": 4249, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.376523, "mean_abs": 0.239481, "relative": 0.875304 }, { "name": "hand_right_joints", "l2": 10.219902, "mean_abs": 0.25501, "relative": 0.805602 }, { "name": "body_joints", "l2": 5.706855, "mean_abs": 0.087078, "relative": 0.77082 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.060314, "mean_abs": 0.009793, "relative": 0.142332 }, { "name": "camera_rotation_matrix", "l2": 3.074962, "mean_abs": 0.252525, "relative": 0.944271 }, { "name": "imu_accel_gyro", "l2": 18.254812, "mean_abs": 1.388879, "relative": 0.98491 }, { "name": "depth_confidence", "l2": 11.932097, "mean_abs": 0.231645, "relative": 0.000396 }, { "name": "video_fisheye_cam0", "l2": 5.190797, "mean_abs": 0.113079, "relative": 0.939942 }, { "name": "video_fisheye_cam1", "l2": 5.769567, "mean_abs": 0.130145, "relative": 0.953372 }, { "name": "video_fisheye_cam2", "l2": 5.728677, "mean_abs": 0.124533, "relative": 0.914498 }, { "name": "video_fisheye_cam3", "l2": 5.646369, "mean_abs": 0.119183, "relative": 0.945416 }, { "name": "video_stereo_left", "l2": 5.602871, "mean_abs": 0.134981, "relative": 0.910478 }, { "name": "video_stereo_right", "l2": 6.859281, "mean_abs": 0.16602, "relative": 0.967015 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.218895, "mean_abs": 0.089242, "relative": 0.794109 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9993657469749451" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.999019980430603" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9995285272598267" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|digital scale|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 849, "start_frame": 4245, "end_frame": 4264, "center_frame": 4254, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.486647, "mean_abs": 0.248789, "relative": 0.884593 }, { "name": "hand_right_joints", "l2": 10.215137, "mean_abs": 0.262788, "relative": 0.805226 }, { "name": "body_joints", "l2": 5.708205, "mean_abs": 0.083555, "relative": 0.771003 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.049799, "mean_abs": 0.007473, "relative": 0.117517 }, { "name": "camera_rotation_matrix", "l2": 3.116375, "mean_abs": 0.257918, "relative": 0.956988 }, { "name": "imu_accel_gyro", "l2": 18.534498, "mean_abs": 1.38667, "relative": 1.0 }, { "name": "depth_confidence", "l2": 13.309201, "mean_abs": 0.238638, "relative": 0.000442 }, { "name": "video_fisheye_cam0", "l2": 5.240887, "mean_abs": 0.114804, "relative": 0.949013 }, { "name": "video_fisheye_cam1", "l2": 5.756474, "mean_abs": 0.130662, "relative": 0.951209 }, { "name": "video_fisheye_cam2", "l2": 5.805277, "mean_abs": 0.126458, "relative": 0.926726 }, { "name": "video_fisheye_cam3", "l2": 5.649643, "mean_abs": 0.119974, "relative": 0.945965 }, { "name": "video_stereo_left", "l2": 5.743121, "mean_abs": 0.139302, "relative": 0.933269 }, { "name": "video_stereo_right", "l2": 7.018764, "mean_abs": 0.170829, "relative": 0.989499 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.258443, "mean_abs": 0.086992, "relative": 0.808263 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9848392009735107" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9999651908874512" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Close bottle cap", "correct": 1, "confidence": "0.9965073466300964" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|digital scale|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 850, "start_frame": 4250, "end_frame": 4269, "center_frame": 4259, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.537419, "mean_abs": 0.245565, "relative": 0.888876 }, { "name": "hand_right_joints", "l2": 10.208558, "mean_abs": 0.261852, "relative": 0.804707 }, { "name": "body_joints", "l2": 5.736383, "mean_abs": 0.082703, "relative": 0.774809 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.047016, "mean_abs": 0.007436, "relative": 0.110951 }, { "name": "camera_rotation_matrix", "l2": 3.056969, "mean_abs": 0.243274, "relative": 0.938745 }, { "name": "imu_accel_gyro", "l2": 17.710741, "mean_abs": 1.308107, "relative": 0.955555 }, { "name": "depth_confidence", "l2": 36.780773, "mean_abs": 0.27279, "relative": 0.001221 }, { "name": "video_fisheye_cam0", "l2": 5.248972, "mean_abs": 0.113595, "relative": 0.950477 }, { "name": "video_fisheye_cam1", "l2": 5.59236, "mean_abs": 0.124997, "relative": 0.92409 }, { "name": "video_fisheye_cam2", "l2": 5.809851, "mean_abs": 0.125671, "relative": 0.927456 }, { "name": "video_fisheye_cam3", "l2": 5.605259, "mean_abs": 0.11875, "relative": 0.938533 }, { "name": "video_stereo_left", "l2": 5.782298, "mean_abs": 0.138854, "relative": 0.939635 }, { "name": "video_stereo_right", "l2": 7.029836, "mean_abs": 0.169872, "relative": 0.991059 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.231209, "mean_abs": 0.0906, "relative": 0.798516 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.9740461111068726" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9999885559082031" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.7284150719642639" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|gooseneck kettle|digital scale", "correct": 0, "confidence": "" } } }, { "window_index": 851, "start_frame": 4255, "end_frame": 4274, "center_frame": 4264, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.550271, "mean_abs": 0.243773, "relative": 0.88996 }, { "name": "hand_right_joints", "l2": 10.155677, "mean_abs": 0.257058, "relative": 0.800539 }, { "name": "body_joints", "l2": 5.740611, "mean_abs": 0.082319, "relative": 0.77538 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.064363, "mean_abs": 0.009055, "relative": 0.151887 }, { "name": "camera_rotation_matrix", "l2": 3.016035, "mean_abs": 0.229892, "relative": 0.926175 }, { "name": "imu_accel_gyro", "l2": 17.310463, "mean_abs": 1.195312, "relative": 0.933959 }, { "name": "depth_confidence", "l2": 36.762177, "mean_abs": 0.265763, "relative": 0.00122 }, { "name": "video_fisheye_cam0", "l2": 5.231858, "mean_abs": 0.111056, "relative": 0.947378 }, { "name": "video_fisheye_cam1", "l2": 5.534863, "mean_abs": 0.121584, "relative": 0.914589 }, { "name": "video_fisheye_cam2", "l2": 5.812135, "mean_abs": 0.122979, "relative": 0.927821 }, { "name": "video_fisheye_cam3", "l2": 5.52984, "mean_abs": 0.114517, "relative": 0.925905 }, { "name": "video_stereo_left", "l2": 5.634394, "mean_abs": 0.130167, "relative": 0.9156 }, { "name": "video_stereo_right", "l2": 6.941193, "mean_abs": 0.161198, "relative": 0.978563 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.080188, "mean_abs": 0.08211, "relative": 0.744468 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.9665635824203491" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.999997615814209" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.507900595664978" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|gooseneck kettle|digital scale", "correct": 0, "confidence": "" } } }, { "window_index": 852, "start_frame": 4260, "end_frame": 4279, "center_frame": 4269, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.565022, "mean_abs": 0.243854, "relative": 0.891205 }, { "name": "hand_right_joints", "l2": 10.147719, "mean_abs": 0.255684, "relative": 0.799912 }, { "name": "body_joints", "l2": 5.739736, "mean_abs": 0.082371, "relative": 0.775261 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.046881, "mean_abs": 0.007044, "relative": 0.110633 }, { "name": "camera_rotation_matrix", "l2": 3.005731, "mean_abs": 0.220962, "relative": 0.923011 }, { "name": "imu_accel_gyro", "l2": 17.263664, "mean_abs": 1.159651, "relative": 0.931434 }, { "name": "depth_confidence", "l2": 36.82832, "mean_abs": 0.263504, "relative": 0.001222 }, { "name": "video_fisheye_cam0", "l2": 5.218552, "mean_abs": 0.109213, "relative": 0.944968 }, { "name": "video_fisheye_cam1", "l2": 5.480076, "mean_abs": 0.117376, "relative": 0.905536 }, { "name": "video_fisheye_cam2", "l2": 5.761014, "mean_abs": 0.119292, "relative": 0.91966 }, { "name": "video_fisheye_cam3", "l2": 5.477303, "mean_abs": 0.110953, "relative": 0.917108 }, { "name": "video_stereo_left", "l2": 5.498897, "mean_abs": 0.123296, "relative": 0.893582 }, { "name": "video_stereo_right", "l2": 6.80578, "mean_abs": 0.152507, "relative": 0.959472 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.109125, "mean_abs": 0.082757, "relative": 0.754824 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.46074700355529785" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9998563528060913" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.8078120350837708" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|gooseneck kettle|digital scale|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 853, "start_frame": 4265, "end_frame": 4284, "center_frame": 4274, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.572693, "mean_abs": 0.245787, "relative": 0.891852 }, { "name": "hand_right_joints", "l2": 10.145485, "mean_abs": 0.256366, "relative": 0.799736 }, { "name": "body_joints", "l2": 5.744169, "mean_abs": 0.082408, "relative": 0.77586 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.038298, "mean_abs": 0.006012, "relative": 0.090377 }, { "name": "camera_rotation_matrix", "l2": 3.00287, "mean_abs": 0.218844, "relative": 0.922133 }, { "name": "imu_accel_gyro", "l2": 17.248005, "mean_abs": 1.157937, "relative": 0.930589 }, { "name": "depth_confidence", "l2": 36.705135, "mean_abs": 0.262261, "relative": 0.001218 }, { "name": "video_fisheye_cam0", "l2": 5.216155, "mean_abs": 0.109162, "relative": 0.944534 }, { "name": "video_fisheye_cam1", "l2": 5.473188, "mean_abs": 0.117012, "relative": 0.904398 }, { "name": "video_fisheye_cam2", "l2": 5.712602, "mean_abs": 0.118517, "relative": 0.911932 }, { "name": "video_fisheye_cam3", "l2": 5.484599, "mean_abs": 0.111263, "relative": 0.91833 }, { "name": "video_stereo_left", "l2": 5.477195, "mean_abs": 0.123806, "relative": 0.890055 }, { "name": "video_stereo_right", "l2": 6.848845, "mean_abs": 0.155184, "relative": 0.965544 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.112119, "mean_abs": 0.084461, "relative": 0.755896 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.39641040563583374" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.9959465861320496" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999997615814209" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.8129513263702393" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|gooseneck kettle|digital scale|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 854, "start_frame": 4270, "end_frame": 4289, "center_frame": 4279, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.562221, "mean_abs": 0.243647, "relative": 0.890968 }, { "name": "hand_right_joints", "l2": 10.148591, "mean_abs": 0.255164, "relative": 0.79998 }, { "name": "body_joints", "l2": 5.75275, "mean_abs": 0.082849, "relative": 0.777019 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028112, "mean_abs": 0.004445, "relative": 0.06634 }, { "name": "camera_rotation_matrix", "l2": 3.000766, "mean_abs": 0.213307, "relative": 0.921487 }, { "name": "imu_accel_gyro", "l2": 17.155882, "mean_abs": 1.116042, "relative": 0.925619 }, { "name": "depth_confidence", "l2": 11.549273, "mean_abs": 0.207565, "relative": 0.000383 }, { "name": "video_fisheye_cam0", "l2": 5.216472, "mean_abs": 0.108676, "relative": 0.944592 }, { "name": "video_fisheye_cam1", "l2": 5.469226, "mean_abs": 0.115588, "relative": 0.903743 }, { "name": "video_fisheye_cam2", "l2": 5.646054, "mean_abs": 0.116071, "relative": 0.901308 }, { "name": "video_fisheye_cam3", "l2": 5.500138, "mean_abs": 0.111174, "relative": 0.920932 }, { "name": "video_stereo_left", "l2": 5.393175, "mean_abs": 0.11984, "relative": 0.876402 }, { "name": "video_stereo_right", "l2": 6.846036, "mean_abs": 0.15194, "relative": 0.965147 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.095412, "mean_abs": 0.086092, "relative": 0.749917 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.7554227709770203" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.8667193055152893" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999946355819702" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.8587966561317444" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|digital scale|table", "correct": 0, "confidence": "" } } }, { "window_index": 855, "start_frame": 4275, "end_frame": 4294, "center_frame": 4284, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.554305, "mean_abs": 0.242119, "relative": 0.890301 }, { "name": "hand_right_joints", "l2": 10.151253, "mean_abs": 0.256691, "relative": 0.80019 }, { "name": "body_joints", "l2": 5.773319, "mean_abs": 0.083951, "relative": 0.779797 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016446, "mean_abs": 0.002503, "relative": 0.038811 }, { "name": "camera_rotation_matrix", "l2": 3.000147, "mean_abs": 0.211003, "relative": 0.921296 }, { "name": "imu_accel_gyro", "l2": 17.093792, "mean_abs": 1.100614, "relative": 0.922269 }, { "name": "depth_confidence", "l2": 11.300899, "mean_abs": 0.204474, "relative": 0.000375 }, { "name": "video_fisheye_cam0", "l2": 5.214342, "mean_abs": 0.108311, "relative": 0.944206 }, { "name": "video_fisheye_cam1", "l2": 5.46574, "mean_abs": 0.114641, "relative": 0.903167 }, { "name": "video_fisheye_cam2", "l2": 5.628612, "mean_abs": 0.115331, "relative": 0.898524 }, { "name": "video_fisheye_cam3", "l2": 5.498835, "mean_abs": 0.110766, "relative": 0.920714 }, { "name": "video_stereo_left", "l2": 5.367307, "mean_abs": 0.118071, "relative": 0.872198 }, { "name": "video_stereo_right", "l2": 6.855346, "mean_abs": 0.149766, "relative": 0.96646 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.336379, "mean_abs": 0.101141, "relative": 0.836155 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.8473094701766968" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.8663994073867798" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999779462814331" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.8798666596412659" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|digital scale|table", "correct": 0, "confidence": "" } } }, { "window_index": 856, "start_frame": 4280, "end_frame": 4299, "center_frame": 4289, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.535436, "mean_abs": 0.245837, "relative": 0.888709 }, { "name": "hand_right_joints", "l2": 10.155782, "mean_abs": 0.260053, "relative": 0.800547 }, { "name": "body_joints", "l2": 5.807495, "mean_abs": 0.085418, "relative": 0.784414 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01386, "mean_abs": 0.002119, "relative": 0.032707 }, { "name": "camera_rotation_matrix", "l2": 3.000379, "mean_abs": 0.210044, "relative": 0.921368 }, { "name": "imu_accel_gyro", "l2": 17.080175, "mean_abs": 1.098289, "relative": 0.921534 }, { "name": "depth_confidence", "l2": 11.253001, "mean_abs": 0.20349, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.215529, "mean_abs": 0.108197, "relative": 0.944421 }, { "name": "video_fisheye_cam1", "l2": 5.459223, "mean_abs": 0.114257, "relative": 0.902091 }, { "name": "video_fisheye_cam2", "l2": 5.619093, "mean_abs": 0.114846, "relative": 0.897005 }, { "name": "video_fisheye_cam3", "l2": 5.49598, "mean_abs": 0.110535, "relative": 0.920236 }, { "name": "video_stereo_left", "l2": 5.366526, "mean_abs": 0.117754, "relative": 0.872071 }, { "name": "video_stereo_right", "l2": 6.86532, "mean_abs": 0.149123, "relative": 0.967866 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.380881, "mean_abs": 0.102431, "relative": 0.852082 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.933936595916748" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.882402777671814" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999951124191284" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9424421787261963" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|digital scale|coffee filter", "correct": 0, "confidence": "" } } }, { "window_index": 857, "start_frame": 4285, "end_frame": 4304, "center_frame": 4294, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.536236, "mean_abs": 0.247357, "relative": 0.888776 }, { "name": "hand_right_joints", "l2": 10.159563, "mean_abs": 0.260825, "relative": 0.800845 }, { "name": "body_joints", "l2": 5.836706, "mean_abs": 0.087034, "relative": 0.788359 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009941, "mean_abs": 0.001587, "relative": 0.023459 }, { "name": "camera_rotation_matrix", "l2": 3.000319, "mean_abs": 0.208162, "relative": 0.921349 }, { "name": "imu_accel_gyro", "l2": 17.072371, "mean_abs": 1.08721, "relative": 0.921113 }, { "name": "depth_confidence", "l2": 11.170074, "mean_abs": 0.201138, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.213725, "mean_abs": 0.107891, "relative": 0.944094 }, { "name": "video_fisheye_cam1", "l2": 5.456725, "mean_abs": 0.113923, "relative": 0.901678 }, { "name": "video_fisheye_cam2", "l2": 5.610647, "mean_abs": 0.114359, "relative": 0.895656 }, { "name": "video_fisheye_cam3", "l2": 5.494803, "mean_abs": 0.110255, "relative": 0.920039 }, { "name": "video_stereo_left", "l2": 5.36808, "mean_abs": 0.116603, "relative": 0.872324 }, { "name": "video_stereo_right", "l2": 6.864576, "mean_abs": 0.148326, "relative": 0.967761 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.34899, "mean_abs": 0.102442, "relative": 0.840668 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9334571361541748" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.8997464776039124" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999935626983643" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9417603611946106" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|digital scale|white mug", "correct": 0, "confidence": "" } } }, { "window_index": 858, "start_frame": 4290, "end_frame": 4309, "center_frame": 4299, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.541792, "mean_abs": 0.248818, "relative": 0.889245 }, { "name": "hand_right_joints", "l2": 10.181931, "mean_abs": 0.261939, "relative": 0.802609 }, { "name": "body_joints", "l2": 5.87822, "mean_abs": 0.087878, "relative": 0.793966 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011131, "mean_abs": 0.001725, "relative": 0.026267 }, { "name": "camera_rotation_matrix", "l2": 3.000155, "mean_abs": 0.207164, "relative": 0.921299 }, { "name": "imu_accel_gyro", "l2": 17.099684, "mean_abs": 1.08387, "relative": 0.922587 }, { "name": "depth_confidence", "l2": 11.139703, "mean_abs": 0.200324, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.212327, "mean_abs": 0.107686, "relative": 0.943841 }, { "name": "video_fisheye_cam1", "l2": 5.452992, "mean_abs": 0.113679, "relative": 0.901061 }, { "name": "video_fisheye_cam2", "l2": 5.602654, "mean_abs": 0.113882, "relative": 0.89438 }, { "name": "video_fisheye_cam3", "l2": 5.494909, "mean_abs": 0.110046, "relative": 0.920056 }, { "name": "video_stereo_left", "l2": 5.3553, "mean_abs": 0.1155, "relative": 0.870247 }, { "name": "video_stereo_right", "l2": 6.857597, "mean_abs": 0.147233, "relative": 0.966777 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.410173, "mean_abs": 0.103893, "relative": 0.862565 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9559752345085144" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.8988319635391235" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999700784683228" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9578930139541626" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|white mug", "correct": 0, "confidence": "" } } }, { "window_index": 859, "start_frame": 4295, "end_frame": 4314, "center_frame": 4304, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.540359, "mean_abs": 0.250693, "relative": 0.889124 }, { "name": "hand_right_joints", "l2": 10.201945, "mean_abs": 0.264316, "relative": 0.804186 }, { "name": "body_joints", "l2": 5.947584, "mean_abs": 0.089654, "relative": 0.803335 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011821, "mean_abs": 0.001886, "relative": 0.027896 }, { "name": "camera_rotation_matrix", "l2": 3.000025, "mean_abs": 0.207029, "relative": 0.921259 }, { "name": "imu_accel_gyro", "l2": 17.101585, "mean_abs": 1.086365, "relative": 0.922689 }, { "name": "depth_confidence", "l2": 11.122416, "mean_abs": 0.200789, "relative": 0.000369 }, { "name": "video_fisheye_cam0", "l2": 5.214654, "mean_abs": 0.10786, "relative": 0.944262 }, { "name": "video_fisheye_cam1", "l2": 5.452598, "mean_abs": 0.113687, "relative": 0.900996 }, { "name": "video_fisheye_cam2", "l2": 5.597106, "mean_abs": 0.11404, "relative": 0.893495 }, { "name": "video_fisheye_cam3", "l2": 5.504066, "mean_abs": 0.110416, "relative": 0.921589 }, { "name": "video_stereo_left", "l2": 5.354443, "mean_abs": 0.115917, "relative": 0.870108 }, { "name": "video_stereo_right", "l2": 6.857575, "mean_abs": 0.147712, "relative": 0.966774 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.299243, "mean_abs": 0.093453, "relative": 0.822865 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9457948207855225" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.8254587054252625" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999991655349731" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9478023648262024" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|digital scale|white mug", "correct": 0, "confidence": "" } } }, { "window_index": 860, "start_frame": 4300, "end_frame": 4319, "center_frame": 4309, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.543394, "mean_abs": 0.250091, "relative": 0.88938 }, { "name": "hand_right_joints", "l2": 10.23122, "mean_abs": 0.264889, "relative": 0.806494 }, { "name": "body_joints", "l2": 5.97369, "mean_abs": 0.090104, "relative": 0.806861 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081672, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017507, "mean_abs": 0.002441, "relative": 0.041314 }, { "name": "camera_rotation_matrix", "l2": 3.001146, "mean_abs": 0.208022, "relative": 0.921603 }, { "name": "imu_accel_gyro", "l2": 17.103975, "mean_abs": 1.091426, "relative": 0.922818 }, { "name": "depth_confidence", "l2": 11.055094, "mean_abs": 0.200497, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.219841, "mean_abs": 0.108156, "relative": 0.945202 }, { "name": "video_fisheye_cam1", "l2": 5.461356, "mean_abs": 0.114376, "relative": 0.902443 }, { "name": "video_fisheye_cam2", "l2": 5.585713, "mean_abs": 0.114466, "relative": 0.891676 }, { "name": "video_fisheye_cam3", "l2": 5.518703, "mean_abs": 0.110739, "relative": 0.92404 }, { "name": "video_stereo_left", "l2": 5.36204, "mean_abs": 0.117192, "relative": 0.871342 }, { "name": "video_stereo_right", "l2": 6.871001, "mean_abs": 0.14888, "relative": 0.968667 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.292517, "mean_abs": 0.093793, "relative": 0.820457 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9580444097518921" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.8718734383583069" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999760389328003" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9628722071647644" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|digital scale|coffee filter", "correct": 0, "confidence": "" } } }, { "window_index": 861, "start_frame": 4305, "end_frame": 4324, "center_frame": 4314, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.54051, "mean_abs": 0.24985, "relative": 0.889137 }, { "name": "hand_right_joints", "l2": 10.231681, "mean_abs": 0.264393, "relative": 0.80653 }, { "name": "body_joints", "l2": 5.958623, "mean_abs": 0.090304, "relative": 0.804826 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081672, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.024535, "mean_abs": 0.003433, "relative": 0.0579 }, { "name": "camera_rotation_matrix", "l2": 3.001693, "mean_abs": 0.208615, "relative": 0.921771 }, { "name": "imu_accel_gyro", "l2": 17.13341, "mean_abs": 1.103275, "relative": 0.924406 }, { "name": "depth_confidence", "l2": 11.018432, "mean_abs": 0.200888, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.223829, "mean_abs": 0.108546, "relative": 0.945924 }, { "name": "video_fisheye_cam1", "l2": 5.471834, "mean_abs": 0.115016, "relative": 0.904174 }, { "name": "video_fisheye_cam2", "l2": 5.578982, "mean_abs": 0.114895, "relative": 0.890602 }, { "name": "video_fisheye_cam3", "l2": 5.535854, "mean_abs": 0.111212, "relative": 0.926912 }, { "name": "video_stereo_left", "l2": 5.362234, "mean_abs": 0.117837, "relative": 0.871374 }, { "name": "video_stereo_right", "l2": 6.875697, "mean_abs": 0.149715, "relative": 0.969329 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.442021, "mean_abs": 0.105064, "relative": 0.873963 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.8951300382614136" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.8687889575958252" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.999985933303833" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.9382264018058777" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 862, "start_frame": 4310, "end_frame": 4329, "center_frame": 4319, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.53884, "mean_abs": 0.249327, "relative": 0.888996 }, { "name": "hand_right_joints", "l2": 10.237764, "mean_abs": 0.26471, "relative": 0.80701 }, { "name": "body_joints", "l2": 5.954464, "mean_abs": 0.090268, "relative": 0.804265 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.026252, "mean_abs": 0.003777, "relative": 0.06195 }, { "name": "camera_rotation_matrix", "l2": 3.000886, "mean_abs": 0.207434, "relative": 0.921523 }, { "name": "imu_accel_gyro", "l2": 17.107094, "mean_abs": 1.09903, "relative": 0.922987 }, { "name": "depth_confidence", "l2": 10.987913, "mean_abs": 0.199974, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.224496, "mean_abs": 0.108395, "relative": 0.946044 }, { "name": "video_fisheye_cam1", "l2": 5.474411, "mean_abs": 0.115001, "relative": 0.9046 }, { "name": "video_fisheye_cam2", "l2": 5.573399, "mean_abs": 0.114531, "relative": 0.88971 }, { "name": "video_fisheye_cam3", "l2": 5.541226, "mean_abs": 0.111303, "relative": 0.927811 }, { "name": "video_stereo_left", "l2": 5.356569, "mean_abs": 0.117487, "relative": 0.870453 }, { "name": "video_stereo_right", "l2": 6.87457, "mean_abs": 0.149392, "relative": 0.96917 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.499665, "mean_abs": 0.103161, "relative": 0.894593 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.6593905687332153" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.5547309517860413" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9998865127563477" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.8875823020935059" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 863, "start_frame": 4315, "end_frame": 4334, "center_frame": 4324, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.534248, "mean_abs": 0.248713, "relative": 0.888609 }, { "name": "hand_right_joints", "l2": 10.25183, "mean_abs": 0.264707, "relative": 0.808118 }, { "name": "body_joints", "l2": 5.954031, "mean_abs": 0.089708, "relative": 0.804206 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020002, "mean_abs": 0.002753, "relative": 0.047201 }, { "name": "camera_rotation_matrix", "l2": 2.999779, "mean_abs": 0.20407, "relative": 0.921183 }, { "name": "imu_accel_gyro", "l2": 17.086126, "mean_abs": 1.089002, "relative": 0.921855 }, { "name": "depth_confidence", "l2": 10.939689, "mean_abs": 0.198001, "relative": 0.000363 }, { "name": "video_fisheye_cam0", "l2": 5.229683, "mean_abs": 0.108006, "relative": 0.946984 }, { "name": "video_fisheye_cam1", "l2": 5.479836, "mean_abs": 0.114048, "relative": 0.905497 }, { "name": "video_fisheye_cam2", "l2": 5.568599, "mean_abs": 0.113691, "relative": 0.888944 }, { "name": "video_fisheye_cam3", "l2": 5.538918, "mean_abs": 0.110947, "relative": 0.927425 }, { "name": "video_stereo_left", "l2": 5.355926, "mean_abs": 0.11609, "relative": 0.870349 }, { "name": "video_stereo_right", "l2": 6.886143, "mean_abs": 0.148181, "relative": 0.970802 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.525822, "mean_abs": 0.101142, "relative": 0.903954 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.474625825881958" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.5616157650947571" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9985621571540833" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.7543085217475891" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 864, "start_frame": 4320, "end_frame": 4339, "center_frame": 4329, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.522708, "mean_abs": 0.247663, "relative": 0.887635 }, { "name": "hand_right_joints", "l2": 10.24668, "mean_abs": 0.264288, "relative": 0.807712 }, { "name": "body_joints", "l2": 5.918918, "mean_abs": 0.088955, "relative": 0.799463 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008825, "mean_abs": 0.001509, "relative": 0.020825 }, { "name": "camera_rotation_matrix", "l2": 3.000494, "mean_abs": 0.204711, "relative": 0.921403 }, { "name": "imu_accel_gyro", "l2": 17.105389, "mean_abs": 1.090708, "relative": 0.922895 }, { "name": "depth_confidence", "l2": 10.938025, "mean_abs": 0.198257, "relative": 0.000363 }, { "name": "video_fisheye_cam0", "l2": 5.233828, "mean_abs": 0.108396, "relative": 0.947734 }, { "name": "video_fisheye_cam1", "l2": 5.478428, "mean_abs": 0.114655, "relative": 0.905264 }, { "name": "video_fisheye_cam2", "l2": 5.570403, "mean_abs": 0.114078, "relative": 0.889232 }, { "name": "video_fisheye_cam3", "l2": 5.524774, "mean_abs": 0.111228, "relative": 0.925057 }, { "name": "video_stereo_left", "l2": 5.362151, "mean_abs": 0.117198, "relative": 0.87136 }, { "name": "video_stereo_right", "l2": 6.879528, "mean_abs": 0.149352, "relative": 0.969869 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.323463, "mean_abs": 0.092448, "relative": 0.831533 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.5083903670310974" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.5058184862136841" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9971573352813721" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.5401791930198669" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 865, "start_frame": 4325, "end_frame": 4344, "center_frame": 4334, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.568734, "mean_abs": 0.247926, "relative": 0.891518 }, { "name": "hand_right_joints", "l2": 10.246186, "mean_abs": 0.262815, "relative": 0.807674 }, { "name": "body_joints", "l2": 5.891843, "mean_abs": 0.088491, "relative": 0.795806 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014515, "mean_abs": 0.002194, "relative": 0.034253 }, { "name": "camera_rotation_matrix", "l2": 3.00095, "mean_abs": 0.20552, "relative": 0.921543 }, { "name": "imu_accel_gyro", "l2": 17.07288, "mean_abs": 1.080073, "relative": 0.921141 }, { "name": "depth_confidence", "l2": 10.951521, "mean_abs": 0.198949, "relative": 0.000363 }, { "name": "video_fisheye_cam0", "l2": 5.235958, "mean_abs": 0.108771, "relative": 0.94812 }, { "name": "video_fisheye_cam1", "l2": 5.477108, "mean_abs": 0.115054, "relative": 0.905046 }, { "name": "video_fisheye_cam2", "l2": 5.574427, "mean_abs": 0.114327, "relative": 0.889874 }, { "name": "video_fisheye_cam3", "l2": 5.521653, "mean_abs": 0.111349, "relative": 0.924534 }, { "name": "video_stereo_left", "l2": 5.369934, "mean_abs": 0.117939, "relative": 0.872625 }, { "name": "video_stereo_right", "l2": 6.881578, "mean_abs": 0.150023, "relative": 0.970158 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.229855, "mean_abs": 0.086161, "relative": 0.798032 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.4714740514755249" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.6029254794120789" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9946126937866211" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.4586814343929291" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 866, "start_frame": 4330, "end_frame": 4349, "center_frame": 4339, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.578597, "mean_abs": 0.248319, "relative": 0.89235 }, { "name": "hand_right_joints", "l2": 10.256558, "mean_abs": 0.26294, "relative": 0.808491 }, { "name": "body_joints", "l2": 5.883697, "mean_abs": 0.088288, "relative": 0.794706 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014166, "mean_abs": 0.001723, "relative": 0.033429 }, { "name": "camera_rotation_matrix", "l2": 3.000491, "mean_abs": 0.205116, "relative": 0.921402 }, { "name": "imu_accel_gyro", "l2": 17.063974, "mean_abs": 1.076011, "relative": 0.92066 }, { "name": "depth_confidence", "l2": 10.954234, "mean_abs": 0.198305, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.23621, "mean_abs": 0.108512, "relative": 0.948166 }, { "name": "video_fisheye_cam1", "l2": 5.470817, "mean_abs": 0.114632, "relative": 0.904006 }, { "name": "video_fisheye_cam2", "l2": 5.579218, "mean_abs": 0.114169, "relative": 0.890639 }, { "name": "video_fisheye_cam3", "l2": 5.504554, "mean_abs": 0.110622, "relative": 0.921671 }, { "name": "video_stereo_left", "l2": 5.371081, "mean_abs": 0.117394, "relative": 0.872811 }, { "name": "video_stereo_right", "l2": 6.881798, "mean_abs": 0.149082, "relative": 0.970189 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.281553, "mean_abs": 0.091726, "relative": 0.816533 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.467812716960907" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.5908212065696716" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9938191771507263" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.5194101929664612" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 867, "start_frame": 4335, "end_frame": 4354, "center_frame": 4344, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.598876, "mean_abs": 0.248268, "relative": 0.89406 }, { "name": "hand_right_joints", "l2": 10.268897, "mean_abs": 0.262805, "relative": 0.809464 }, { "name": "body_joints", "l2": 5.879404, "mean_abs": 0.087798, "relative": 0.794126 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009996, "mean_abs": 0.001373, "relative": 0.023589 }, { "name": "camera_rotation_matrix", "l2": 2.999493, "mean_abs": 0.204672, "relative": 0.921096 }, { "name": "imu_accel_gyro", "l2": 17.058558, "mean_abs": 1.075949, "relative": 0.920368 }, { "name": "depth_confidence", "l2": 10.971042, "mean_abs": 0.198294, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.2333, "mean_abs": 0.10814, "relative": 0.947639 }, { "name": "video_fisheye_cam1", "l2": 5.465347, "mean_abs": 0.113873, "relative": 0.903103 }, { "name": "video_fisheye_cam2", "l2": 5.583256, "mean_abs": 0.11389, "relative": 0.891284 }, { "name": "video_fisheye_cam3", "l2": 5.484491, "mean_abs": 0.109822, "relative": 0.918312 }, { "name": "video_stereo_left", "l2": 5.371745, "mean_abs": 0.116793, "relative": 0.872919 }, { "name": "video_stereo_right", "l2": 6.878015, "mean_abs": 0.1481, "relative": 0.969656 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.418461, "mean_abs": 0.097951, "relative": 0.865531 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.530930757522583" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.4602386951446533" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9993311166763306" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.49029603600502014" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 868, "start_frame": 4340, "end_frame": 4359, "center_frame": 4349, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.614948, "mean_abs": 0.24983, "relative": 0.895416 }, { "name": "hand_right_joints", "l2": 10.265453, "mean_abs": 0.263482, "relative": 0.809192 }, { "name": "body_joints", "l2": 5.868166, "mean_abs": 0.087482, "relative": 0.792608 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007185, "mean_abs": 0.001101, "relative": 0.016955 }, { "name": "camera_rotation_matrix", "l2": 3.000524, "mean_abs": 0.2048, "relative": 0.921412 }, { "name": "imu_accel_gyro", "l2": 17.097563, "mean_abs": 1.082121, "relative": 0.922472 }, { "name": "depth_confidence", "l2": 10.965847, "mean_abs": 0.197891, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.230785, "mean_abs": 0.108031, "relative": 0.947183 }, { "name": "video_fisheye_cam1", "l2": 5.449314, "mean_abs": 0.113463, "relative": 0.900453 }, { "name": "video_fisheye_cam2", "l2": 5.583489, "mean_abs": 0.113611, "relative": 0.891321 }, { "name": "video_fisheye_cam3", "l2": 5.470623, "mean_abs": 0.109534, "relative": 0.91599 }, { "name": "video_stereo_left", "l2": 5.37612, "mean_abs": 0.116631, "relative": 0.87363 }, { "name": "video_stereo_right", "l2": 6.87348, "mean_abs": 0.148024, "relative": 0.969016 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.271025, "mean_abs": 0.089935, "relative": 0.812766 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.009986, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.4573417901992798" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.5442340970039368" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9987164735794067" }, "next_action": { "true": "Close bottle cap", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.4316399097442627" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 869, "start_frame": 4345, "end_frame": 4364, "center_frame": 4354, "action": "Close bottle cap", "subtask": "Pour and close white bottle", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.620158, "mean_abs": 0.250531, "relative": 0.895856 }, { "name": "hand_right_joints", "l2": 10.265939, "mean_abs": 0.263666, "relative": 0.809231 }, { "name": "body_joints", "l2": 5.876598, "mean_abs": 0.087861, "relative": 0.793747 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007327, "mean_abs": 0.001107, "relative": 0.01729 }, { "name": "camera_rotation_matrix", "l2": 2.999574, "mean_abs": 0.20475, "relative": 0.92112 }, { "name": "imu_accel_gyro", "l2": 17.076124, "mean_abs": 1.077759, "relative": 0.921316 }, { "name": "depth_confidence", "l2": 10.953418, "mean_abs": 0.197966, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.228138, "mean_abs": 0.108037, "relative": 0.946704 }, { "name": "video_fisheye_cam1", "l2": 5.440907, "mean_abs": 0.113426, "relative": 0.899064 }, { "name": "video_fisheye_cam2", "l2": 5.578168, "mean_abs": 0.113537, "relative": 0.890472 }, { "name": "video_fisheye_cam3", "l2": 5.463185, "mean_abs": 0.1096, "relative": 0.914744 }, { "name": "video_stereo_left", "l2": 5.37165, "mean_abs": 0.116474, "relative": 0.872904 }, { "name": "video_stereo_right", "l2": 6.865054, "mean_abs": 0.147999, "relative": 0.967829 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.233559, "mean_abs": 0.08852, "relative": 0.799357 }, { "name": "caption_objects_interaction_text", "l2": 1.847926, "mean_abs": 0.013466, "relative": 0.848523 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.35056272149086" }, "timeline_subtask": { "true": "Pour and close white bottle", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.5411785244941711" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9643778800964355" }, "next_action": { "true": "Close bottle cap", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.4901554584503174" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|table|coffee filter|white mug", "correct": 0, "confidence": "" } } }, { "window_index": 870, "start_frame": 4350, "end_frame": 4369, "center_frame": 4359, "action": "Close bottle cap", "subtask": "", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.619654, "mean_abs": 0.250991, "relative": 0.895813 }, { "name": "hand_right_joints", "l2": 10.270012, "mean_abs": 0.264836, "relative": 0.809552 }, { "name": "body_joints", "l2": 5.896848, "mean_abs": 0.088518, "relative": 0.796483 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006392, "mean_abs": 0.001059, "relative": 0.015084 }, { "name": "camera_rotation_matrix", "l2": 2.999757, "mean_abs": 0.204746, "relative": 0.921177 }, { "name": "imu_accel_gyro", "l2": 17.075415, "mean_abs": 1.075408, "relative": 0.921277 }, { "name": "depth_confidence", "l2": 10.956563, "mean_abs": 0.197462, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.224517, "mean_abs": 0.107881, "relative": 0.946048 }, { "name": "video_fisheye_cam1", "l2": 5.430801, "mean_abs": 0.113145, "relative": 0.897394 }, { "name": "video_fisheye_cam2", "l2": 5.574403, "mean_abs": 0.113288, "relative": 0.889871 }, { "name": "video_fisheye_cam3", "l2": 5.453154, "mean_abs": 0.109171, "relative": 0.913065 }, { "name": "video_stereo_left", "l2": 5.364296, "mean_abs": 0.115803, "relative": 0.871709 }, { "name": "video_stereo_right", "l2": 6.850258, "mean_abs": 0.147172, "relative": 0.965743 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.256107, "mean_abs": 0.091355, "relative": 0.807427 }, { "name": "caption_objects_interaction_text", "l2": 1.847926, "mean_abs": 0.013606, "relative": 0.848523 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.35964709520339966" }, "timeline_subtask": null, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9947720170021057" }, "next_action": { "true": "Close bottle cap", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.5876492261886597" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|table|coffee filter|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 871, "start_frame": 4355, "end_frame": 4374, "center_frame": 4364, "action": "Close bottle cap", "subtask": "Move bottle to coffee equipment", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.61601, "mean_abs": 0.251512, "relative": 0.895506 }, { "name": "hand_right_joints", "l2": 10.278042, "mean_abs": 0.265677, "relative": 0.810185 }, { "name": "body_joints", "l2": 5.914977, "mean_abs": 0.08873, "relative": 0.798931 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003356, "mean_abs": 0.000549, "relative": 0.007919 }, { "name": "camera_rotation_matrix", "l2": 2.999959, "mean_abs": 0.204159, "relative": 0.921239 }, { "name": "imu_accel_gyro", "l2": 17.082094, "mean_abs": 1.076161, "relative": 0.921638 }, { "name": "depth_confidence", "l2": 10.957184, "mean_abs": 0.197259, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.22338, "mean_abs": 0.107797, "relative": 0.945842 }, { "name": "video_fisheye_cam1", "l2": 5.420958, "mean_abs": 0.112748, "relative": 0.895768 }, { "name": "video_fisheye_cam2", "l2": 5.572766, "mean_abs": 0.113166, "relative": 0.889609 }, { "name": "video_fisheye_cam3", "l2": 5.442265, "mean_abs": 0.10907, "relative": 0.911242 }, { "name": "video_stereo_left", "l2": 5.359917, "mean_abs": 0.115363, "relative": 0.870997 }, { "name": "video_stereo_right", "l2": 6.839584, "mean_abs": 0.146523, "relative": 0.964238 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.249936, "mean_abs": 0.090957, "relative": 0.805218 }, { "name": "caption_objects_interaction_text", "l2": 1.847926, "mean_abs": 0.013476, "relative": 0.848523 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.4491121768951416" }, "timeline_subtask": { "true": "Move bottle to coffee equipment", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.5403767228126526" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9786983728408813" }, "next_action": { "true": "Close bottle cap", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.7296859622001648" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "kettle|coffee dripper|scale|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 872, "start_frame": 4360, "end_frame": 4379, "center_frame": 4369, "action": "Close bottle cap", "subtask": "Move bottle to coffee equipment", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.616114, "mean_abs": 0.251894, "relative": 0.895514 }, { "name": "hand_right_joints", "l2": 10.251781, "mean_abs": 0.265055, "relative": 0.808115 }, { "name": "body_joints", "l2": 5.927996, "mean_abs": 0.08915, "relative": 0.80069 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.002594, "mean_abs": 0.000453, "relative": 0.006122 }, { "name": "camera_rotation_matrix", "l2": 3.00021, "mean_abs": 0.203915, "relative": 0.921316 }, { "name": "imu_accel_gyro", "l2": 17.071516, "mean_abs": 1.077376, "relative": 0.921067 }, { "name": "depth_confidence", "l2": 10.9572, "mean_abs": 0.196872, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.219472, "mean_abs": 0.107656, "relative": 0.945135 }, { "name": "video_fisheye_cam1", "l2": 5.409733, "mean_abs": 0.112357, "relative": 0.893913 }, { "name": "video_fisheye_cam2", "l2": 5.568027, "mean_abs": 0.11293, "relative": 0.888853 }, { "name": "video_fisheye_cam3", "l2": 5.428602, "mean_abs": 0.108864, "relative": 0.908954 }, { "name": "video_stereo_left", "l2": 5.353738, "mean_abs": 0.115043, "relative": 0.869993 }, { "name": "video_stereo_right", "l2": 6.825415, "mean_abs": 0.146137, "relative": 0.96224 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.180142, "mean_abs": 0.088012, "relative": 0.78024 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.5572973489761353" }, "timeline_subtask": { "true": "Move bottle to coffee equipment", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.5159992575645447" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9990345239639282" }, "next_action": { "true": "Close bottle cap", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.4257010221481323" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|scale|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 873, "start_frame": 4365, "end_frame": 4384, "center_frame": 4374, "action": "Close bottle cap", "subtask": "Move bottle to coffee equipment", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.612768, "mean_abs": 0.252307, "relative": 0.895232 }, { "name": "hand_right_joints", "l2": 10.257115, "mean_abs": 0.265331, "relative": 0.808535 }, { "name": "body_joints", "l2": 5.933349, "mean_abs": 0.089249, "relative": 0.801413 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003126, "mean_abs": 0.000528, "relative": 0.007376 }, { "name": "camera_rotation_matrix", "l2": 3.000099, "mean_abs": 0.203727, "relative": 0.921282 }, { "name": "imu_accel_gyro", "l2": 17.05859, "mean_abs": 1.078432, "relative": 0.92037 }, { "name": "depth_confidence", "l2": 10.971573, "mean_abs": 0.197452, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.212559, "mean_abs": 0.107488, "relative": 0.943883 }, { "name": "video_fisheye_cam1", "l2": 5.396057, "mean_abs": 0.112085, "relative": 0.891653 }, { "name": "video_fisheye_cam2", "l2": 5.564337, "mean_abs": 0.112796, "relative": 0.888264 }, { "name": "video_fisheye_cam3", "l2": 5.39729, "mean_abs": 0.108095, "relative": 0.903711 }, { "name": "video_stereo_left", "l2": 5.346336, "mean_abs": 0.114787, "relative": 0.86879 }, { "name": "video_stereo_right", "l2": 6.799174, "mean_abs": 0.145668, "relative": 0.958541 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.16996, "mean_abs": 0.086848, "relative": 0.776596 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Move kettle away", "correct": 0, "confidence": "0.6749051809310913" }, "timeline_subtask": { "true": "Move bottle to coffee equipment", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.55208820104599" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9937294721603394" }, "next_action": { "true": "Place item on table", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.49570807814598083" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|scale|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 874, "start_frame": 4370, "end_frame": 4389, "center_frame": 4379, "action": "Close bottle cap", "subtask": "Move bottle to coffee equipment", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.615052, "mean_abs": 0.25179, "relative": 0.895425 }, { "name": "hand_right_joints", "l2": 10.262856, "mean_abs": 0.265771, "relative": 0.808988 }, { "name": "body_joints", "l2": 5.931659, "mean_abs": 0.08962, "relative": 0.801184 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004412, "mean_abs": 0.000682, "relative": 0.010411 }, { "name": "camera_rotation_matrix", "l2": 3.000287, "mean_abs": 0.203957, "relative": 0.921339 }, { "name": "imu_accel_gyro", "l2": 17.062056, "mean_abs": 1.07746, "relative": 0.920557 }, { "name": "depth_confidence", "l2": 10.995009, "mean_abs": 0.197428, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.215461, "mean_abs": 0.107666, "relative": 0.944408 }, { "name": "video_fisheye_cam1", "l2": 5.40475, "mean_abs": 0.112442, "relative": 0.893089 }, { "name": "video_fisheye_cam2", "l2": 5.567194, "mean_abs": 0.113055, "relative": 0.88872 }, { "name": "video_fisheye_cam3", "l2": 5.44314, "mean_abs": 0.109604, "relative": 0.911388 }, { "name": "video_stereo_left", "l2": 5.346537, "mean_abs": 0.114908, "relative": 0.868823 }, { "name": "video_stereo_right", "l2": 6.810825, "mean_abs": 0.146083, "relative": 0.960183 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.114868, "mean_abs": 0.082516, "relative": 0.756879 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.6722816228866577" }, "timeline_subtask": { "true": "Move bottle to coffee equipment", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.7330989241600037" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9963638186454773" }, "next_action": { "true": "Place item on table", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.35026469826698303" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|scale|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 875, "start_frame": 4375, "end_frame": 4394, "center_frame": 4384, "action": "Close bottle cap", "subtask": "Move bottle to coffee equipment", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.621385, "mean_abs": 0.251866, "relative": 0.895959 }, { "name": "hand_right_joints", "l2": 10.258162, "mean_abs": 0.265231, "relative": 0.808618 }, { "name": "body_joints", "l2": 5.921546, "mean_abs": 0.089212, "relative": 0.799818 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00349, "mean_abs": 0.000593, "relative": 0.008237 }, { "name": "camera_rotation_matrix", "l2": 3.00033, "mean_abs": 0.203885, "relative": 0.921353 }, { "name": "imu_accel_gyro", "l2": 17.06237, "mean_abs": 1.076879, "relative": 0.920574 }, { "name": "depth_confidence", "l2": 10.998648, "mean_abs": 0.197823, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.214386, "mean_abs": 0.10757, "relative": 0.944214 }, { "name": "video_fisheye_cam1", "l2": 5.399018, "mean_abs": 0.112314, "relative": 0.892142 }, { "name": "video_fisheye_cam2", "l2": 5.567091, "mean_abs": 0.112948, "relative": 0.888703 }, { "name": "video_fisheye_cam3", "l2": 5.447854, "mean_abs": 0.10973, "relative": 0.912177 }, { "name": "video_stereo_left", "l2": 5.343844, "mean_abs": 0.114709, "relative": 0.868385 }, { "name": "video_stereo_right", "l2": 6.804449, "mean_abs": 0.145847, "relative": 0.959285 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.074206, "mean_abs": 0.081809, "relative": 0.742327 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.6498131155967712" }, "timeline_subtask": { "true": "Move bottle to coffee equipment", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.7760382294654846" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.998038113117218" }, "next_action": { "true": "Place item on table", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.47853797674179077" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|scale|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 876, "start_frame": 4380, "end_frame": 4399, "center_frame": 4389, "action": "Close bottle cap", "subtask": "Move bottle to coffee equipment", "objects": [ "white bottle", "coffee dripper", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.619637, "mean_abs": 0.251465, "relative": 0.895812 }, { "name": "hand_right_joints", "l2": 10.270851, "mean_abs": 0.265126, "relative": 0.809618 }, { "name": "body_joints", "l2": 5.913624, "mean_abs": 0.089089, "relative": 0.798748 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004262, "mean_abs": 0.0007, "relative": 0.010057 }, { "name": "camera_rotation_matrix", "l2": 3.000127, "mean_abs": 0.203755, "relative": 0.92129 }, { "name": "imu_accel_gyro", "l2": 17.064264, "mean_abs": 1.079999, "relative": 0.920676 }, { "name": "depth_confidence", "l2": 11.000221, "mean_abs": 0.197788, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.217219, "mean_abs": 0.107644, "relative": 0.944727 }, { "name": "video_fisheye_cam1", "l2": 5.401508, "mean_abs": 0.112521, "relative": 0.892554 }, { "name": "video_fisheye_cam2", "l2": 5.569332, "mean_abs": 0.11313, "relative": 0.889061 }, { "name": "video_fisheye_cam3", "l2": 5.483031, "mean_abs": 0.110594, "relative": 0.918067 }, { "name": "video_stereo_left", "l2": 5.345811, "mean_abs": 0.114894, "relative": 0.868705 }, { "name": "video_stereo_right", "l2": 6.809369, "mean_abs": 0.146096, "relative": 0.959978 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.120108, "mean_abs": 0.08465, "relative": 0.758755 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010045, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.5780695676803589" }, "timeline_subtask": { "true": "Move bottle to coffee equipment", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.8253584504127502" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9995396137237549" }, "next_action": { "true": "Place item on table", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.6192188262939453" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white bottle", "predicted": "coffee dripper|scale|gooseneck kettle|table|coffee filter|white mug", "correct": 0, "confidence": "" } } }, { "window_index": 877, "start_frame": 4385, "end_frame": 4404, "center_frame": 4394, "action": "Close bottle cap", "subtask": "Move bottle to coffee equipment", "objects": [ "white bottle", "coffee dripper", "scale", "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.618572, "mean_abs": 0.251082, "relative": 0.895722 }, { "name": "hand_right_joints", "l2": 10.288548, "mean_abs": 0.265653, "relative": 0.811013 }, { "name": "body_joints", "l2": 5.910118, "mean_abs": 0.088811, "relative": 0.798275 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003883, "mean_abs": 0.000639, "relative": 0.009163 }, { "name": "camera_rotation_matrix", "l2": 2.99991, "mean_abs": 0.203642, "relative": 0.921224 }, { "name": "imu_accel_gyro", "l2": 17.082249, "mean_abs": 1.079628, "relative": 0.921646 }, { "name": "depth_confidence", "l2": 10.993821, "mean_abs": 0.197303, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.217947, "mean_abs": 0.107675, "relative": 0.944859 }, { "name": "video_fisheye_cam1", "l2": 5.40274, "mean_abs": 0.112595, "relative": 0.892757 }, { "name": "video_fisheye_cam2", "l2": 5.571508, "mean_abs": 0.113148, "relative": 0.889408 }, { "name": "video_fisheye_cam3", "l2": 5.492898, "mean_abs": 0.110822, "relative": 0.919719 }, { "name": "video_stereo_left", "l2": 5.344201, "mean_abs": 0.114765, "relative": 0.868443 }, { "name": "video_stereo_right", "l2": 6.807387, "mean_abs": 0.146121, "relative": 0.959699 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.052278, "mean_abs": 0.079229, "relative": 0.73448 }, { "name": "caption_objects_interaction_text", "l2": 1.946363, "mean_abs": 0.016766, "relative": 0.893723 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Close bottle cap", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.5807220339775085" }, "timeline_subtask": { "true": "Move bottle to coffee equipment", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.7143031358718872" }, "transition_detection": { "true": "transition", "predicted": "steady", "correct": 0, "confidence": "0.54218989610672" }, "next_action": { "true": "Place item on table", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.6155366897583008" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white mug|white bottle|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|bottle|gooseneck kettle|table|coffee filter|white mug", "correct": 0, "confidence": "" } } }, { "window_index": 878, "start_frame": 4390, "end_frame": 4409, "center_frame": 4399, "action": "", "subtask": "", "objects": [ "white bottle", "coffee dripper", "scale", "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.622157, "mean_abs": 0.251754, "relative": 0.896024 }, { "name": "hand_right_joints", "l2": 10.280558, "mean_abs": 0.265361, "relative": 0.810383 }, { "name": "body_joints", "l2": 5.898264, "mean_abs": 0.088849, "relative": 0.796674 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005858, "mean_abs": 0.000867, "relative": 0.013823 }, { "name": "camera_rotation_matrix", "l2": 2.999787, "mean_abs": 0.203849, "relative": 0.921186 }, { "name": "imu_accel_gyro", "l2": 17.083704, "mean_abs": 1.083022, "relative": 0.921725 }, { "name": "depth_confidence", "l2": 10.990012, "mean_abs": 0.197412, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.218227, "mean_abs": 0.107758, "relative": 0.944909 }, { "name": "video_fisheye_cam1", "l2": 5.392715, "mean_abs": 0.112337, "relative": 0.891101 }, { "name": "video_fisheye_cam2", "l2": 5.570493, "mean_abs": 0.113186, "relative": 0.889246 }, { "name": "video_fisheye_cam3", "l2": 5.492024, "mean_abs": 0.110869, "relative": 0.919573 }, { "name": "video_stereo_left", "l2": 5.340645, "mean_abs": 0.115006, "relative": 0.867865 }, { "name": "video_stereo_right", "l2": 6.78183, "mean_abs": 0.145887, "relative": 0.956096 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.054257, "mean_abs": 0.079907, "relative": 0.735188 }, { "name": "caption_objects_interaction_text", "l2": 1.946363, "mean_abs": 0.017152, "relative": 0.893723 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": { "true": "transition", "predicted": "transition", "correct": 1, "confidence": "0.9148183465003967" }, "next_action": { "true": "Place item on table", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.5739760398864746" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white mug|white bottle|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|table|coffee filter|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 879, "start_frame": 4395, "end_frame": 4414, "center_frame": 4404, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "white bottle", "coffee dripper", "scale", "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.637351, "mean_abs": 0.252492, "relative": 0.897306 }, { "name": "hand_right_joints", "l2": 10.24526, "mean_abs": 0.265865, "relative": 0.807601 }, { "name": "body_joints", "l2": 5.886996, "mean_abs": 0.089301, "relative": 0.795152 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014978, "mean_abs": 0.001797, "relative": 0.035346 }, { "name": "camera_rotation_matrix", "l2": 3.000767, "mean_abs": 0.206447, "relative": 0.921487 }, { "name": "imu_accel_gyro", "l2": 17.094643, "mean_abs": 1.092314, "relative": 0.922315 }, { "name": "depth_confidence", "l2": 10.975515, "mean_abs": 0.198354, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.217411, "mean_abs": 0.108363, "relative": 0.944761 }, { "name": "video_fisheye_cam1", "l2": 5.391446, "mean_abs": 0.113465, "relative": 0.890891 }, { "name": "video_fisheye_cam2", "l2": 5.578839, "mean_abs": 0.113925, "relative": 0.890579 }, { "name": "video_fisheye_cam3", "l2": 5.486614, "mean_abs": 0.111752, "relative": 0.918667 }, { "name": "video_stereo_left", "l2": 5.350115, "mean_abs": 0.116748, "relative": 0.869404 }, { "name": "video_stereo_right", "l2": 6.767528, "mean_abs": 0.147756, "relative": 0.954079 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.013649, "mean_abs": 0.076679, "relative": 0.720655 }, { "name": "caption_objects_interaction_text", "l2": 1.946363, "mean_abs": 0.01707, "relative": 0.893723 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.34716734290122986" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.4783322811126709" }, "transition_detection": { "true": "transition", "predicted": "transition", "correct": 1, "confidence": "0.9438948631286621" }, "next_action": { "true": "Place item on table", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.44366154074668884" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "coffee dripper|scale|white mug|white bottle|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|white mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 880, "start_frame": 4400, "end_frame": 4419, "center_frame": 4409, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.655862, "mean_abs": 0.251791, "relative": 0.898867 }, { "name": "hand_right_joints", "l2": 10.210726, "mean_abs": 0.268298, "relative": 0.804878 }, { "name": "body_joints", "l2": 5.884796, "mean_abs": 0.089391, "relative": 0.794855 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.02412, "mean_abs": 0.002964, "relative": 0.05692 }, { "name": "camera_rotation_matrix", "l2": 3.001867, "mean_abs": 0.209495, "relative": 0.921825 }, { "name": "imu_accel_gyro", "l2": 17.099409, "mean_abs": 1.086311, "relative": 0.922572 }, { "name": "depth_confidence", "l2": 10.985048, "mean_abs": 0.200137, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.214704, "mean_abs": 0.108711, "relative": 0.944271 }, { "name": "video_fisheye_cam1", "l2": 5.365742, "mean_abs": 0.114318, "relative": 0.886644 }, { "name": "video_fisheye_cam2", "l2": 5.585379, "mean_abs": 0.11472, "relative": 0.891623 }, { "name": "video_fisheye_cam3", "l2": 5.432402, "mean_abs": 0.110665, "relative": 0.90959 }, { "name": "video_stereo_left", "l2": 5.359657, "mean_abs": 0.118411, "relative": 0.870955 }, { "name": "video_stereo_right", "l2": 6.697318, "mean_abs": 0.148759, "relative": 0.944181 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.003654, "mean_abs": 0.076352, "relative": 0.717077 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.48304322361946106" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.5013586282730103" }, "transition_detection": { "true": "transition", "predicted": "steady", "correct": 0, "confidence": "0.9959709048271179" }, "next_action": { "true": "Place item on table", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.3325023353099823" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 881, "start_frame": 4405, "end_frame": 4424, "center_frame": 4414, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.660891, "mean_abs": 0.251462, "relative": 0.899292 }, { "name": "hand_right_joints", "l2": 10.185116, "mean_abs": 0.267809, "relative": 0.80286 }, { "name": "body_joints", "l2": 5.886851, "mean_abs": 0.089206, "relative": 0.795132 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.029652, "mean_abs": 0.003868, "relative": 0.069974 }, { "name": "camera_rotation_matrix", "l2": 3.002845, "mean_abs": 0.211378, "relative": 0.922125 }, { "name": "imu_accel_gyro", "l2": 17.070229, "mean_abs": 1.083381, "relative": 0.920998 }, { "name": "depth_confidence", "l2": 11.00514, "mean_abs": 0.201137, "relative": 0.000365 }, { "name": "video_fisheye_cam0", "l2": 5.217042, "mean_abs": 0.109086, "relative": 0.944695 }, { "name": "video_fisheye_cam1", "l2": 5.352989, "mean_abs": 0.11475, "relative": 0.884536 }, { "name": "video_fisheye_cam2", "l2": 5.59083, "mean_abs": 0.1153, "relative": 0.892493 }, { "name": "video_fisheye_cam3", "l2": 5.426476, "mean_abs": 0.110714, "relative": 0.908598 }, { "name": "video_stereo_left", "l2": 5.365218, "mean_abs": 0.119338, "relative": 0.871859 }, { "name": "video_stereo_right", "l2": 6.651021, "mean_abs": 0.149191, "relative": 0.937654 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.991961, "mean_abs": 0.076509, "relative": 0.712893 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.3992617726325989" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9182792901992798" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9968043565750122" }, "next_action": { "true": "Place item on table", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.36838576197624207" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 882, "start_frame": 4410, "end_frame": 4429, "center_frame": 4419, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.648468, "mean_abs": 0.250709, "relative": 0.898244 }, { "name": "hand_right_joints", "l2": 10.152904, "mean_abs": 0.266335, "relative": 0.80032 }, { "name": "body_joints", "l2": 5.900481, "mean_abs": 0.089557, "relative": 0.796973 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030006, "mean_abs": 0.004134, "relative": 0.07081 }, { "name": "camera_rotation_matrix", "l2": 3.002777, "mean_abs": 0.212215, "relative": 0.922104 }, { "name": "imu_accel_gyro", "l2": 17.070118, "mean_abs": 1.088797, "relative": 0.920992 }, { "name": "depth_confidence", "l2": 11.034939, "mean_abs": 0.201739, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.216049, "mean_abs": 0.108883, "relative": 0.944515 }, { "name": "video_fisheye_cam1", "l2": 5.345075, "mean_abs": 0.114775, "relative": 0.883228 }, { "name": "video_fisheye_cam2", "l2": 5.594165, "mean_abs": 0.115285, "relative": 0.893025 }, { "name": "video_fisheye_cam3", "l2": 5.425295, "mean_abs": 0.110381, "relative": 0.9084 }, { "name": "video_stereo_left", "l2": 5.364243, "mean_abs": 0.119106, "relative": 0.8717 }, { "name": "video_stereo_right", "l2": 6.610762, "mean_abs": 0.148682, "relative": 0.931979 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.042438, "mean_abs": 0.07866, "relative": 0.730958 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.5430309772491455" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.7371112704277039" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9955625534057617" }, "next_action": { "true": "Place item on table", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.24680756032466888" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 883, "start_frame": 4415, "end_frame": 4434, "center_frame": 4424, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.651172, "mean_abs": 0.250697, "relative": 0.898472 }, { "name": "hand_right_joints", "l2": 10.115466, "mean_abs": 0.262854, "relative": 0.797369 }, { "name": "body_joints", "l2": 5.919017, "mean_abs": 0.089118, "relative": 0.799477 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021237, "mean_abs": 0.003034, "relative": 0.050116 }, { "name": "camera_rotation_matrix", "l2": 3.00137, "mean_abs": 0.21144, "relative": 0.921672 }, { "name": "imu_accel_gyro", "l2": 17.083704, "mean_abs": 1.082695, "relative": 0.921725 }, { "name": "depth_confidence", "l2": 11.061275, "mean_abs": 0.201364, "relative": 0.000367 }, { "name": "video_fisheye_cam0", "l2": 5.211133, "mean_abs": 0.108416, "relative": 0.943625 }, { "name": "video_fisheye_cam1", "l2": 5.311304, "mean_abs": 0.113072, "relative": 0.877648 }, { "name": "video_fisheye_cam2", "l2": 5.599463, "mean_abs": 0.11459, "relative": 0.893871 }, { "name": "video_fisheye_cam3", "l2": 5.408424, "mean_abs": 0.109584, "relative": 0.905575 }, { "name": "video_stereo_left", "l2": 5.355408, "mean_abs": 0.117306, "relative": 0.870264 }, { "name": "video_stereo_right", "l2": 6.514615, "mean_abs": 0.144597, "relative": 0.918424 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.014642, "mean_abs": 0.07715, "relative": 0.72101 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.6120333671569824" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.6091059446334839" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9988614320755005" }, "next_action": { "true": "Place item on table", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.3748154044151306" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 884, "start_frame": 4420, "end_frame": 4439, "center_frame": 4429, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.644446, "mean_abs": 0.249535, "relative": 0.897904 }, { "name": "hand_right_joints", "l2": 10.138288, "mean_abs": 0.262823, "relative": 0.799168 }, { "name": "body_joints", "l2": 5.941038, "mean_abs": 0.088734, "relative": 0.802451 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015898, "mean_abs": 0.002241, "relative": 0.037517 }, { "name": "camera_rotation_matrix", "l2": 3.000431, "mean_abs": 0.21075, "relative": 0.921384 }, { "name": "imu_accel_gyro", "l2": 17.047279, "mean_abs": 1.071865, "relative": 0.919759 }, { "name": "depth_confidence", "l2": 11.107701, "mean_abs": 0.201325, "relative": 0.000369 }, { "name": "video_fisheye_cam0", "l2": 5.209931, "mean_abs": 0.10816, "relative": 0.943407 }, { "name": "video_fisheye_cam1", "l2": 5.29885, "mean_abs": 0.111793, "relative": 0.87559 }, { "name": "video_fisheye_cam2", "l2": 5.604962, "mean_abs": 0.113986, "relative": 0.894749 }, { "name": "video_fisheye_cam3", "l2": 5.444949, "mean_abs": 0.11033, "relative": 0.911691 }, { "name": "video_stereo_left", "l2": 5.346252, "mean_abs": 0.115692, "relative": 0.868777 }, { "name": "video_stereo_right", "l2": 6.438422, "mean_abs": 0.141375, "relative": 0.907682 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.987886, "mean_abs": 0.074585, "relative": 0.711434 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.6792599558830261" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.5893995761871338" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9925568103790283" }, "next_action": { "true": "Place item on table", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.5984087586402893" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 885, "start_frame": 4425, "end_frame": 4444, "center_frame": 4434, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.661516, "mean_abs": 0.249547, "relative": 0.899344 }, { "name": "hand_right_joints", "l2": 10.227201, "mean_abs": 0.267512, "relative": 0.806177 }, { "name": "body_joints", "l2": 5.955504, "mean_abs": 0.088893, "relative": 0.804405 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015385, "mean_abs": 0.001869, "relative": 0.036307 }, { "name": "camera_rotation_matrix", "l2": 3.000457, "mean_abs": 0.210713, "relative": 0.921392 }, { "name": "imu_accel_gyro", "l2": 17.082436, "mean_abs": 1.057064, "relative": 0.921656 }, { "name": "depth_confidence", "l2": 11.136023, "mean_abs": 0.201446, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.210428, "mean_abs": 0.1081, "relative": 0.943497 }, { "name": "video_fisheye_cam1", "l2": 5.283793, "mean_abs": 0.111392, "relative": 0.873102 }, { "name": "video_fisheye_cam2", "l2": 5.608104, "mean_abs": 0.113933, "relative": 0.89525 }, { "name": "video_fisheye_cam3", "l2": 5.463421, "mean_abs": 0.110685, "relative": 0.914784 }, { "name": "video_stereo_left", "l2": 5.334822, "mean_abs": 0.114922, "relative": 0.866919 }, { "name": "video_stereo_right", "l2": 6.371315, "mean_abs": 0.139716, "relative": 0.898222 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.076742, "mean_abs": 0.082106, "relative": 0.743235 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.6812241077423096" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.617874264717102" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9747772216796875" }, "next_action": { "true": "Place item on table", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.5721173882484436" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 886, "start_frame": 4430, "end_frame": 4449, "center_frame": 4439, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.676023, "mean_abs": 0.249949, "relative": 0.900568 }, { "name": "hand_right_joints", "l2": 10.347258, "mean_abs": 0.273993, "relative": 0.815641 }, { "name": "body_joints", "l2": 5.935285, "mean_abs": 0.088699, "relative": 0.801674 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.022125, "mean_abs": 0.002606, "relative": 0.052212 }, { "name": "camera_rotation_matrix", "l2": 3.000983, "mean_abs": 0.211901, "relative": 0.921553 }, { "name": "imu_accel_gyro", "l2": 17.051601, "mean_abs": 1.048857, "relative": 0.919993 }, { "name": "depth_confidence", "l2": 11.15503, "mean_abs": 0.201997, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.213947, "mean_abs": 0.108418, "relative": 0.944134 }, { "name": "video_fisheye_cam1", "l2": 5.273895, "mean_abs": 0.111741, "relative": 0.871467 }, { "name": "video_fisheye_cam2", "l2": 5.613841, "mean_abs": 0.114449, "relative": 0.896166 }, { "name": "video_fisheye_cam3", "l2": 5.491686, "mean_abs": 0.111298, "relative": 0.919517 }, { "name": "video_stereo_left", "l2": 5.331507, "mean_abs": 0.115769, "relative": 0.866381 }, { "name": "video_stereo_right", "l2": 6.336414, "mean_abs": 0.140504, "relative": 0.893301 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.117892, "mean_abs": 0.085762, "relative": 0.757962 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.6840025782585144" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.6078978776931763" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.6901268362998962" }, "next_action": { "true": "Place item on table", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.3972393274307251" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 887, "start_frame": 4435, "end_frame": 4454, "center_frame": 4444, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.691214, "mean_abs": 0.249947, "relative": 0.90185 }, { "name": "hand_right_joints", "l2": 10.47658, "mean_abs": 0.278487, "relative": 0.825835 }, { "name": "body_joints", "l2": 5.902078, "mean_abs": 0.088563, "relative": 0.797189 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028601, "mean_abs": 0.003164, "relative": 0.067495 }, { "name": "camera_rotation_matrix", "l2": 3.00188, "mean_abs": 0.214463, "relative": 0.921829 }, { "name": "imu_accel_gyro", "l2": 17.081989, "mean_abs": 1.054796, "relative": 0.921632 }, { "name": "depth_confidence", "l2": 11.171109, "mean_abs": 0.202782, "relative": 0.000371 }, { "name": "video_fisheye_cam0", "l2": 5.215105, "mean_abs": 0.108919, "relative": 0.944344 }, { "name": "video_fisheye_cam1", "l2": 5.261846, "mean_abs": 0.112435, "relative": 0.869476 }, { "name": "video_fisheye_cam2", "l2": 5.620086, "mean_abs": 0.115297, "relative": 0.897163 }, { "name": "video_fisheye_cam3", "l2": 5.516734, "mean_abs": 0.11217, "relative": 0.923711 }, { "name": "video_stereo_left", "l2": 5.329151, "mean_abs": 0.117234, "relative": 0.865998 }, { "name": "video_stereo_right", "l2": 6.317321, "mean_abs": 0.141963, "relative": 0.89061 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.190976, "mean_abs": 0.093682, "relative": 0.784118 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.5029498338699341" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9316630363464355" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.6695892214775085" }, "next_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.5672380924224854" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 888, "start_frame": 4440, "end_frame": 4459, "center_frame": 4449, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.717397, "mean_abs": 0.250803, "relative": 0.904058 }, { "name": "hand_right_joints", "l2": 10.612707, "mean_abs": 0.278387, "relative": 0.836565 }, { "name": "body_joints", "l2": 5.849841, "mean_abs": 0.087443, "relative": 0.790133 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.041322, "mean_abs": 0.005066, "relative": 0.097513 }, { "name": "camera_rotation_matrix", "l2": 3.005217, "mean_abs": 0.218892, "relative": 0.922853 }, { "name": "imu_accel_gyro", "l2": 17.081608, "mean_abs": 1.063392, "relative": 0.921612 }, { "name": "depth_confidence", "l2": 11.208265, "mean_abs": 0.205105, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.225113, "mean_abs": 0.109934, "relative": 0.946156 }, { "name": "video_fisheye_cam1", "l2": 5.25986, "mean_abs": 0.113883, "relative": 0.869147 }, { "name": "video_fisheye_cam2", "l2": 5.648984, "mean_abs": 0.11702, "relative": 0.901776 }, { "name": "video_fisheye_cam3", "l2": 5.550614, "mean_abs": 0.114014, "relative": 0.929383 }, { "name": "video_stereo_left", "l2": 5.347818, "mean_abs": 0.119685, "relative": 0.869031 }, { "name": "video_stereo_right", "l2": 6.317947, "mean_abs": 0.143911, "relative": 0.890698 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.151362, "mean_abs": 0.088544, "relative": 0.76994 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.40355581045150757" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9392425417900085" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.6822054386138916" }, "next_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.7816019654273987" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|table|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 889, "start_frame": 4445, "end_frame": 4464, "center_frame": 4454, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.698029, "mean_abs": 0.251013, "relative": 0.902424 }, { "name": "hand_right_joints", "l2": 10.743038, "mean_abs": 0.275107, "relative": 0.846839 }, { "name": "body_joints", "l2": 5.799204, "mean_abs": 0.085982, "relative": 0.783294 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.069382, "mean_abs": 0.008616, "relative": 0.163733 }, { "name": "camera_rotation_matrix", "l2": 3.011069, "mean_abs": 0.223474, "relative": 0.92465 }, { "name": "imu_accel_gyro", "l2": 17.061077, "mean_abs": 1.069797, "relative": 0.920504 }, { "name": "depth_confidence", "l2": 11.228032, "mean_abs": 0.207435, "relative": 0.000373 }, { "name": "video_fisheye_cam0", "l2": 5.238515, "mean_abs": 0.110822, "relative": 0.948583 }, { "name": "video_fisheye_cam1", "l2": 5.256261, "mean_abs": 0.115576, "relative": 0.868553 }, { "name": "video_fisheye_cam2", "l2": 5.69045, "mean_abs": 0.118585, "relative": 0.908396 }, { "name": "video_fisheye_cam3", "l2": 5.583374, "mean_abs": 0.115954, "relative": 0.934869 }, { "name": "video_stereo_left", "l2": 5.421436, "mean_abs": 0.123089, "relative": 0.880994 }, { "name": "video_stereo_right", "l2": 6.337507, "mean_abs": 0.145622, "relative": 0.893456 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.180365, "mean_abs": 0.08945, "relative": 0.78032 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.8008579015731812" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9597190618515015" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.831466794013977" }, "next_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.8572208285331726" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 890, "start_frame": 4450, "end_frame": 4469, "center_frame": 4459, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.600677, "mean_abs": 0.249916, "relative": 0.894212 }, { "name": "hand_right_joints", "l2": 10.824824, "mean_abs": 0.279658, "relative": 0.853286 }, { "name": "body_joints", "l2": 5.857195, "mean_abs": 0.092091, "relative": 0.791127 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.11512, "mean_abs": 0.013923, "relative": 0.271667 }, { "name": "camera_rotation_matrix", "l2": 3.014457, "mean_abs": 0.226024, "relative": 0.925691 }, { "name": "imu_accel_gyro", "l2": 16.949696, "mean_abs": 1.058858, "relative": 0.914494 }, { "name": "depth_confidence", "l2": 11.162231, "mean_abs": 0.208825, "relative": 0.00037 }, { "name": "video_fisheye_cam0", "l2": 5.252006, "mean_abs": 0.111272, "relative": 0.951026 }, { "name": "video_fisheye_cam1", "l2": 5.2498, "mean_abs": 0.116309, "relative": 0.867485 }, { "name": "video_fisheye_cam2", "l2": 5.728663, "mean_abs": 0.119356, "relative": 0.914496 }, { "name": "video_fisheye_cam3", "l2": 5.649372, "mean_abs": 0.118058, "relative": 0.945919 }, { "name": "video_stereo_left", "l2": 5.504519, "mean_abs": 0.126698, "relative": 0.894495 }, { "name": "video_stereo_right", "l2": 6.374597, "mean_abs": 0.147763, "relative": 0.898684 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.221346, "mean_abs": 0.091388, "relative": 0.794986 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9514385461807251" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9953666925430298" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999661445617676" }, "next_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.8659430146217346" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 891, "start_frame": 4455, "end_frame": 4474, "center_frame": 4464, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.473006, "mean_abs": 0.2492, "relative": 0.883443 }, { "name": "hand_right_joints", "l2": 11.052782, "mean_abs": 0.291457, "relative": 0.871255 }, { "name": "body_joints", "l2": 5.94953, "mean_abs": 0.09768, "relative": 0.803598 }, { "name": "body_contacts", "l2": 3.464165, "mean_abs": 0.081726, "relative": 0.999643 }, { "name": "camera_translation", "l2": 0.177144, "mean_abs": 0.024583, "relative": 0.418034 }, { "name": "camera_rotation_matrix", "l2": 3.005535, "mean_abs": 0.222785, "relative": 0.922951 }, { "name": "imu_accel_gyro", "l2": 16.915064, "mean_abs": 1.101355, "relative": 0.912626 }, { "name": "depth_confidence", "l2": 11.091701, "mean_abs": 0.208929, "relative": 0.000368 }, { "name": "video_fisheye_cam0", "l2": 5.258595, "mean_abs": 0.110874, "relative": 0.952219 }, { "name": "video_fisheye_cam1", "l2": 5.228898, "mean_abs": 0.115793, "relative": 0.864031 }, { "name": "video_fisheye_cam2", "l2": 5.718262, "mean_abs": 0.118528, "relative": 0.912835 }, { "name": "video_fisheye_cam3", "l2": 5.710657, "mean_abs": 0.119445, "relative": 0.956181 }, { "name": "video_stereo_left", "l2": 5.490739, "mean_abs": 0.126607, "relative": 0.892256 }, { "name": "video_stereo_right", "l2": 6.384505, "mean_abs": 0.148351, "relative": 0.900081 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.108984, "mean_abs": 0.087172, "relative": 0.754774 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.986695408821106" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9865444302558899" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.6526451110839844" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 892, "start_frame": 4460, "end_frame": 4479, "center_frame": 4469, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.331118, "mean_abs": 0.246911, "relative": 0.871474 }, { "name": "hand_right_joints", "l2": 11.351719, "mean_abs": 0.305875, "relative": 0.894819 }, { "name": "body_joints", "l2": 6.093373, "mean_abs": 0.103183, "relative": 0.823027 }, { "name": "body_contacts", "l2": 3.464278, "mean_abs": 0.081706, "relative": 0.999676 }, { "name": "camera_translation", "l2": 0.246993, "mean_abs": 0.036586, "relative": 0.582868 }, { "name": "camera_rotation_matrix", "l2": 3.001361, "mean_abs": 0.218021, "relative": 0.921669 }, { "name": "imu_accel_gyro", "l2": 17.060936, "mean_abs": 1.126161, "relative": 0.920496 }, { "name": "depth_confidence", "l2": 10.976864, "mean_abs": 0.208318, "relative": 0.000364 }, { "name": "video_fisheye_cam0", "l2": 5.280448, "mean_abs": 0.111182, "relative": 0.956176 }, { "name": "video_fisheye_cam1", "l2": 5.198872, "mean_abs": 0.114092, "relative": 0.85907 }, { "name": "video_fisheye_cam2", "l2": 5.71809, "mean_abs": 0.11784, "relative": 0.912808 }, { "name": "video_fisheye_cam3", "l2": 5.773883, "mean_abs": 0.120415, "relative": 0.966767 }, { "name": "video_stereo_left", "l2": 5.485531, "mean_abs": 0.124465, "relative": 0.89141 }, { "name": "video_stereo_right", "l2": 6.403835, "mean_abs": 0.147933, "relative": 0.902806 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.93846, "mean_abs": 0.074796, "relative": 0.693746 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.011869, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9973645806312561" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9911562204360962" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.877063512802124" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 893, "start_frame": 4465, "end_frame": 4484, "center_frame": 4474, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.162949, "mean_abs": 0.240014, "relative": 0.857288 }, { "name": "hand_right_joints", "l2": 11.487903, "mean_abs": 0.311294, "relative": 0.905554 }, { "name": "body_joints", "l2": 6.150537, "mean_abs": 0.10366, "relative": 0.830748 }, { "name": "body_contacts", "l2": 3.464278, "mean_abs": 0.081705, "relative": 0.999676 }, { "name": "camera_translation", "l2": 0.27235, "mean_abs": 0.042012, "relative": 0.642708 }, { "name": "camera_rotation_matrix", "l2": 3.007445, "mean_abs": 0.217553, "relative": 0.923538 }, { "name": "imu_accel_gyro", "l2": 17.255001, "mean_abs": 1.158751, "relative": 0.930967 }, { "name": "depth_confidence", "l2": 10.784263, "mean_abs": 0.205641, "relative": 0.000358 }, { "name": "video_fisheye_cam0", "l2": 5.298547, "mean_abs": 0.111972, "relative": 0.959454 }, { "name": "video_fisheye_cam1", "l2": 5.166074, "mean_abs": 0.11278, "relative": 0.85365 }, { "name": "video_fisheye_cam2", "l2": 5.710757, "mean_abs": 0.118401, "relative": 0.911637 }, { "name": "video_fisheye_cam3", "l2": 5.840788, "mean_abs": 0.121112, "relative": 0.977969 }, { "name": "video_stereo_left", "l2": 5.533947, "mean_abs": 0.123901, "relative": 0.899277 }, { "name": "video_stereo_right", "l2": 6.441528, "mean_abs": 0.147843, "relative": 0.90812 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.111738, "mean_abs": 0.084954, "relative": 0.755759 }, { "name": "caption_objects_interaction_text", "l2": 1.983461, "mean_abs": 0.019076, "relative": 0.910757 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9996217489242554" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9947870969772339" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999991655349731" }, "next_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.7328874468803406" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 894, "start_frame": 4470, "end_frame": 4489, "center_frame": 4479, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.95421, "mean_abs": 0.224537, "relative": 0.83968 }, { "name": "hand_right_joints", "l2": 11.683076, "mean_abs": 0.309572, "relative": 0.920939 }, { "name": "body_joints", "l2": 6.115332, "mean_abs": 0.100366, "relative": 0.825993 }, { "name": "body_contacts", "l2": 3.464278, "mean_abs": 0.081705, "relative": 0.999676 }, { "name": "camera_translation", "l2": 0.22864, "mean_abs": 0.036137, "relative": 0.539559 }, { "name": "camera_rotation_matrix", "l2": 3.006411, "mean_abs": 0.216336, "relative": 0.92322 }, { "name": "imu_accel_gyro", "l2": 17.180826, "mean_abs": 1.137107, "relative": 0.926965 }, { "name": "depth_confidence", "l2": 10.615407, "mean_abs": 0.201592, "relative": 0.000352 }, { "name": "video_fisheye_cam0", "l2": 5.30931, "mean_abs": 0.112067, "relative": 0.961402 }, { "name": "video_fisheye_cam1", "l2": 5.135797, "mean_abs": 0.11193, "relative": 0.848647 }, { "name": "video_fisheye_cam2", "l2": 5.681795, "mean_abs": 0.11784, "relative": 0.907014 }, { "name": "video_fisheye_cam3", "l2": 5.819454, "mean_abs": 0.120747, "relative": 0.974397 }, { "name": "video_stereo_left", "l2": 5.5263, "mean_abs": 0.122339, "relative": 0.898035 }, { "name": "video_stereo_right", "l2": 6.426417, "mean_abs": 0.145233, "relative": 0.90599 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.13075, "mean_abs": 0.088168, "relative": 0.762564 }, { "name": "caption_objects_interaction_text", "l2": 1.983461, "mean_abs": 0.019379, "relative": 0.910757 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9996557235717773" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9878416657447815" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999996423721313" }, "next_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.7295282483100891" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle", "correct": 0, "confidence": "" } } }, { "window_index": 895, "start_frame": 4475, "end_frame": 4494, "center_frame": 4484, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.830138, "mean_abs": 0.210489, "relative": 0.829214 }, { "name": "hand_right_joints", "l2": 11.866701, "mean_abs": 0.298543, "relative": 0.935413 }, { "name": "body_joints", "l2": 6.107476, "mean_abs": 0.096505, "relative": 0.824932 }, { "name": "body_contacts", "l2": 3.464306, "mean_abs": 0.081726, "relative": 0.999684 }, { "name": "camera_translation", "l2": 0.136757, "mean_abs": 0.021503, "relative": 0.322728 }, { "name": "camera_rotation_matrix", "l2": 3.004587, "mean_abs": 0.212861, "relative": 0.92266 }, { "name": "imu_accel_gyro", "l2": 17.330612, "mean_abs": 1.161357, "relative": 0.935046 }, { "name": "depth_confidence", "l2": 10.42182, "mean_abs": 0.19553, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.317725, "mean_abs": 0.111342, "relative": 0.962926 }, { "name": "video_fisheye_cam1", "l2": 5.114737, "mean_abs": 0.110963, "relative": 0.845167 }, { "name": "video_fisheye_cam2", "l2": 5.647903, "mean_abs": 0.116157, "relative": 0.901604 }, { "name": "video_fisheye_cam3", "l2": 5.797319, "mean_abs": 0.120011, "relative": 0.970691 }, { "name": "video_stereo_left", "l2": 5.509789, "mean_abs": 0.120413, "relative": 0.895352 }, { "name": "video_stereo_right", "l2": 6.411236, "mean_abs": 0.141798, "relative": 0.90385 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.125843, "mean_abs": 0.084981, "relative": 0.760807 }, { "name": "caption_objects_interaction_text", "l2": 1.983461, "mean_abs": 0.019128, "relative": 0.910757 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9918096661567688" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.8541319966316223" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999933242797852" }, "next_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.538226842880249" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|dripper", "correct": 0, "confidence": "" } } }, { "window_index": 896, "start_frame": 4480, "end_frame": 4499, "center_frame": 4489, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.863692, "mean_abs": 0.212446, "relative": 0.832045 }, { "name": "hand_right_joints", "l2": 11.930534, "mean_abs": 0.294914, "relative": 0.940445 }, { "name": "body_joints", "l2": 6.093277, "mean_abs": 0.096916, "relative": 0.823014 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.094958, "mean_abs": 0.014028, "relative": 0.224089 }, { "name": "camera_rotation_matrix", "l2": 3.006519, "mean_abs": 0.210289, "relative": 0.923253 }, { "name": "imu_accel_gyro", "l2": 17.305758, "mean_abs": 1.15324, "relative": 0.933705 }, { "name": "depth_confidence", "l2": 10.301408, "mean_abs": 0.191953, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.318089, "mean_abs": 0.111272, "relative": 0.962992 }, { "name": "video_fisheye_cam1", "l2": 5.134715, "mean_abs": 0.112608, "relative": 0.848468 }, { "name": "video_fisheye_cam2", "l2": 5.619043, "mean_abs": 0.115965, "relative": 0.896997 }, { "name": "video_fisheye_cam3", "l2": 5.74458, "mean_abs": 0.119234, "relative": 0.961861 }, { "name": "video_stereo_left", "l2": 5.483521, "mean_abs": 0.12141, "relative": 0.891083 }, { "name": "video_stereo_right", "l2": 6.390573, "mean_abs": 0.142932, "relative": 0.900937 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.272407, "mean_abs": 0.096489, "relative": 0.81326 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9538761377334595" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Secure coffee container", "correct": 0, "confidence": "0.7272000908851624" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.7392315864562988" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|dripper", "correct": 0, "confidence": "" } } }, { "window_index": 897, "start_frame": 4485, "end_frame": 4504, "center_frame": 4494, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.147047, "mean_abs": 0.224922, "relative": 0.855947 }, { "name": "hand_right_joints", "l2": 11.936083, "mean_abs": 0.304101, "relative": 0.940883 }, { "name": "body_joints", "l2": 5.999155, "mean_abs": 0.099254, "relative": 0.810301 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.168369, "mean_abs": 0.020573, "relative": 0.397326 }, { "name": "camera_rotation_matrix", "l2": 3.010795, "mean_abs": 0.212033, "relative": 0.924566 }, { "name": "imu_accel_gyro", "l2": 17.260187, "mean_abs": 1.168132, "relative": 0.931247 }, { "name": "depth_confidence", "l2": 10.345585, "mean_abs": 0.19684, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.309804, "mean_abs": 0.111498, "relative": 0.961492 }, { "name": "video_fisheye_cam1", "l2": 5.134906, "mean_abs": 0.113399, "relative": 0.8485 }, { "name": "video_fisheye_cam2", "l2": 5.581506, "mean_abs": 0.115705, "relative": 0.891004 }, { "name": "video_fisheye_cam3", "l2": 5.704366, "mean_abs": 0.120103, "relative": 0.955127 }, { "name": "video_stereo_left", "l2": 5.444031, "mean_abs": 0.123048, "relative": 0.884666 }, { "name": "video_stereo_right", "l2": 6.374411, "mean_abs": 0.146977, "relative": 0.898658 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.033068, "mean_abs": 0.078624, "relative": 0.727604 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9307644963264465" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Secure coffee container", "correct": 0, "confidence": "0.9694623947143555" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.8848164081573486" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 898, "start_frame": 4490, "end_frame": 4509, "center_frame": 4499, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.488077, "mean_abs": 0.237269, "relative": 0.884714 }, { "name": "hand_right_joints", "l2": 11.962999, "mean_abs": 0.317439, "relative": 0.943004 }, { "name": "body_joints", "l2": 5.981694, "mean_abs": 0.101686, "relative": 0.807943 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.195944, "mean_abs": 0.024994, "relative": 0.462401 }, { "name": "camera_rotation_matrix", "l2": 3.005405, "mean_abs": 0.210632, "relative": 0.922911 }, { "name": "imu_accel_gyro", "l2": 17.276438, "mean_abs": 1.172525, "relative": 0.932123 }, { "name": "depth_confidence", "l2": 10.398275, "mean_abs": 0.199293, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.302518, "mean_abs": 0.111383, "relative": 0.960173 }, { "name": "video_fisheye_cam1", "l2": 5.14817, "mean_abs": 0.113526, "relative": 0.850692 }, { "name": "video_fisheye_cam2", "l2": 5.54925, "mean_abs": 0.11483, "relative": 0.885855 }, { "name": "video_fisheye_cam3", "l2": 5.645666, "mean_abs": 0.118739, "relative": 0.945299 }, { "name": "video_stereo_left", "l2": 5.387281, "mean_abs": 0.12253, "relative": 0.875444 }, { "name": "video_stereo_right", "l2": 6.360578, "mean_abs": 0.147845, "relative": 0.896708 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.093798, "mean_abs": 0.083184, "relative": 0.749339 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.81791090965271" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Secure coffee container", "correct": 0, "confidence": "0.9942604303359985" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999902248382568" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9342560768127441" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 899, "start_frame": 4495, "end_frame": 4514, "center_frame": 4504, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.815442, "mean_abs": 0.2442, "relative": 0.912329 }, { "name": "hand_right_joints", "l2": 11.934392, "mean_abs": 0.323986, "relative": 0.940749 }, { "name": "body_joints", "l2": 5.898724, "mean_abs": 0.10133, "relative": 0.796736 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.165474, "mean_abs": 0.021, "relative": 0.390496 }, { "name": "camera_rotation_matrix", "l2": 3.005, "mean_abs": 0.207462, "relative": 0.922787 }, { "name": "imu_accel_gyro", "l2": 17.12392, "mean_abs": 1.155089, "relative": 0.923894 }, { "name": "depth_confidence", "l2": 10.41651, "mean_abs": 0.201512, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.288695, "mean_abs": 0.110891, "relative": 0.95767 }, { "name": "video_fisheye_cam1", "l2": 5.171534, "mean_abs": 0.113309, "relative": 0.854552 }, { "name": "video_fisheye_cam2", "l2": 5.492589, "mean_abs": 0.11357, "relative": 0.87681 }, { "name": "video_fisheye_cam3", "l2": 5.590963, "mean_abs": 0.116191, "relative": 0.936139 }, { "name": "video_stereo_left", "l2": 5.309664, "mean_abs": 0.121393, "relative": 0.862831 }, { "name": "video_stereo_right", "l2": 6.315266, "mean_abs": 0.148087, "relative": 0.89032 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.9862, "mean_abs": 0.078093, "relative": 0.710831 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.5764226317405701" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Secure coffee container", "correct": 0, "confidence": "0.9791902303695679" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9937536120414734" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.914411187171936" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|closed coffee container|mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 900, "start_frame": 4500, "end_frame": 4519, "center_frame": 4509, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.108475, "mean_abs": 0.249346, "relative": 0.937047 }, { "name": "hand_right_joints", "l2": 11.779434, "mean_abs": 0.315328, "relative": 0.928535 }, { "name": "body_joints", "l2": 5.677338, "mean_abs": 0.095717, "relative": 0.766833 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081687, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.102728, "mean_abs": 0.013097, "relative": 0.242423 }, { "name": "camera_rotation_matrix", "l2": 3.005959, "mean_abs": 0.207235, "relative": 0.923081 }, { "name": "imu_accel_gyro", "l2": 17.029427, "mean_abs": 1.099165, "relative": 0.918796 }, { "name": "depth_confidence", "l2": 10.446294, "mean_abs": 0.20198, "relative": 0.000347 }, { "name": "video_fisheye_cam0", "l2": 5.263007, "mean_abs": 0.110309, "relative": 0.953018 }, { "name": "video_fisheye_cam1", "l2": 5.177892, "mean_abs": 0.112667, "relative": 0.855603 }, { "name": "video_fisheye_cam2", "l2": 5.441461, "mean_abs": 0.111932, "relative": 0.868648 }, { "name": "video_fisheye_cam3", "l2": 5.536248, "mean_abs": 0.113708, "relative": 0.926978 }, { "name": "video_stereo_left", "l2": 5.217883, "mean_abs": 0.118802, "relative": 0.847916 }, { "name": "video_stereo_right", "l2": 6.248459, "mean_abs": 0.146279, "relative": 0.880902 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.952324, "mean_abs": 0.076752, "relative": 0.698707 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.8426812887191772" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Secure coffee container", "correct": 0, "confidence": "0.8924335837364197" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.7938292026519775" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.865336537361145" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|closed coffee container|mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 901, "start_frame": 4505, "end_frame": 4524, "center_frame": 4514, "action": "Place item on table", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.431203, "mean_abs": 0.251636, "relative": 0.964271 }, { "name": "hand_right_joints", "l2": 11.58462, "mean_abs": 0.298987, "relative": 0.913178 }, { "name": "body_joints", "l2": 5.3459, "mean_abs": 0.084589, "relative": 0.722066 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.0641, "mean_abs": 0.01018, "relative": 0.151266 }, { "name": "camera_rotation_matrix", "l2": 3.003827, "mean_abs": 0.206329, "relative": 0.922427 }, { "name": "imu_accel_gyro", "l2": 16.957685, "mean_abs": 1.064593, "relative": 0.914926 }, { "name": "depth_confidence", "l2": 10.434401, "mean_abs": 0.199715, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.262308, "mean_abs": 0.109969, "relative": 0.952891 }, { "name": "video_fisheye_cam1", "l2": 5.158761, "mean_abs": 0.111644, "relative": 0.852442 }, { "name": "video_fisheye_cam2", "l2": 5.419727, "mean_abs": 0.110798, "relative": 0.865179 }, { "name": "video_fisheye_cam3", "l2": 5.493137, "mean_abs": 0.112476, "relative": 0.91976 }, { "name": "video_stereo_left", "l2": 5.165001, "mean_abs": 0.115227, "relative": 0.839323 }, { "name": "video_stereo_right", "l2": 6.197313, "mean_abs": 0.142527, "relative": 0.873691 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.019895, "mean_abs": 0.078155, "relative": 0.72289 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Place item on table", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.6585087776184082" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.6958653926849365" }, "transition_detection": { "true": "transition", "predicted": "steady", "correct": 0, "confidence": "0.9996511936187744" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.5828813910484314" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|closed coffee container|water bottle|mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 902, "start_frame": 4510, "end_frame": 4529, "center_frame": 4519, "action": "", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.681479, "mean_abs": 0.253469, "relative": 0.985383 }, { "name": "hand_right_joints", "l2": 11.413663, "mean_abs": 0.285275, "relative": 0.899702 }, { "name": "body_joints", "l2": 5.210726, "mean_abs": 0.07801, "relative": 0.703808 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081689, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.060046, "mean_abs": 0.008307, "relative": 0.1417 }, { "name": "camera_rotation_matrix", "l2": 3.00182, "mean_abs": 0.203012, "relative": 0.92181 }, { "name": "imu_accel_gyro", "l2": 17.077078, "mean_abs": 1.079655, "relative": 0.921367 }, { "name": "depth_confidence", "l2": 10.406549, "mean_abs": 0.197388, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.258649, "mean_abs": 0.109112, "relative": 0.952229 }, { "name": "video_fisheye_cam1", "l2": 5.164152, "mean_abs": 0.110843, "relative": 0.853333 }, { "name": "video_fisheye_cam2", "l2": 5.406436, "mean_abs": 0.109603, "relative": 0.863057 }, { "name": "video_fisheye_cam3", "l2": 5.465036, "mean_abs": 0.110865, "relative": 0.915054 }, { "name": "video_stereo_left", "l2": 5.126144, "mean_abs": 0.113086, "relative": 0.833009 }, { "name": "video_stereo_right", "l2": 6.173199, "mean_abs": 0.140271, "relative": 0.870292 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.228107, "mean_abs": 0.089493, "relative": 0.797406 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.9700832962989807" }, "transition_detection": { "true": "transition", "predicted": "steady", "correct": 0, "confidence": "0.9999969005584717" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.7506520748138428" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|water bottle|mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 903, "start_frame": 4515, "end_frame": 4534, "center_frame": 4524, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.83317, "mean_abs": 0.249125, "relative": 0.998178 }, { "name": "hand_right_joints", "l2": 11.323061, "mean_abs": 0.281161, "relative": 0.89256 }, { "name": "body_joints", "l2": 5.213445, "mean_abs": 0.07664, "relative": 0.704176 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081689, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.041189, "mean_abs": 0.005666, "relative": 0.097199 }, { "name": "camera_rotation_matrix", "l2": 3.000446, "mean_abs": 0.200381, "relative": 0.921388 }, { "name": "imu_accel_gyro", "l2": 17.126579, "mean_abs": 1.04015, "relative": 0.924038 }, { "name": "depth_confidence", "l2": 10.42449, "mean_abs": 0.195384, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.261561, "mean_abs": 0.10879, "relative": 0.952756 }, { "name": "video_fisheye_cam1", "l2": 5.196391, "mean_abs": 0.110865, "relative": 0.85866 }, { "name": "video_fisheye_cam2", "l2": 5.398241, "mean_abs": 0.109055, "relative": 0.861749 }, { "name": "video_fisheye_cam3", "l2": 5.463194, "mean_abs": 0.109889, "relative": 0.914746 }, { "name": "video_stereo_left", "l2": 5.094167, "mean_abs": 0.110719, "relative": 0.827812 }, { "name": "video_stereo_right", "l2": 6.16817, "mean_abs": 0.137544, "relative": 0.869583 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.146349, "mean_abs": 0.084181, "relative": 0.768146 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.8695270419120789" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.9894618391990662" }, "transition_detection": { "true": "transition", "predicted": "steady", "correct": 0, "confidence": "0.9999983310699463" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.8958048820495605" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|water bottle|mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 904, "start_frame": 4520, "end_frame": 4539, "center_frame": 4529, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.854765, "mean_abs": 0.248848, "relative": 1.0 }, { "name": "hand_right_joints", "l2": 11.23517, "mean_abs": 0.278813, "relative": 0.885632 }, { "name": "body_joints", "l2": 5.274897, "mean_abs": 0.077757, "relative": 0.712476 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081689, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021697, "mean_abs": 0.003201, "relative": 0.051201 }, { "name": "camera_rotation_matrix", "l2": 3.00198, "mean_abs": 0.202271, "relative": 0.921859 }, { "name": "imu_accel_gyro", "l2": 17.146044, "mean_abs": 1.033312, "relative": 0.925088 }, { "name": "depth_confidence", "l2": 10.409813, "mean_abs": 0.19553, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.263731, "mean_abs": 0.109262, "relative": 0.953149 }, { "name": "video_fisheye_cam1", "l2": 5.229649, "mean_abs": 0.112463, "relative": 0.864155 }, { "name": "video_fisheye_cam2", "l2": 5.408175, "mean_abs": 0.11031, "relative": 0.863335 }, { "name": "video_fisheye_cam3", "l2": 5.463959, "mean_abs": 0.110693, "relative": 0.914874 }, { "name": "video_stereo_left", "l2": 5.112122, "mean_abs": 0.112527, "relative": 0.83073 }, { "name": "video_stereo_right", "l2": 6.181665, "mean_abs": 0.138738, "relative": 0.871485 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.136445, "mean_abs": 0.084732, "relative": 0.764602 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.8987473249435425" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.9810553789138794" }, "transition_detection": { "true": "transition", "predicted": "steady", "correct": 0, "confidence": "0.9999994039535522" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.9505394697189331" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|water bottle|mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 905, "start_frame": 4525, "end_frame": 4544, "center_frame": 4534, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.851484, "mean_abs": 0.250678, "relative": 0.999723 }, { "name": "hand_right_joints", "l2": 11.167945, "mean_abs": 0.275198, "relative": 0.880333 }, { "name": "body_joints", "l2": 5.315349, "mean_abs": 0.078038, "relative": 0.71794 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081689, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.050393, "mean_abs": 0.007656, "relative": 0.118921 }, { "name": "camera_rotation_matrix", "l2": 3.002467, "mean_abs": 0.203214, "relative": 0.922009 }, { "name": "imu_accel_gyro", "l2": 17.163734, "mean_abs": 1.039569, "relative": 0.926043 }, { "name": "depth_confidence", "l2": 10.46543, "mean_abs": 0.197272, "relative": 0.000347 }, { "name": "video_fisheye_cam0", "l2": 5.270262, "mean_abs": 0.109436, "relative": 0.954332 }, { "name": "video_fisheye_cam1", "l2": 5.279729, "mean_abs": 0.11418, "relative": 0.872431 }, { "name": "video_fisheye_cam2", "l2": 5.436827, "mean_abs": 0.11162, "relative": 0.867909 }, { "name": "video_fisheye_cam3", "l2": 5.479024, "mean_abs": 0.112158, "relative": 0.917397 }, { "name": "video_stereo_left", "l2": 5.140575, "mean_abs": 0.114113, "relative": 0.835354 }, { "name": "video_stereo_right", "l2": 6.215163, "mean_abs": 0.140351, "relative": 0.876208 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.160375, "mean_abs": 0.091616, "relative": 0.773166 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.9153809547424316" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.9248914122581482" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.8685920834541321" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|coffee filter|water bottle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 906, "start_frame": 4530, "end_frame": 4549, "center_frame": 4539, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.761001, "mean_abs": 0.249632, "relative": 0.992091 }, { "name": "hand_right_joints", "l2": 11.13009, "mean_abs": 0.27159, "relative": 0.877349 }, { "name": "body_joints", "l2": 5.336427, "mean_abs": 0.079202, "relative": 0.720787 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081689, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.108717, "mean_abs": 0.015672, "relative": 0.256557 }, { "name": "camera_rotation_matrix", "l2": 3.004788, "mean_abs": 0.20511, "relative": 0.922722 }, { "name": "imu_accel_gyro", "l2": 17.176943, "mean_abs": 1.065133, "relative": 0.926755 }, { "name": "depth_confidence", "l2": 10.590488, "mean_abs": 0.201388, "relative": 0.000351 }, { "name": "video_fisheye_cam0", "l2": 5.26487, "mean_abs": 0.10968, "relative": 0.953355 }, { "name": "video_fisheye_cam1", "l2": 5.338282, "mean_abs": 0.115629, "relative": 0.882106 }, { "name": "video_fisheye_cam2", "l2": 5.493661, "mean_abs": 0.113468, "relative": 0.876981 }, { "name": "video_fisheye_cam3", "l2": 5.489048, "mean_abs": 0.112666, "relative": 0.919075 }, { "name": "video_stereo_left", "l2": 5.210435, "mean_abs": 0.117039, "relative": 0.846706 }, { "name": "video_stereo_right", "l2": 6.276728, "mean_abs": 0.143432, "relative": 0.884887 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.073961, "mean_abs": 0.083764, "relative": 0.742239 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.5445832014083862" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.6761091947555542" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.7457084655761719" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|coffee filter|water bottle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 907, "start_frame": 4535, "end_frame": 4554, "center_frame": 4544, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.593863, "mean_abs": 0.24884, "relative": 0.977992 }, { "name": "hand_right_joints", "l2": 11.108562, "mean_abs": 0.272612, "relative": 0.875652 }, { "name": "body_joints", "l2": 5.369207, "mean_abs": 0.081071, "relative": 0.725214 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081688, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.168111, "mean_abs": 0.023192, "relative": 0.396719 }, { "name": "camera_rotation_matrix", "l2": 3.009139, "mean_abs": 0.208269, "relative": 0.924058 }, { "name": "imu_accel_gyro", "l2": 17.132923, "mean_abs": 1.083605, "relative": 0.92438 }, { "name": "depth_confidence", "l2": 10.760592, "mean_abs": 0.205568, "relative": 0.000357 }, { "name": "video_fisheye_cam0", "l2": 5.251133, "mean_abs": 0.110122, "relative": 0.950868 }, { "name": "video_fisheye_cam1", "l2": 5.415785, "mean_abs": 0.117858, "relative": 0.894913 }, { "name": "video_fisheye_cam2", "l2": 5.557756, "mean_abs": 0.115646, "relative": 0.887213 }, { "name": "video_fisheye_cam3", "l2": 5.478194, "mean_abs": 0.113196, "relative": 0.917258 }, { "name": "video_stereo_left", "l2": 5.290356, "mean_abs": 0.120516, "relative": 0.859693 }, { "name": "video_stereo_right", "l2": 6.385402, "mean_abs": 0.148012, "relative": 0.900208 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.932284, "mean_abs": 0.074298, "relative": 0.691535 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.9048243761062622" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.3683694303035736" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle", "correct": 0, "confidence": "0.3441237807273865" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 908, "start_frame": 4540, "end_frame": 4559, "center_frame": 4549, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.400764, "mean_abs": 0.249773, "relative": 0.961703 }, { "name": "hand_right_joints", "l2": 11.054377, "mean_abs": 0.273516, "relative": 0.871381 }, { "name": "body_joints", "l2": 5.419452, "mean_abs": 0.082831, "relative": 0.732001 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.201295, "mean_abs": 0.026108, "relative": 0.475027 }, { "name": "camera_rotation_matrix", "l2": 3.007367, "mean_abs": 0.209578, "relative": 0.923513 }, { "name": "imu_accel_gyro", "l2": 17.151495, "mean_abs": 1.096216, "relative": 0.925382 }, { "name": "depth_confidence", "l2": 10.919, "mean_abs": 0.208535, "relative": 0.000362 }, { "name": "video_fisheye_cam0", "l2": 5.239074, "mean_abs": 0.110188, "relative": 0.948684 }, { "name": "video_fisheye_cam1", "l2": 5.507749, "mean_abs": 0.119931, "relative": 0.910109 }, { "name": "video_fisheye_cam2", "l2": 5.607884, "mean_abs": 0.11737, "relative": 0.895215 }, { "name": "video_fisheye_cam3", "l2": 5.471974, "mean_abs": 0.113304, "relative": 0.916216 }, { "name": "video_stereo_left", "l2": 5.346537, "mean_abs": 0.122792, "relative": 0.868823 }, { "name": "video_stereo_right", "l2": 6.480643, "mean_abs": 0.151334, "relative": 0.913635 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.94405, "mean_abs": 0.075462, "relative": 0.695746 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.9595302939414978" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.31534504890441895" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.34304994344711304" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 909, "start_frame": 4545, "end_frame": 4564, "center_frame": 4554, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.148026, "mean_abs": 0.24772, "relative": 0.940383 }, { "name": "hand_right_joints", "l2": 11.013946, "mean_abs": 0.271512, "relative": 0.868194 }, { "name": "body_joints", "l2": 5.419505, "mean_abs": 0.082256, "relative": 0.732008 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.21064, "mean_abs": 0.028757, "relative": 0.49708 }, { "name": "camera_rotation_matrix", "l2": 3.00507, "mean_abs": 0.211283, "relative": 0.922808 }, { "name": "imu_accel_gyro", "l2": 17.117155, "mean_abs": 1.0766, "relative": 0.923529 }, { "name": "depth_confidence", "l2": 11.202847, "mean_abs": 0.21123, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.231707, "mean_abs": 0.109811, "relative": 0.94735 }, { "name": "video_fisheye_cam1", "l2": 5.571932, "mean_abs": 0.120949, "relative": 0.920715 }, { "name": "video_fisheye_cam2", "l2": 5.66112, "mean_abs": 0.118526, "relative": 0.903714 }, { "name": "video_fisheye_cam3", "l2": 5.456347, "mean_abs": 0.112643, "relative": 0.913599 }, { "name": "video_stereo_left", "l2": 5.393197, "mean_abs": 0.123601, "relative": 0.876405 }, { "name": "video_stereo_right", "l2": 6.543885, "mean_abs": 0.152362, "relative": 0.92255 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.973544, "mean_abs": 0.077881, "relative": 0.706302 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.8726866841316223" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.4842537045478821" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999998807907104" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.6205340623855591" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 910, "start_frame": 4550, "end_frame": 4569, "center_frame": 4559, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.918293, "mean_abs": 0.24542, "relative": 0.921005 }, { "name": "hand_right_joints", "l2": 10.898141, "mean_abs": 0.269329, "relative": 0.859065 }, { "name": "body_joints", "l2": 5.387863, "mean_abs": 0.079486, "relative": 0.727734 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.179347, "mean_abs": 0.025273, "relative": 0.423233 }, { "name": "camera_rotation_matrix", "l2": 3.005229, "mean_abs": 0.213765, "relative": 0.922857 }, { "name": "imu_accel_gyro", "l2": 17.497719, "mean_abs": 1.082492, "relative": 0.944062 }, { "name": "depth_confidence", "l2": 11.469417, "mean_abs": 0.213463, "relative": 0.000381 }, { "name": "video_fisheye_cam0", "l2": 5.21986, "mean_abs": 0.109431, "relative": 0.945205 }, { "name": "video_fisheye_cam1", "l2": 5.645026, "mean_abs": 0.1219, "relative": 0.932793 }, { "name": "video_fisheye_cam2", "l2": 5.709583, "mean_abs": 0.118856, "relative": 0.91145 }, { "name": "video_fisheye_cam3", "l2": 5.452427, "mean_abs": 0.11265, "relative": 0.912943 }, { "name": "video_stereo_left", "l2": 5.415386, "mean_abs": 0.12307, "relative": 0.880011 }, { "name": "video_stereo_right", "l2": 6.582904, "mean_abs": 0.152043, "relative": 0.928051 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.016583, "mean_abs": 0.07719, "relative": 0.721705 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8780595064163208" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.8423803448677063" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.999998927116394" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9833893179893494" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|coffee mug|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 911, "start_frame": 4555, "end_frame": 4574, "center_frame": 4564, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.693426, "mean_abs": 0.23747, "relative": 0.902036 }, { "name": "hand_right_joints", "l2": 10.811156, "mean_abs": 0.262871, "relative": 0.852208 }, { "name": "body_joints", "l2": 5.406996, "mean_abs": 0.0777, "relative": 0.730319 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.124501, "mean_abs": 0.017973, "relative": 0.293804 }, { "name": "camera_rotation_matrix", "l2": 3.002115, "mean_abs": 0.213839, "relative": 0.921901 }, { "name": "imu_accel_gyro", "l2": 17.364172, "mean_abs": 1.060095, "relative": 0.936857 }, { "name": "depth_confidence", "l2": 11.700607, "mean_abs": 0.214677, "relative": 0.000388 }, { "name": "video_fisheye_cam0", "l2": 5.210519, "mean_abs": 0.108785, "relative": 0.943514 }, { "name": "video_fisheye_cam1", "l2": 5.690317, "mean_abs": 0.121329, "relative": 0.940277 }, { "name": "video_fisheye_cam2", "l2": 5.722008, "mean_abs": 0.11841, "relative": 0.913433 }, { "name": "video_fisheye_cam3", "l2": 5.446788, "mean_abs": 0.111825, "relative": 0.911999 }, { "name": "video_stereo_left", "l2": 5.429362, "mean_abs": 0.122043, "relative": 0.882282 }, { "name": "video_stereo_right", "l2": 6.610223, "mean_abs": 0.150609, "relative": 0.931903 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.015696, "mean_abs": 0.077216, "relative": 0.721387 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.990176260471344" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.978188157081604" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999430179595947" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.994249701499939" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 912, "start_frame": 4560, "end_frame": 4579, "center_frame": 4569, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.463283, "mean_abs": 0.225433, "relative": 0.882623 }, { "name": "hand_right_joints", "l2": 10.750871, "mean_abs": 0.256175, "relative": 0.847456 }, { "name": "body_joints", "l2": 5.422744, "mean_abs": 0.076954, "relative": 0.732446 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.069705, "mean_abs": 0.010796, "relative": 0.164494 }, { "name": "camera_rotation_matrix", "l2": 2.997634, "mean_abs": 0.212037, "relative": 0.920525 }, { "name": "imu_accel_gyro", "l2": 17.303503, "mean_abs": 1.029212, "relative": 0.933584 }, { "name": "depth_confidence", "l2": 11.79259, "mean_abs": 0.214447, "relative": 0.000391 }, { "name": "video_fisheye_cam0", "l2": 5.198378, "mean_abs": 0.108004, "relative": 0.941315 }, { "name": "video_fisheye_cam1", "l2": 5.702193, "mean_abs": 0.120539, "relative": 0.942239 }, { "name": "video_fisheye_cam2", "l2": 5.727269, "mean_abs": 0.117849, "relative": 0.914273 }, { "name": "video_fisheye_cam3", "l2": 5.4516, "mean_abs": 0.110952, "relative": 0.912805 }, { "name": "video_stereo_left", "l2": 5.425643, "mean_abs": 0.120216, "relative": 0.881678 }, { "name": "video_stereo_right", "l2": 6.60411, "mean_abs": 0.148567, "relative": 0.931041 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.049954, "mean_abs": 0.079919, "relative": 0.733648 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9969286322593689" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9794410467147827" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9985472559928894" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9974302649497986" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 913, "start_frame": 4565, "end_frame": 4584, "center_frame": 4574, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.389913, "mean_abs": 0.22538, "relative": 0.876433 }, { "name": "hand_right_joints", "l2": 10.724921, "mean_abs": 0.259139, "relative": 0.845411 }, { "name": "body_joints", "l2": 5.444532, "mean_abs": 0.077591, "relative": 0.735389 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.07112, "mean_abs": 0.009346, "relative": 0.167834 }, { "name": "camera_rotation_matrix", "l2": 3.000639, "mean_abs": 0.214466, "relative": 0.921448 }, { "name": "imu_accel_gyro", "l2": 17.402567, "mean_abs": 1.01048, "relative": 0.938928 }, { "name": "depth_confidence", "l2": 11.877187, "mean_abs": 0.215164, "relative": 0.000394 }, { "name": "video_fisheye_cam0", "l2": 5.188468, "mean_abs": 0.10853, "relative": 0.939521 }, { "name": "video_fisheye_cam1", "l2": 5.674682, "mean_abs": 0.120471, "relative": 0.937693 }, { "name": "video_fisheye_cam2", "l2": 5.731967, "mean_abs": 0.1181, "relative": 0.915023 }, { "name": "video_fisheye_cam3", "l2": 5.44599, "mean_abs": 0.111164, "relative": 0.911865 }, { "name": "video_stereo_left", "l2": 5.419973, "mean_abs": 0.119926, "relative": 0.880756 }, { "name": "video_stereo_right", "l2": 6.556792, "mean_abs": 0.147197, "relative": 0.92437 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.999708, "mean_abs": 0.07625, "relative": 0.715666 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9940643906593323" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9734013080596924" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.994596540927887" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9971926808357239" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|coffee mug|mug", "correct": 0, "confidence": "" } } }, { "window_index": 914, "start_frame": 4570, "end_frame": 4589, "center_frame": 4579, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.364904, "mean_abs": 0.226066, "relative": 0.874324 }, { "name": "hand_right_joints", "l2": 10.672832, "mean_abs": 0.257885, "relative": 0.841305 }, { "name": "body_joints", "l2": 5.465716, "mean_abs": 0.077555, "relative": 0.73825 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081686, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.093318, "mean_abs": 0.011845, "relative": 0.220218 }, { "name": "camera_rotation_matrix", "l2": 3.003951, "mean_abs": 0.216127, "relative": 0.922464 }, { "name": "imu_accel_gyro", "l2": 17.197163, "mean_abs": 0.992218, "relative": 0.927846 }, { "name": "depth_confidence", "l2": 11.922527, "mean_abs": 0.215953, "relative": 0.000396 }, { "name": "video_fisheye_cam0", "l2": 5.179069, "mean_abs": 0.108892, "relative": 0.937819 }, { "name": "video_fisheye_cam1", "l2": 5.642939, "mean_abs": 0.121022, "relative": 0.932448 }, { "name": "video_fisheye_cam2", "l2": 5.73353, "mean_abs": 0.118597, "relative": 0.915273 }, { "name": "video_fisheye_cam3", "l2": 5.427614, "mean_abs": 0.111854, "relative": 0.908789 }, { "name": "video_stereo_left", "l2": 5.424313, "mean_abs": 0.121077, "relative": 0.881462 }, { "name": "video_stereo_right", "l2": 6.510315, "mean_abs": 0.14677, "relative": 0.917818 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.924094, "mean_abs": 0.073639, "relative": 0.688605 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9583333134651184" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.8996859788894653" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9576320648193359" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9891235828399658" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|coffee mug|mug", "correct": 0, "confidence": "" } } }, { "window_index": 915, "start_frame": 4575, "end_frame": 4594, "center_frame": 4584, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.381279, "mean_abs": 0.227192, "relative": 0.875705 }, { "name": "hand_right_joints", "l2": 10.70096, "mean_abs": 0.257482, "relative": 0.843522 }, { "name": "body_joints", "l2": 5.462093, "mean_abs": 0.07716, "relative": 0.73776 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.070756, "mean_abs": 0.009251, "relative": 0.166975 }, { "name": "camera_rotation_matrix", "l2": 3.002658, "mean_abs": 0.216155, "relative": 0.922067 }, { "name": "imu_accel_gyro", "l2": 17.027716, "mean_abs": 0.965927, "relative": 0.918704 }, { "name": "depth_confidence", "l2": 11.889992, "mean_abs": 0.214437, "relative": 0.000395 }, { "name": "video_fisheye_cam0", "l2": 5.161825, "mean_abs": 0.108465, "relative": 0.934696 }, { "name": "video_fisheye_cam1", "l2": 5.610677, "mean_abs": 0.120038, "relative": 0.927117 }, { "name": "video_fisheye_cam2", "l2": 5.731861, "mean_abs": 0.11792, "relative": 0.915006 }, { "name": "video_fisheye_cam3", "l2": 5.407587, "mean_abs": 0.110819, "relative": 0.905435 }, { "name": "video_stereo_left", "l2": 5.403044, "mean_abs": 0.120045, "relative": 0.878005 }, { "name": "video_stereo_right", "l2": 6.468472, "mean_abs": 0.1449, "relative": 0.911919 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.95351, "mean_abs": 0.076754, "relative": 0.699132 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9850825071334839" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9813926815986633" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9930939674377441" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9953872561454773" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 916, "start_frame": 4580, "end_frame": 4599, "center_frame": 4589, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.381848, "mean_abs": 0.225994, "relative": 0.875753 }, { "name": "hand_right_joints", "l2": 10.722936, "mean_abs": 0.255142, "relative": 0.845254 }, { "name": "body_joints", "l2": 5.447048, "mean_abs": 0.07702, "relative": 0.735728 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.036316, "mean_abs": 0.004945, "relative": 0.085702 }, { "name": "camera_rotation_matrix", "l2": 3.000627, "mean_abs": 0.215136, "relative": 0.921444 }, { "name": "imu_accel_gyro", "l2": 17.091095, "mean_abs": 0.965693, "relative": 0.922123 }, { "name": "depth_confidence", "l2": 11.903061, "mean_abs": 0.213748, "relative": 0.000395 }, { "name": "video_fisheye_cam0", "l2": 5.144821, "mean_abs": 0.107577, "relative": 0.931617 }, { "name": "video_fisheye_cam1", "l2": 5.575714, "mean_abs": 0.118232, "relative": 0.92134 }, { "name": "video_fisheye_cam2", "l2": 5.732159, "mean_abs": 0.117231, "relative": 0.915054 }, { "name": "video_fisheye_cam3", "l2": 5.405823, "mean_abs": 0.110022, "relative": 0.90514 }, { "name": "video_stereo_left", "l2": 5.388685, "mean_abs": 0.118504, "relative": 0.875672 }, { "name": "video_stereo_right", "l2": 6.433662, "mean_abs": 0.142947, "relative": 0.907011 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.045619, "mean_abs": 0.079343, "relative": 0.732096 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9915817379951477" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9955487251281738" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9962140917778015" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9976450800895691" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 917, "start_frame": 4585, "end_frame": 4604, "center_frame": 4594, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.376609, "mean_abs": 0.224939, "relative": 0.875311 }, { "name": "hand_right_joints", "l2": 10.74659, "mean_abs": 0.252966, "relative": 0.847119 }, { "name": "body_joints", "l2": 5.419487, "mean_abs": 0.076366, "relative": 0.732006 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018182, "mean_abs": 0.00272, "relative": 0.042908 }, { "name": "camera_rotation_matrix", "l2": 2.999272, "mean_abs": 0.214151, "relative": 0.921028 }, { "name": "imu_accel_gyro", "l2": 17.057468, "mean_abs": 0.9805, "relative": 0.920309 }, { "name": "depth_confidence", "l2": 11.890959, "mean_abs": 0.213378, "relative": 0.000395 }, { "name": "video_fisheye_cam0", "l2": 5.133748, "mean_abs": 0.106888, "relative": 0.929612 }, { "name": "video_fisheye_cam1", "l2": 5.547805, "mean_abs": 0.116859, "relative": 0.916728 }, { "name": "video_fisheye_cam2", "l2": 5.735109, "mean_abs": 0.116915, "relative": 0.915525 }, { "name": "video_fisheye_cam3", "l2": 5.404015, "mean_abs": 0.109642, "relative": 0.904837 }, { "name": "video_stereo_left", "l2": 5.379452, "mean_abs": 0.1179, "relative": 0.874172 }, { "name": "video_stereo_right", "l2": 6.413963, "mean_abs": 0.141957, "relative": 0.904234 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.999768, "mean_abs": 0.075192, "relative": 0.715687 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9924134612083435" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9978922009468079" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9995833039283752" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9967604279518127" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 918, "start_frame": 4590, "end_frame": 4609, "center_frame": 4599, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.388195, "mean_abs": 0.224935, "relative": 0.876289 }, { "name": "hand_right_joints", "l2": 10.757051, "mean_abs": 0.252518, "relative": 0.847943 }, { "name": "body_joints", "l2": 5.387956, "mean_abs": 0.076024, "relative": 0.727747 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012415, "mean_abs": 0.001871, "relative": 0.029297 }, { "name": "camera_rotation_matrix", "l2": 2.999632, "mean_abs": 0.214058, "relative": 0.921138 }, { "name": "imu_accel_gyro", "l2": 17.052563, "mean_abs": 0.960446, "relative": 0.920044 }, { "name": "depth_confidence", "l2": 11.871632, "mean_abs": 0.212786, "relative": 0.000394 }, { "name": "video_fisheye_cam0", "l2": 5.134589, "mean_abs": 0.106917, "relative": 0.929764 }, { "name": "video_fisheye_cam1", "l2": 5.552886, "mean_abs": 0.116911, "relative": 0.917567 }, { "name": "video_fisheye_cam2", "l2": 5.742647, "mean_abs": 0.117071, "relative": 0.916728 }, { "name": "video_fisheye_cam3", "l2": 5.43881, "mean_abs": 0.111069, "relative": 0.910663 }, { "name": "video_stereo_left", "l2": 5.390448, "mean_abs": 0.118478, "relative": 0.875958 }, { "name": "video_stereo_right", "l2": 6.439392, "mean_abs": 0.143199, "relative": 0.907819 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.055212, "mean_abs": 0.079696, "relative": 0.735529 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9979178309440613" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.998519241809845" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999051094055176" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9989615678787231" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 919, "start_frame": 4595, "end_frame": 4614, "center_frame": 4604, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.410859, "mean_abs": 0.226594, "relative": 0.8782 }, { "name": "hand_right_joints", "l2": 10.781861, "mean_abs": 0.254667, "relative": 0.849899 }, { "name": "body_joints", "l2": 5.365881, "mean_abs": 0.076228, "relative": 0.724765 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016019, "mean_abs": 0.002302, "relative": 0.037801 }, { "name": "camera_rotation_matrix", "l2": 3.000409, "mean_abs": 0.214242, "relative": 0.921377 }, { "name": "imu_accel_gyro", "l2": 17.05868, "mean_abs": 0.971892, "relative": 0.920375 }, { "name": "depth_confidence", "l2": 11.854939, "mean_abs": 0.212774, "relative": 0.000393 }, { "name": "video_fisheye_cam0", "l2": 5.132188, "mean_abs": 0.106993, "relative": 0.929329 }, { "name": "video_fisheye_cam1", "l2": 5.550754, "mean_abs": 0.116873, "relative": 0.917215 }, { "name": "video_fisheye_cam2", "l2": 5.744701, "mean_abs": 0.116988, "relative": 0.917056 }, { "name": "video_fisheye_cam3", "l2": 5.46732, "mean_abs": 0.110956, "relative": 0.915437 }, { "name": "video_stereo_left", "l2": 5.391822, "mean_abs": 0.118269, "relative": 0.876182 }, { "name": "video_stereo_right", "l2": 6.450228, "mean_abs": 0.143191, "relative": 0.909347 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.053756, "mean_abs": 0.079553, "relative": 0.735008 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9941985607147217" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9896185994148254" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9852744340896606" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9988300204277039" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 920, "start_frame": 4600, "end_frame": 4619, "center_frame": 4609, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.422826, "mean_abs": 0.227682, "relative": 0.87921 }, { "name": "hand_right_joints", "l2": 10.794606, "mean_abs": 0.255589, "relative": 0.850904 }, { "name": "body_joints", "l2": 5.347019, "mean_abs": 0.076649, "relative": 0.722218 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014195, "mean_abs": 0.002025, "relative": 0.033499 }, { "name": "camera_rotation_matrix", "l2": 3.000638, "mean_abs": 0.21364, "relative": 0.921447 }, { "name": "imu_accel_gyro", "l2": 17.093945, "mean_abs": 0.982187, "relative": 0.922277 }, { "name": "depth_confidence", "l2": 11.774242, "mean_abs": 0.210925, "relative": 0.000391 }, { "name": "video_fisheye_cam0", "l2": 5.13136, "mean_abs": 0.106653, "relative": 0.92918 }, { "name": "video_fisheye_cam1", "l2": 5.545373, "mean_abs": 0.116393, "relative": 0.916326 }, { "name": "video_fisheye_cam2", "l2": 5.745841, "mean_abs": 0.116804, "relative": 0.917238 }, { "name": "video_fisheye_cam3", "l2": 5.47136, "mean_abs": 0.110086, "relative": 0.916113 }, { "name": "video_stereo_left", "l2": 5.392694, "mean_abs": 0.117722, "relative": 0.876323 }, { "name": "video_stereo_right", "l2": 6.459698, "mean_abs": 0.14275, "relative": 0.910682 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.008665, "mean_abs": 0.078976, "relative": 0.718871 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9954262375831604" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9958788156509399" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9730302691459656" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9994189739227295" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 921, "start_frame": 4605, "end_frame": 4624, "center_frame": 4614, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.468436, "mean_abs": 0.22914, "relative": 0.883057 }, { "name": "hand_right_joints", "l2": 10.81024, "mean_abs": 0.257188, "relative": 0.852136 }, { "name": "body_joints", "l2": 5.31793, "mean_abs": 0.076641, "relative": 0.718288 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013314, "mean_abs": 0.001775, "relative": 0.031419 }, { "name": "camera_rotation_matrix", "l2": 2.99983, "mean_abs": 0.212778, "relative": 0.921199 }, { "name": "imu_accel_gyro", "l2": 17.102982, "mean_abs": 0.970784, "relative": 0.922765 }, { "name": "depth_confidence", "l2": 11.724821, "mean_abs": 0.20992, "relative": 0.000389 }, { "name": "video_fisheye_cam0", "l2": 5.132964, "mean_abs": 0.106821, "relative": 0.92947 }, { "name": "video_fisheye_cam1", "l2": 5.554211, "mean_abs": 0.116675, "relative": 0.917786 }, { "name": "video_fisheye_cam2", "l2": 5.755437, "mean_abs": 0.11691, "relative": 0.91877 }, { "name": "video_fisheye_cam3", "l2": 5.474241, "mean_abs": 0.110497, "relative": 0.916596 }, { "name": "video_stereo_left", "l2": 5.40441, "mean_abs": 0.117423, "relative": 0.878227 }, { "name": "video_stereo_right", "l2": 6.475426, "mean_abs": 0.14273, "relative": 0.912899 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.997585, "mean_abs": 0.076562, "relative": 0.714906 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9974287152290344" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9989450573921204" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.8863182663917542" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9992449283599854" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 922, "start_frame": 4610, "end_frame": 4629, "center_frame": 4619, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.447267, "mean_abs": 0.230027, "relative": 0.881271 }, { "name": "hand_right_joints", "l2": 10.846135, "mean_abs": 0.258001, "relative": 0.854966 }, { "name": "body_joints", "l2": 5.288373, "mean_abs": 0.076925, "relative": 0.714296 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01122, "mean_abs": 0.001455, "relative": 0.026478 }, { "name": "camera_rotation_matrix", "l2": 2.999977, "mean_abs": 0.213223, "relative": 0.921244 }, { "name": "imu_accel_gyro", "l2": 17.098606, "mean_abs": 0.971284, "relative": 0.922529 }, { "name": "depth_confidence", "l2": 11.71592, "mean_abs": 0.209728, "relative": 0.000389 }, { "name": "video_fisheye_cam0", "l2": 5.130479, "mean_abs": 0.106796, "relative": 0.92902 }, { "name": "video_fisheye_cam1", "l2": 5.554231, "mean_abs": 0.116707, "relative": 0.91779 }, { "name": "video_fisheye_cam2", "l2": 5.76237, "mean_abs": 0.117111, "relative": 0.919877 }, { "name": "video_fisheye_cam3", "l2": 5.459409, "mean_abs": 0.110185, "relative": 0.914112 }, { "name": "video_stereo_left", "l2": 5.408596, "mean_abs": 0.117481, "relative": 0.878908 }, { "name": "video_stereo_right", "l2": 6.479433, "mean_abs": 0.142757, "relative": 0.913464 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.003373, "mean_abs": 0.076787, "relative": 0.716977 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9982560276985168" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9991081357002258" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.5454618334770203" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9997126460075378" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 923, "start_frame": 4615, "end_frame": 4634, "center_frame": 4624, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.459283, "mean_abs": 0.231747, "relative": 0.882285 }, { "name": "hand_right_joints", "l2": 10.867474, "mean_abs": 0.259107, "relative": 0.856648 }, { "name": "body_joints", "l2": 5.229632, "mean_abs": 0.076931, "relative": 0.706362 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008949, "mean_abs": 0.001275, "relative": 0.021119 }, { "name": "camera_rotation_matrix", "l2": 3.000172, "mean_abs": 0.213505, "relative": 0.921304 }, { "name": "imu_accel_gyro", "l2": 17.108927, "mean_abs": 0.969152, "relative": 0.923086 }, { "name": "depth_confidence", "l2": 11.734919, "mean_abs": 0.210367, "relative": 0.000389 }, { "name": "video_fisheye_cam0", "l2": 5.127012, "mean_abs": 0.106583, "relative": 0.928392 }, { "name": "video_fisheye_cam1", "l2": 5.548065, "mean_abs": 0.116501, "relative": 0.916771 }, { "name": "video_fisheye_cam2", "l2": 5.761675, "mean_abs": 0.117036, "relative": 0.919766 }, { "name": "video_fisheye_cam3", "l2": 5.456513, "mean_abs": 0.110047, "relative": 0.913627 }, { "name": "video_stereo_left", "l2": 5.404446, "mean_abs": 0.117136, "relative": 0.878233 }, { "name": "video_stereo_right", "l2": 6.470996, "mean_abs": 0.142351, "relative": 0.912275 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.013295, "mean_abs": 0.075752, "relative": 0.720528 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9992092847824097" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9996412992477417" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.764386773109436" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9997736811637878" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 924, "start_frame": 4620, "end_frame": 4639, "center_frame": 4629, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.467037, "mean_abs": 0.231944, "relative": 0.882939 }, { "name": "hand_right_joints", "l2": 10.882881, "mean_abs": 0.2598, "relative": 0.857862 }, { "name": "body_joints", "l2": 5.177197, "mean_abs": 0.07624, "relative": 0.69928 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00986, "mean_abs": 0.001279, "relative": 0.023267 }, { "name": "camera_rotation_matrix", "l2": 3.000252, "mean_abs": 0.213652, "relative": 0.921329 }, { "name": "imu_accel_gyro", "l2": 17.097759, "mean_abs": 0.957933, "relative": 0.922483 }, { "name": "depth_confidence", "l2": 11.766392, "mean_abs": 0.211144, "relative": 0.000391 }, { "name": "video_fisheye_cam0", "l2": 5.123177, "mean_abs": 0.106388, "relative": 0.927698 }, { "name": "video_fisheye_cam1", "l2": 5.542519, "mean_abs": 0.116307, "relative": 0.915855 }, { "name": "video_fisheye_cam2", "l2": 5.763511, "mean_abs": 0.11705, "relative": 0.920059 }, { "name": "video_fisheye_cam3", "l2": 5.463862, "mean_abs": 0.110017, "relative": 0.914858 }, { "name": "video_stereo_left", "l2": 5.400321, "mean_abs": 0.116854, "relative": 0.877563 }, { "name": "video_stereo_right", "l2": 6.456567, "mean_abs": 0.141718, "relative": 0.910241 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.027548, "mean_abs": 0.076911, "relative": 0.725629 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.999803364276886" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9997490048408508" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.8961238265037537" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.999883770942688" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 925, "start_frame": 4625, "end_frame": 4644, "center_frame": 4634, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.476148, "mean_abs": 0.231978, "relative": 0.883708 }, { "name": "hand_right_joints", "l2": 10.893518, "mean_abs": 0.260202, "relative": 0.858701 }, { "name": "body_joints", "l2": 5.118106, "mean_abs": 0.075347, "relative": 0.691298 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014775, "mean_abs": 0.001979, "relative": 0.034867 }, { "name": "camera_rotation_matrix", "l2": 3.000143, "mean_abs": 0.213836, "relative": 0.921295 }, { "name": "imu_accel_gyro", "l2": 17.092012, "mean_abs": 0.958311, "relative": 0.922173 }, { "name": "depth_confidence", "l2": 11.786975, "mean_abs": 0.210993, "relative": 0.000391 }, { "name": "video_fisheye_cam0", "l2": 5.122912, "mean_abs": 0.106307, "relative": 0.92765 }, { "name": "video_fisheye_cam1", "l2": 5.536072, "mean_abs": 0.115921, "relative": 0.914789 }, { "name": "video_fisheye_cam2", "l2": 5.765366, "mean_abs": 0.116962, "relative": 0.920355 }, { "name": "video_fisheye_cam3", "l2": 5.484297, "mean_abs": 0.110526, "relative": 0.918279 }, { "name": "video_stereo_left", "l2": 5.394131, "mean_abs": 0.116429, "relative": 0.876557 }, { "name": "video_stereo_right", "l2": 6.444902, "mean_abs": 0.141046, "relative": 0.908596 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.051494, "mean_abs": 0.080408, "relative": 0.734199 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9998875856399536" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9998478889465332" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.7177802324295044" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.999863862991333" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 926, "start_frame": 4630, "end_frame": 4649, "center_frame": 4639, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.52126, "mean_abs": 0.234161, "relative": 0.887513 }, { "name": "hand_right_joints", "l2": 10.903567, "mean_abs": 0.261349, "relative": 0.859493 }, { "name": "body_joints", "l2": 5.047937, "mean_abs": 0.073917, "relative": 0.681821 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014628, "mean_abs": 0.002115, "relative": 0.034519 }, { "name": "camera_rotation_matrix", "l2": 3.00011, "mean_abs": 0.214552, "relative": 0.921285 }, { "name": "imu_accel_gyro", "l2": 17.084555, "mean_abs": 0.956779, "relative": 0.921771 }, { "name": "depth_confidence", "l2": 11.85704, "mean_abs": 0.21211, "relative": 0.000394 }, { "name": "video_fisheye_cam0", "l2": 5.122473, "mean_abs": 0.106181, "relative": 0.92757 }, { "name": "video_fisheye_cam1", "l2": 5.532895, "mean_abs": 0.115888, "relative": 0.914264 }, { "name": "video_fisheye_cam2", "l2": 5.764807, "mean_abs": 0.116975, "relative": 0.920266 }, { "name": "video_fisheye_cam3", "l2": 5.493836, "mean_abs": 0.110348, "relative": 0.919877 }, { "name": "video_stereo_left", "l2": 5.389614, "mean_abs": 0.116773, "relative": 0.875823 }, { "name": "video_stereo_right", "l2": 6.433795, "mean_abs": 0.141137, "relative": 0.90703 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.039915, "mean_abs": 0.080018, "relative": 0.730055 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999442100524902" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9998739957809448" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.9225466847419739" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999237060546875" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 927, "start_frame": 4635, "end_frame": 4654, "center_frame": 4644, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.529628, "mean_abs": 0.233846, "relative": 0.888219 }, { "name": "hand_right_joints", "l2": 10.901726, "mean_abs": 0.261138, "relative": 0.859348 }, { "name": "body_joints", "l2": 4.963608, "mean_abs": 0.071842, "relative": 0.67043 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016578, "mean_abs": 0.002448, "relative": 0.039122 }, { "name": "camera_rotation_matrix", "l2": 3.001973, "mean_abs": 0.218952, "relative": 0.921857 }, { "name": "imu_accel_gyro", "l2": 17.052477, "mean_abs": 0.992374, "relative": 0.92004 }, { "name": "depth_confidence", "l2": 11.93334, "mean_abs": 0.216346, "relative": 0.000396 }, { "name": "video_fisheye_cam0", "l2": 5.125942, "mean_abs": 0.10691, "relative": 0.928198 }, { "name": "video_fisheye_cam1", "l2": 5.530517, "mean_abs": 0.11695, "relative": 0.913871 }, { "name": "video_fisheye_cam2", "l2": 5.791889, "mean_abs": 0.118794, "relative": 0.924589 }, { "name": "video_fisheye_cam3", "l2": 5.526805, "mean_abs": 0.111002, "relative": 0.925397 }, { "name": "video_stereo_left", "l2": 5.411686, "mean_abs": 0.119501, "relative": 0.87941 }, { "name": "video_stereo_right", "l2": 6.457534, "mean_abs": 0.144902, "relative": 0.910377 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.964753, "mean_abs": 0.078282, "relative": 0.703156 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999759197235107" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9998562335968018" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9880685210227966" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999648332595825" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 928, "start_frame": 4640, "end_frame": 4659, "center_frame": 4649, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.581849, "mean_abs": 0.241612, "relative": 0.892624 }, { "name": "hand_right_joints", "l2": 10.858301, "mean_abs": 0.265649, "relative": 0.855925 }, { "name": "body_joints", "l2": 4.907533, "mean_abs": 0.071249, "relative": 0.662857 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.032967, "mean_abs": 0.004369, "relative": 0.077797 }, { "name": "camera_rotation_matrix", "l2": 3.014647, "mean_abs": 0.230639, "relative": 0.925749 }, { "name": "imu_accel_gyro", "l2": 17.135206, "mean_abs": 1.093861, "relative": 0.924503 }, { "name": "depth_confidence", "l2": 12.118387, "mean_abs": 0.223699, "relative": 0.000402 }, { "name": "video_fisheye_cam0", "l2": 5.159096, "mean_abs": 0.10975, "relative": 0.934202 }, { "name": "video_fisheye_cam1", "l2": 5.560709, "mean_abs": 0.120959, "relative": 0.91886 }, { "name": "video_fisheye_cam2", "l2": 5.870203, "mean_abs": 0.123377, "relative": 0.937091 }, { "name": "video_fisheye_cam3", "l2": 5.523995, "mean_abs": 0.113514, "relative": 0.924926 }, { "name": "video_stereo_left", "l2": 5.582584, "mean_abs": 0.12995, "relative": 0.907181 }, { "name": "video_stereo_right", "l2": 6.64184, "mean_abs": 0.156859, "relative": 0.93636 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.96852, "mean_abs": 0.078113, "relative": 0.704504 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9997264742851257" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9983356595039368" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999276399612427" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 929, "start_frame": 4645, "end_frame": 4664, "center_frame": 4654, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.609571, "mean_abs": 0.251978, "relative": 0.894963 }, { "name": "hand_right_joints", "l2": 10.791257, "mean_abs": 0.271996, "relative": 0.85064 }, { "name": "body_joints", "l2": 4.861114, "mean_abs": 0.070748, "relative": 0.656587 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.071684, "mean_abs": 0.010837, "relative": 0.169164 }, { "name": "camera_rotation_matrix", "l2": 3.048984, "mean_abs": 0.246821, "relative": 0.936294 }, { "name": "imu_accel_gyro", "l2": 17.910686, "mean_abs": 1.266328, "relative": 0.966343 }, { "name": "depth_confidence", "l2": 30131.119141, "mean_abs": 54.387814, "relative": 1.0 }, { "name": "video_fisheye_cam0", "l2": 5.208192, "mean_abs": 0.113154, "relative": 0.943092 }, { "name": "video_fisheye_cam1", "l2": 5.640358, "mean_abs": 0.126812, "relative": 0.932022 }, { "name": "video_fisheye_cam2", "l2": 5.952249, "mean_abs": 0.12821, "relative": 0.950188 }, { "name": "video_fisheye_cam3", "l2": 5.573204, "mean_abs": 0.117734, "relative": 0.933166 }, { "name": "video_stereo_left", "l2": 5.782244, "mean_abs": 0.137834, "relative": 0.939626 }, { "name": "video_stereo_right", "l2": 6.897186, "mean_abs": 0.168066, "relative": 0.972358 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.076056, "mean_abs": 0.082693, "relative": 0.742989 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "1.0" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Move bottle to coffee equipment", "correct": 0, "confidence": "1.0" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "1.0" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|digital scale|glass carafe|coffee scale|white mug|coffee jar|lid|mug|white cup|metal pitcher|carafe|stainless steel milk pitcher", "correct": 0, "confidence": "" } } }, { "window_index": 930, "start_frame": 4650, "end_frame": 4669, "center_frame": 4659, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.624292, "mean_abs": 0.254694, "relative": 0.896204 }, { "name": "hand_right_joints", "l2": 10.754062, "mean_abs": 0.27151, "relative": 0.847708 }, { "name": "body_joints", "l2": 4.815681, "mean_abs": 0.069918, "relative": 0.65045 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.080316, "mean_abs": 0.012683, "relative": 0.189535 }, { "name": "camera_rotation_matrix", "l2": 3.040599, "mean_abs": 0.246528, "relative": 0.933719 }, { "name": "imu_accel_gyro", "l2": 17.65593, "mean_abs": 1.263363, "relative": 0.952598 }, { "name": "depth_confidence", "l2": 22269.148438, "mean_abs": 34.060272, "relative": 0.739075 }, { "name": "video_fisheye_cam0", "l2": 5.207621, "mean_abs": 0.113056, "relative": 0.942989 }, { "name": "video_fisheye_cam1", "l2": 5.623857, "mean_abs": 0.126819, "relative": 0.929295 }, { "name": "video_fisheye_cam2", "l2": 5.931718, "mean_abs": 0.127841, "relative": 0.946911 }, { "name": "video_fisheye_cam3", "l2": 5.565721, "mean_abs": 0.117755, "relative": 0.931913 }, { "name": "video_stereo_left", "l2": 5.76222, "mean_abs": 0.137526, "relative": 0.936372 }, { "name": "video_stereo_right", "l2": 6.904555, "mean_abs": 0.167507, "relative": 0.973397 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.036471, "mean_abs": 0.078181, "relative": 0.728822 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Pour coffee", "correct": 0, "confidence": "1.0" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Prepare for pouring", "correct": 1, "confidence": "1.0" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Pour coffee", "correct": 0, "confidence": "1.0" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|dripper|coffee filter|glass carafe|coffee scale|wooden scoop|coffee jar|coffee scoop|coffee container|lid|water bottle|mug|small bottle|white coffee cup|metal pitcher|coffee cup|stainless steel milk pitcher", "correct": 0, "confidence": "" } } }, { "window_index": 931, "start_frame": 4655, "end_frame": 4674, "center_frame": 4664, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.631986, "mean_abs": 0.251628, "relative": 0.896853 }, { "name": "hand_right_joints", "l2": 10.712507, "mean_abs": 0.265968, "relative": 0.844432 }, { "name": "body_joints", "l2": 4.78073, "mean_abs": 0.068273, "relative": 0.645729 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.083036, "mean_abs": 0.012755, "relative": 0.195954 }, { "name": "camera_rotation_matrix", "l2": 3.028113, "mean_abs": 0.241778, "relative": 0.929884 }, { "name": "imu_accel_gyro", "l2": 17.391222, "mean_abs": 1.237687, "relative": 0.938316 }, { "name": "depth_confidence", "l2": 22269.146484, "mean_abs": 34.043224, "relative": 0.739075 }, { "name": "video_fisheye_cam0", "l2": 5.191463, "mean_abs": 0.111557, "relative": 0.940063 }, { "name": "video_fisheye_cam1", "l2": 5.587579, "mean_abs": 0.124731, "relative": 0.9233 }, { "name": "video_fisheye_cam2", "l2": 5.922142, "mean_abs": 0.12602, "relative": 0.945382 }, { "name": "video_fisheye_cam3", "l2": 5.530183, "mean_abs": 0.115743, "relative": 0.925963 }, { "name": "video_stereo_left", "l2": 5.718397, "mean_abs": 0.135171, "relative": 0.929251 }, { "name": "video_stereo_right", "l2": 6.87796, "mean_abs": 0.164426, "relative": 0.969648 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.108909, "mean_abs": 0.083235, "relative": 0.754747 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Pour coffee", "correct": 0, "confidence": "1.0" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Prepare for pouring", "correct": 1, "confidence": "1.0" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Pour coffee", "correct": 0, "confidence": "1.0" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|dripper|coffee filter|glass carafe|coffee scale|wooden scoop|coffee jar|coffee scoop|coffee container|lid|water bottle|mug|small bottle|white coffee cup|metal pitcher|coffee cup|stainless steel milk pitcher", "correct": 0, "confidence": "" } } }, { "window_index": 932, "start_frame": 4660, "end_frame": 4679, "center_frame": 4669, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.618808, "mean_abs": 0.253308, "relative": 0.895742 }, { "name": "hand_right_joints", "l2": 10.672758, "mean_abs": 0.267053, "relative": 0.841299 }, { "name": "body_joints", "l2": 4.804663, "mean_abs": 0.068556, "relative": 0.648962 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.048215, "mean_abs": 0.006714, "relative": 0.113779 }, { "name": "camera_rotation_matrix", "l2": 3.013895, "mean_abs": 0.233593, "relative": 0.925518 }, { "name": "imu_accel_gyro", "l2": 17.417725, "mean_abs": 1.245015, "relative": 0.939746 }, { "name": "depth_confidence", "l2": 22269.146484, "mean_abs": 34.041168, "relative": 0.739075 }, { "name": "video_fisheye_cam0", "l2": 5.17084, "mean_abs": 0.109789, "relative": 0.936328 }, { "name": "video_fisheye_cam1", "l2": 5.52876, "mean_abs": 0.121539, "relative": 0.913581 }, { "name": "video_fisheye_cam2", "l2": 5.832151, "mean_abs": 0.122681, "relative": 0.931016 }, { "name": "video_fisheye_cam3", "l2": 5.493688, "mean_abs": 0.112139, "relative": 0.919852 }, { "name": "video_stereo_left", "l2": 5.542769, "mean_abs": 0.129561, "relative": 0.900711 }, { "name": "video_stereo_right", "l2": 6.742698, "mean_abs": 0.158638, "relative": 0.950579 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.203946, "mean_abs": 0.086528, "relative": 0.788759 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Pour coffee", "correct": 0, "confidence": "1.0" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Prepare for pouring", "correct": 1, "confidence": "1.0" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Pour coffee", "correct": 0, "confidence": "1.0" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "kettle|coffee dripper|scale|bottle|gooseneck kettle|dripper|coffee filter|glass carafe|coffee scale|wooden scoop|coffee jar|coffee scoop|coffee container|lid|water bottle|mug|small bottle|white coffee cup|metal pitcher|coffee cup|stainless steel milk pitcher", "correct": 0, "confidence": "" } } }, { "window_index": 933, "start_frame": 4665, "end_frame": 4684, "center_frame": 4674, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.611782, "mean_abs": 0.255657, "relative": 0.895149 }, { "name": "hand_right_joints", "l2": 10.700247, "mean_abs": 0.270421, "relative": 0.843466 }, { "name": "body_joints", "l2": 4.905135, "mean_abs": 0.073052, "relative": 0.662533 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.043476, "mean_abs": 0.006132, "relative": 0.102598 }, { "name": "camera_rotation_matrix", "l2": 3.033458, "mean_abs": 0.241073, "relative": 0.931526 }, { "name": "imu_accel_gyro", "l2": 17.430471, "mean_abs": 1.206665, "relative": 0.940434 }, { "name": "depth_confidence", "l2": 1166.266724, "mean_abs": 2.379988, "relative": 0.038706 }, { "name": "video_fisheye_cam0", "l2": 5.178192, "mean_abs": 0.111674, "relative": 0.93766 }, { "name": "video_fisheye_cam1", "l2": 5.555505, "mean_abs": 0.124175, "relative": 0.918 }, { "name": "video_fisheye_cam2", "l2": 5.723403, "mean_abs": 0.122537, "relative": 0.913656 }, { "name": "video_fisheye_cam3", "l2": 5.537379, "mean_abs": 0.11516, "relative": 0.927167 }, { "name": "video_stereo_left", "l2": 5.536708, "mean_abs": 0.131773, "relative": 0.899726 }, { "name": "video_stereo_right", "l2": 6.760143, "mean_abs": 0.162766, "relative": 0.953038 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.041977, "mean_abs": 0.080255, "relative": 0.730793 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle away", "correct": 0, "confidence": "0.9999998807907104" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.5474713444709778" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Move kettle away", "correct": 0, "confidence": "0.9095953106880188" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "gooseneck kettle|table|dripper|coffee filter|glass carafe|wooden scoop|coffee jar|coffee scoop|coffee container|water bottle|coffee mug|white bottle|coffee equipment|small bottle|weighing scale|coffee cup|stainless steel milk pitcher", "correct": 0, "confidence": "" } } }, { "window_index": 934, "start_frame": 4670, "end_frame": 4689, "center_frame": 4679, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.577395, "mean_abs": 0.252172, "relative": 0.892248 }, { "name": "hand_right_joints", "l2": 10.701952, "mean_abs": 0.269494, "relative": 0.8436 }, { "name": "body_joints", "l2": 5.063248, "mean_abs": 0.077549, "relative": 0.683889 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.066474, "mean_abs": 0.009443, "relative": 0.15687 }, { "name": "camera_rotation_matrix", "l2": 3.036112, "mean_abs": 0.240325, "relative": 0.932341 }, { "name": "imu_accel_gyro", "l2": 17.428358, "mean_abs": 1.168297, "relative": 0.94032 }, { "name": "depth_confidence", "l2": 13.273207, "mean_abs": 0.236673, "relative": 0.000441 }, { "name": "video_fisheye_cam0", "l2": 5.181003, "mean_abs": 0.11177, "relative": 0.938169 }, { "name": "video_fisheye_cam1", "l2": 5.563927, "mean_abs": 0.123835, "relative": 0.919392 }, { "name": "video_fisheye_cam2", "l2": 5.664948, "mean_abs": 0.121459, "relative": 0.904325 }, { "name": "video_fisheye_cam3", "l2": 5.559733, "mean_abs": 0.115905, "relative": 0.93091 }, { "name": "video_stereo_left", "l2": 5.470314, "mean_abs": 0.129482, "relative": 0.888937 }, { "name": "video_stereo_right", "l2": 6.681062, "mean_abs": 0.160136, "relative": 0.94189 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.088804, "mean_abs": 0.085759, "relative": 0.747552 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.6351548433303833" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.8306093811988831" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.3670908808708191" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|gooseneck kettle|digital scale|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 935, "start_frame": 4675, "end_frame": 4694, "center_frame": 4684, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.485737, "mean_abs": 0.244153, "relative": 0.884517 }, { "name": "hand_right_joints", "l2": 10.70561, "mean_abs": 0.262927, "relative": 0.843888 }, { "name": "body_joints", "l2": 5.204028, "mean_abs": 0.080562, "relative": 0.702904 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.102802, "mean_abs": 0.013782, "relative": 0.242599 }, { "name": "camera_rotation_matrix", "l2": 3.025578, "mean_abs": 0.234244, "relative": 0.929106 }, { "name": "imu_accel_gyro", "l2": 17.308027, "mean_abs": 1.117351, "relative": 0.933828 }, { "name": "depth_confidence", "l2": 12.565044, "mean_abs": 0.228604, "relative": 0.000417 }, { "name": "video_fisheye_cam0", "l2": 5.163232, "mean_abs": 0.110233, "relative": 0.934951 }, { "name": "video_fisheye_cam1", "l2": 5.56443, "mean_abs": 0.121939, "relative": 0.919475 }, { "name": "video_fisheye_cam2", "l2": 5.597714, "mean_abs": 0.118826, "relative": 0.893592 }, { "name": "video_fisheye_cam3", "l2": 5.580051, "mean_abs": 0.115345, "relative": 0.934312 }, { "name": "video_stereo_left", "l2": 5.355145, "mean_abs": 0.125255, "relative": 0.870222 }, { "name": "video_stereo_right", "l2": 6.541957, "mean_abs": 0.154153, "relative": 0.922279 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.266536, "mean_abs": 0.092813, "relative": 0.811159 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9386245608329773" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9225940704345703" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.788848876953125" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 936, "start_frame": 4680, "end_frame": 4699, "center_frame": 4689, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.354381, "mean_abs": 0.236818, "relative": 0.873436 }, { "name": "hand_right_joints", "l2": 10.661678, "mean_abs": 0.26268, "relative": 0.840425 }, { "name": "body_joints", "l2": 5.360606, "mean_abs": 0.083557, "relative": 0.724053 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.140786, "mean_abs": 0.017943, "relative": 0.332236 }, { "name": "camera_rotation_matrix", "l2": 3.005661, "mean_abs": 0.223477, "relative": 0.92299 }, { "name": "imu_accel_gyro", "l2": 17.080177, "mean_abs": 1.037018, "relative": 0.921534 }, { "name": "depth_confidence", "l2": 12.501756, "mean_abs": 0.222694, "relative": 0.000415 }, { "name": "video_fisheye_cam0", "l2": 5.122488, "mean_abs": 0.107612, "relative": 0.927573 }, { "name": "video_fisheye_cam1", "l2": 5.563198, "mean_abs": 0.119331, "relative": 0.919271 }, { "name": "video_fisheye_cam2", "l2": 5.520399, "mean_abs": 0.114803, "relative": 0.88125 }, { "name": "video_fisheye_cam3", "l2": 5.593537, "mean_abs": 0.113865, "relative": 0.93657 }, { "name": "video_stereo_left", "l2": 5.181177, "mean_abs": 0.116723, "relative": 0.841951 }, { "name": "video_stereo_right", "l2": 6.388437, "mean_abs": 0.145454, "relative": 0.900636 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.145775, "mean_abs": 0.089014, "relative": 0.767941 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9712947010993958" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9069050550460815" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999998807907104" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9436051249504089" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 937, "start_frame": 4685, "end_frame": 4704, "center_frame": 4694, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.295162, "mean_abs": 0.230819, "relative": 0.868441 }, { "name": "hand_right_joints", "l2": 10.604074, "mean_abs": 0.259785, "relative": 0.835885 }, { "name": "body_joints", "l2": 5.446162, "mean_abs": 0.084483, "relative": 0.735609 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.170255, "mean_abs": 0.021864, "relative": 0.401779 }, { "name": "camera_rotation_matrix", "l2": 3.002764, "mean_abs": 0.219749, "relative": 0.9221 }, { "name": "imu_accel_gyro", "l2": 17.439226, "mean_abs": 1.008209, "relative": 0.940906 }, { "name": "depth_confidence", "l2": 12.869873, "mean_abs": 0.223547, "relative": 0.000427 }, { "name": "video_fisheye_cam0", "l2": 5.119193, "mean_abs": 0.107053, "relative": 0.926976 }, { "name": "video_fisheye_cam1", "l2": 5.620434, "mean_abs": 0.120552, "relative": 0.928729 }, { "name": "video_fisheye_cam2", "l2": 5.503836, "mean_abs": 0.11364, "relative": 0.878606 }, { "name": "video_fisheye_cam3", "l2": 5.602065, "mean_abs": 0.114704, "relative": 0.937998 }, { "name": "video_stereo_left", "l2": 5.156582, "mean_abs": 0.114407, "relative": 0.837955 }, { "name": "video_stereo_right", "l2": 6.372477, "mean_abs": 0.143201, "relative": 0.898385 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.1828, "mean_abs": 0.089076, "relative": 0.781191 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9704196453094482" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9687022566795349" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9802709817886353" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 938, "start_frame": 4690, "end_frame": 4709, "center_frame": 4699, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.224007, "mean_abs": 0.227057, "relative": 0.862439 }, { "name": "hand_right_joints", "l2": 10.558454, "mean_abs": 0.256456, "relative": 0.832289 }, { "name": "body_joints", "l2": 5.539577, "mean_abs": 0.085681, "relative": 0.748226 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.164039, "mean_abs": 0.021498, "relative": 0.38711 }, { "name": "camera_rotation_matrix", "l2": 3.001817, "mean_abs": 0.220773, "relative": 0.921809 }, { "name": "imu_accel_gyro", "l2": 17.372709, "mean_abs": 1.018123, "relative": 0.937317 }, { "name": "depth_confidence", "l2": 12.897431, "mean_abs": 0.225487, "relative": 0.000428 }, { "name": "video_fisheye_cam0", "l2": 5.130044, "mean_abs": 0.107399, "relative": 0.928941 }, { "name": "video_fisheye_cam1", "l2": 5.686253, "mean_abs": 0.12258, "relative": 0.939605 }, { "name": "video_fisheye_cam2", "l2": 5.511259, "mean_abs": 0.113849, "relative": 0.87979 }, { "name": "video_fisheye_cam3", "l2": 5.623611, "mean_abs": 0.11534, "relative": 0.941606 }, { "name": "video_stereo_left", "l2": 5.166259, "mean_abs": 0.114691, "relative": 0.839527 }, { "name": "video_stereo_right", "l2": 6.396315, "mean_abs": 0.143746, "relative": 0.901746 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.295532, "mean_abs": 0.097117, "relative": 0.821537 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9753908514976501" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9868647456169128" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.994113028049469" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 939, "start_frame": 4695, "end_frame": 4714, "center_frame": 4704, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.144068, "mean_abs": 0.218296, "relative": 0.855695 }, { "name": "hand_right_joints", "l2": 10.525909, "mean_abs": 0.251933, "relative": 0.829723 }, { "name": "body_joints", "l2": 5.620268, "mean_abs": 0.085743, "relative": 0.759125 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.107399, "mean_abs": 0.014711, "relative": 0.253446 }, { "name": "camera_rotation_matrix", "l2": 3.000495, "mean_abs": 0.220033, "relative": 0.921403 }, { "name": "imu_accel_gyro", "l2": 17.316833, "mean_abs": 0.993235, "relative": 0.934303 }, { "name": "depth_confidence", "l2": 12.893333, "mean_abs": 0.226065, "relative": 0.000428 }, { "name": "video_fisheye_cam0", "l2": 5.131163, "mean_abs": 0.107228, "relative": 0.929144 }, { "name": "video_fisheye_cam1", "l2": 5.702202, "mean_abs": 0.122112, "relative": 0.942241 }, { "name": "video_fisheye_cam2", "l2": 5.510907, "mean_abs": 0.113728, "relative": 0.879734 }, { "name": "video_fisheye_cam3", "l2": 5.62606, "mean_abs": 0.114602, "relative": 0.942016 }, { "name": "video_stereo_left", "l2": 5.167383, "mean_abs": 0.114416, "relative": 0.83971 }, { "name": "video_stereo_right", "l2": 6.37373, "mean_abs": 0.143102, "relative": 0.898562 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.201341, "mean_abs": 0.091848, "relative": 0.787827 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9717796444892883" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.991688072681427" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999997615814209" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9960814118385315" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 940, "start_frame": 4700, "end_frame": 4719, "center_frame": 4709, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.125946, "mean_abs": 0.213174, "relative": 0.854167 }, { "name": "hand_right_joints", "l2": 10.532995, "mean_abs": 0.253575, "relative": 0.830282 }, { "name": "body_joints", "l2": 5.633894, "mean_abs": 0.085818, "relative": 0.760965 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.046932, "mean_abs": 0.00593, "relative": 0.110752 }, { "name": "camera_rotation_matrix", "l2": 3.003374, "mean_abs": 0.220382, "relative": 0.922287 }, { "name": "imu_accel_gyro", "l2": 17.287766, "mean_abs": 0.974442, "relative": 0.932734 }, { "name": "depth_confidence", "l2": 12.870131, "mean_abs": 0.225669, "relative": 0.000427 }, { "name": "video_fisheye_cam0", "l2": 5.134562, "mean_abs": 0.10752, "relative": 0.929759 }, { "name": "video_fisheye_cam1", "l2": 5.707196, "mean_abs": 0.122119, "relative": 0.943066 }, { "name": "video_fisheye_cam2", "l2": 5.510443, "mean_abs": 0.11433, "relative": 0.87966 }, { "name": "video_fisheye_cam3", "l2": 5.623145, "mean_abs": 0.115194, "relative": 0.941528 }, { "name": "video_stereo_left", "l2": 5.164282, "mean_abs": 0.115151, "relative": 0.839206 }, { "name": "video_stereo_right", "l2": 6.3676, "mean_abs": 0.14385, "relative": 0.897698 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.176015, "mean_abs": 0.08797, "relative": 0.778763 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9616092443466187" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.978553295135498" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999988079071045" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9861900806427002" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 941, "start_frame": 4705, "end_frame": 4724, "center_frame": 4714, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.121667, "mean_abs": 0.217397, "relative": 0.853806 }, { "name": "hand_right_joints", "l2": 10.547919, "mean_abs": 0.255227, "relative": 0.831458 }, { "name": "body_joints", "l2": 5.648262, "mean_abs": 0.085981, "relative": 0.762906 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.070161, "mean_abs": 0.009663, "relative": 0.165569 }, { "name": "camera_rotation_matrix", "l2": 3.004977, "mean_abs": 0.219285, "relative": 0.92278 }, { "name": "imu_accel_gyro", "l2": 17.208984, "mean_abs": 0.964942, "relative": 0.928484 }, { "name": "depth_confidence", "l2": 12.517374, "mean_abs": 0.224187, "relative": 0.000415 }, { "name": "video_fisheye_cam0", "l2": 5.135596, "mean_abs": 0.10767, "relative": 0.929947 }, { "name": "video_fisheye_cam1", "l2": 5.717722, "mean_abs": 0.122436, "relative": 0.944805 }, { "name": "video_fisheye_cam2", "l2": 5.511565, "mean_abs": 0.114161, "relative": 0.879839 }, { "name": "video_fisheye_cam3", "l2": 5.648005, "mean_abs": 0.11554, "relative": 0.94569 }, { "name": "video_stereo_left", "l2": 5.177549, "mean_abs": 0.114876, "relative": 0.841362 }, { "name": "video_stereo_right", "l2": 6.367452, "mean_abs": 0.142875, "relative": 0.897677 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.060126, "mean_abs": 0.080902, "relative": 0.737288 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.829465389251709" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.7167705297470093" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.999910831451416" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.6213365793228149" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 942, "start_frame": 4710, "end_frame": 4729, "center_frame": 4719, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.116249, "mean_abs": 0.219366, "relative": 0.853349 }, { "name": "hand_right_joints", "l2": 10.56541, "mean_abs": 0.25496, "relative": 0.832837 }, { "name": "body_joints", "l2": 5.662063, "mean_abs": 0.08579, "relative": 0.76477 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.080594, "mean_abs": 0.010688, "relative": 0.190192 }, { "name": "camera_rotation_matrix", "l2": 3.001068, "mean_abs": 0.219499, "relative": 0.921579 }, { "name": "imu_accel_gyro", "l2": 17.422789, "mean_abs": 0.974183, "relative": 0.940019 }, { "name": "depth_confidence", "l2": 12.475896, "mean_abs": 0.223661, "relative": 0.000414 }, { "name": "video_fisheye_cam0", "l2": 5.122578, "mean_abs": 0.107376, "relative": 0.927589 }, { "name": "video_fisheye_cam1", "l2": 5.689135, "mean_abs": 0.121564, "relative": 0.940082 }, { "name": "video_fisheye_cam2", "l2": 5.507875, "mean_abs": 0.113637, "relative": 0.87925 }, { "name": "video_fisheye_cam3", "l2": 5.621467, "mean_abs": 0.114834, "relative": 0.941247 }, { "name": "video_stereo_left", "l2": 5.167965, "mean_abs": 0.113868, "relative": 0.839805 }, { "name": "video_stereo_right", "l2": 6.336775, "mean_abs": 0.14105, "relative": 0.893352 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.091319, "mean_abs": 0.084537, "relative": 0.748452 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8503955602645874" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.7395211458206177" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999101161956787" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.7265777587890625" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|coffee mug|mug", "correct": 0, "confidence": "" } } }, { "window_index": 943, "start_frame": 4715, "end_frame": 4734, "center_frame": 4724, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.139065, "mean_abs": 0.219263, "relative": 0.855273 }, { "name": "hand_right_joints", "l2": 10.600145, "mean_abs": 0.256147, "relative": 0.835575 }, { "name": "body_joints", "l2": 5.647565, "mean_abs": 0.084766, "relative": 0.762812 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.049324, "mean_abs": 0.006764, "relative": 0.116397 }, { "name": "camera_rotation_matrix", "l2": 3.000993, "mean_abs": 0.218333, "relative": 0.921556 }, { "name": "imu_accel_gyro", "l2": 17.122515, "mean_abs": 0.923416, "relative": 0.923819 }, { "name": "depth_confidence", "l2": 12.391349, "mean_abs": 0.222046, "relative": 0.000411 }, { "name": "video_fisheye_cam0", "l2": 5.116785, "mean_abs": 0.106917, "relative": 0.92654 }, { "name": "video_fisheye_cam1", "l2": 5.650119, "mean_abs": 0.119565, "relative": 0.933635 }, { "name": "video_fisheye_cam2", "l2": 5.510966, "mean_abs": 0.113256, "relative": 0.879744 }, { "name": "video_fisheye_cam3", "l2": 5.59534, "mean_abs": 0.113228, "relative": 0.936872 }, { "name": "video_stereo_left", "l2": 5.175174, "mean_abs": 0.113093, "relative": 0.840976 }, { "name": "video_stereo_right", "l2": 6.328218, "mean_abs": 0.139911, "relative": 0.892146 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.176257, "mean_abs": 0.08979, "relative": 0.77885 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9389151930809021" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.8253526091575623" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9996911287307739" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8810917139053345" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 944, "start_frame": 4720, "end_frame": 4739, "center_frame": 4729, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.153924, "mean_abs": 0.218694, "relative": 0.856527 }, { "name": "hand_right_joints", "l2": 10.647171, "mean_abs": 0.256689, "relative": 0.839282 }, { "name": "body_joints", "l2": 5.595102, "mean_abs": 0.084625, "relative": 0.755726 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013853, "mean_abs": 0.002029, "relative": 0.032692 }, { "name": "camera_rotation_matrix", "l2": 3.000626, "mean_abs": 0.215728, "relative": 0.921444 }, { "name": "imu_accel_gyro", "l2": 16.980564, "mean_abs": 0.934606, "relative": 0.91616 }, { "name": "depth_confidence", "l2": 12.389565, "mean_abs": 0.220922, "relative": 0.000411 }, { "name": "video_fisheye_cam0", "l2": 5.107687, "mean_abs": 0.106192, "relative": 0.924893 }, { "name": "video_fisheye_cam1", "l2": 5.609146, "mean_abs": 0.117814, "relative": 0.926864 }, { "name": "video_fisheye_cam2", "l2": 5.51405, "mean_abs": 0.112798, "relative": 0.880236 }, { "name": "video_fisheye_cam3", "l2": 5.561036, "mean_abs": 0.112148, "relative": 0.931128 }, { "name": "video_stereo_left", "l2": 5.170947, "mean_abs": 0.111977, "relative": 0.840289 }, { "name": "video_stereo_right", "l2": 6.316187, "mean_abs": 0.138561, "relative": 0.89045 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.155176, "mean_abs": 0.089775, "relative": 0.771305 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9803535342216492" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9863730669021606" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9936291575431824" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9689993262290955" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 945, "start_frame": 4725, "end_frame": 4744, "center_frame": 4734, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.144561, "mean_abs": 0.223788, "relative": 0.855737 }, { "name": "hand_right_joints", "l2": 10.663548, "mean_abs": 0.263123, "relative": 0.840573 }, { "name": "body_joints", "l2": 5.543159, "mean_abs": 0.084704, "relative": 0.74871 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030866, "mean_abs": 0.003998, "relative": 0.072839 }, { "name": "camera_rotation_matrix", "l2": 3.011122, "mean_abs": 0.229825, "relative": 0.924667 }, { "name": "imu_accel_gyro", "l2": 17.036936, "mean_abs": 1.011456, "relative": 0.919201 }, { "name": "depth_confidence", "l2": 12.749959, "mean_abs": 0.232574, "relative": 0.000423 }, { "name": "video_fisheye_cam0", "l2": 5.108267, "mean_abs": 0.107955, "relative": 0.924998 }, { "name": "video_fisheye_cam1", "l2": 5.717821, "mean_abs": 0.124604, "relative": 0.944822 }, { "name": "video_fisheye_cam2", "l2": 5.517598, "mean_abs": 0.115615, "relative": 0.880802 }, { "name": "video_fisheye_cam3", "l2": 5.57605, "mean_abs": 0.114926, "relative": 0.933642 }, { "name": "video_stereo_left", "l2": 5.1816, "mean_abs": 0.118777, "relative": 0.84202 }, { "name": "video_stereo_right", "l2": 6.314088, "mean_abs": 0.145845, "relative": 0.890154 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.315216, "mean_abs": 0.095419, "relative": 0.828581 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999958276748657" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9999899864196777" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999991655349731" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 946, "start_frame": 4730, "end_frame": 4749, "center_frame": 4739, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.131569, "mean_abs": 0.227395, "relative": 0.854641 }, { "name": "hand_right_joints", "l2": 10.673145, "mean_abs": 0.264687, "relative": 0.841329 }, { "name": "body_joints", "l2": 5.510329, "mean_abs": 0.084195, "relative": 0.744276 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.048886, "mean_abs": 0.005966, "relative": 0.115363 }, { "name": "camera_rotation_matrix", "l2": 3.019655, "mean_abs": 0.237021, "relative": 0.927287 }, { "name": "imu_accel_gyro", "l2": 17.236958, "mean_abs": 1.002118, "relative": 0.929993 }, { "name": "depth_confidence", "l2": 13.145885, "mean_abs": 0.23922, "relative": 0.000436 }, { "name": "video_fisheye_cam0", "l2": 5.109189, "mean_abs": 0.108732, "relative": 0.925165 }, { "name": "video_fisheye_cam1", "l2": 5.819896, "mean_abs": 0.128937, "relative": 0.961689 }, { "name": "video_fisheye_cam2", "l2": 5.529384, "mean_abs": 0.116933, "relative": 0.882684 }, { "name": "video_fisheye_cam3", "l2": 5.596538, "mean_abs": 0.116306, "relative": 0.937073 }, { "name": "video_stereo_left", "l2": 5.196944, "mean_abs": 0.12143, "relative": 0.844514 }, { "name": "video_stereo_right", "l2": 6.311211, "mean_abs": 0.149474, "relative": 0.889748 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.182932, "mean_abs": 0.086809, "relative": 0.781238 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.999997615814209" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9999923706054688" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999984502792358" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|coffee mug|mug", "correct": 0, "confidence": "" } } }, { "window_index": 947, "start_frame": 4735, "end_frame": 4754, "center_frame": 4744, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.095942, "mean_abs": 0.224778, "relative": 0.851636 }, { "name": "hand_right_joints", "l2": 10.659452, "mean_abs": 0.262903, "relative": 0.84025 }, { "name": "body_joints", "l2": 5.453139, "mean_abs": 0.08269, "relative": 0.736551 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.054943, "mean_abs": 0.007041, "relative": 0.129658 }, { "name": "camera_rotation_matrix", "l2": 3.019352, "mean_abs": 0.237532, "relative": 0.927194 }, { "name": "imu_accel_gyro", "l2": 17.217033, "mean_abs": 0.996213, "relative": 0.928918 }, { "name": "depth_confidence", "l2": 13.141758, "mean_abs": 0.240191, "relative": 0.000436 }, { "name": "video_fisheye_cam0", "l2": 5.106705, "mean_abs": 0.108683, "relative": 0.924715 }, { "name": "video_fisheye_cam1", "l2": 5.841369, "mean_abs": 0.12934, "relative": 0.965237 }, { "name": "video_fisheye_cam2", "l2": 5.528237, "mean_abs": 0.117086, "relative": 0.882501 }, { "name": "video_fisheye_cam3", "l2": 5.598129, "mean_abs": 0.116381, "relative": 0.937339 }, { "name": "video_stereo_left", "l2": 5.187426, "mean_abs": 0.12137, "relative": 0.842967 }, { "name": "video_stereo_right", "l2": 6.278505, "mean_abs": 0.149238, "relative": 0.885137 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.224606, "mean_abs": 0.086558, "relative": 0.796153 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999982118606567" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9999927282333374" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999986886978149" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|coffee mug|mug", "correct": 0, "confidence": "" } } }, { "window_index": 948, "start_frame": 4740, "end_frame": 4759, "center_frame": 4749, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.043444, "mean_abs": 0.218177, "relative": 0.847207 }, { "name": "hand_right_joints", "l2": 10.640785, "mean_abs": 0.258789, "relative": 0.838779 }, { "name": "body_joints", "l2": 5.394765, "mean_abs": 0.080678, "relative": 0.728667 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.052206, "mean_abs": 0.007057, "relative": 0.123199 }, { "name": "camera_rotation_matrix", "l2": 3.013475, "mean_abs": 0.233825, "relative": 0.925389 }, { "name": "imu_accel_gyro", "l2": 17.318262, "mean_abs": 1.01121, "relative": 0.93438 }, { "name": "depth_confidence", "l2": 13.541567, "mean_abs": 0.242555, "relative": 0.000449 }, { "name": "video_fisheye_cam0", "l2": 5.093881, "mean_abs": 0.107573, "relative": 0.922393 }, { "name": "video_fisheye_cam1", "l2": 5.855549, "mean_abs": 0.128355, "relative": 0.96758 }, { "name": "video_fisheye_cam2", "l2": 5.517542, "mean_abs": 0.115925, "relative": 0.880793 }, { "name": "video_fisheye_cam3", "l2": 5.589797, "mean_abs": 0.11547, "relative": 0.935944 }, { "name": "video_stereo_left", "l2": 5.133292, "mean_abs": 0.118988, "relative": 0.83417 }, { "name": "video_stereo_right", "l2": 6.195363, "mean_abs": 0.145683, "relative": 0.873416 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.101012, "mean_abs": 0.081942, "relative": 0.751921 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012181, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999998807907104" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9999980926513672" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999998807907104" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|coffee mug|mug", "correct": 0, "confidence": "" } } }, { "window_index": 949, "start_frame": 4745, "end_frame": 4764, "center_frame": 4754, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.970454, "mean_abs": 0.205732, "relative": 0.84105 }, { "name": "hand_right_joints", "l2": 10.605894, "mean_abs": 0.249374, "relative": 0.836028 }, { "name": "body_joints", "l2": 5.352168, "mean_abs": 0.078983, "relative": 0.722913 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021595, "mean_abs": 0.003342, "relative": 0.050961 }, { "name": "camera_rotation_matrix", "l2": 3.000194, "mean_abs": 0.220859, "relative": 0.921311 }, { "name": "imu_accel_gyro", "l2": 17.228168, "mean_abs": 0.857446, "relative": 0.929519 }, { "name": "depth_confidence", "l2": 13.782761, "mean_abs": 0.238071, "relative": 0.000457 }, { "name": "video_fisheye_cam0", "l2": 5.072964, "mean_abs": 0.104579, "relative": 0.918605 }, { "name": "video_fisheye_cam1", "l2": 5.903501, "mean_abs": 0.125148, "relative": 0.975504 }, { "name": "video_fisheye_cam2", "l2": 5.490903, "mean_abs": 0.112433, "relative": 0.876541 }, { "name": "video_fisheye_cam3", "l2": 5.588157, "mean_abs": 0.112565, "relative": 0.935669 }, { "name": "video_stereo_left", "l2": 5.01967, "mean_abs": 0.110835, "relative": 0.815706 }, { "name": "video_stereo_right", "l2": 6.031975, "mean_abs": 0.135279, "relative": 0.850382 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.22263, "mean_abs": 0.086313, "relative": 0.795446 }, { "name": "caption_objects_interaction_text", "l2": 1.91718, "mean_abs": 0.018477, "relative": 0.880323 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999994039535522" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9999852180480957" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9973185658454895" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.999998927116394" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 950, "start_frame": 4750, "end_frame": 4769, "center_frame": 4759, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.991075, "mean_abs": 0.208156, "relative": 0.84279 }, { "name": "hand_right_joints", "l2": 10.615807, "mean_abs": 0.249131, "relative": 0.83681 }, { "name": "body_joints", "l2": 5.350893, "mean_abs": 0.078872, "relative": 0.722741 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012153, "mean_abs": 0.001522, "relative": 0.02868 }, { "name": "camera_rotation_matrix", "l2": 2.99994, "mean_abs": 0.221316, "relative": 0.921233 }, { "name": "imu_accel_gyro", "l2": 17.101252, "mean_abs": 0.799934, "relative": 0.922671 }, { "name": "depth_confidence", "l2": 13.711551, "mean_abs": 0.236827, "relative": 0.000455 }, { "name": "video_fisheye_cam0", "l2": 5.077038, "mean_abs": 0.104979, "relative": 0.919343 }, { "name": "video_fisheye_cam1", "l2": 5.844198, "mean_abs": 0.124202, "relative": 0.965704 }, { "name": "video_fisheye_cam2", "l2": 5.514617, "mean_abs": 0.113272, "relative": 0.880327 }, { "name": "video_fisheye_cam3", "l2": 5.590029, "mean_abs": 0.112626, "relative": 0.935983 }, { "name": "video_stereo_left", "l2": 5.046871, "mean_abs": 0.11201, "relative": 0.820127 }, { "name": "video_stereo_right", "l2": 5.942829, "mean_abs": 0.133034, "relative": 0.837814 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.017057, "mean_abs": 0.080762, "relative": 0.721875 }, { "name": "caption_objects_interaction_text", "l2": 1.91718, "mean_abs": 0.018761, "relative": 0.880323 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999986886978149" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9999692440032959" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9270350933074951" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999982118606567" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 951, "start_frame": 4755, "end_frame": 4774, "center_frame": 4764, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.018545, "mean_abs": 0.208939, "relative": 0.845107 }, { "name": "hand_right_joints", "l2": 10.620955, "mean_abs": 0.249497, "relative": 0.837215 }, { "name": "body_joints", "l2": 5.341891, "mean_abs": 0.07839, "relative": 0.721525 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013907, "mean_abs": 0.001843, "relative": 0.032819 }, { "name": "camera_rotation_matrix", "l2": 3.000307, "mean_abs": 0.221687, "relative": 0.921346 }, { "name": "imu_accel_gyro", "l2": 17.099289, "mean_abs": 0.795826, "relative": 0.922566 }, { "name": "depth_confidence", "l2": 13.777342, "mean_abs": 0.237128, "relative": 0.000457 }, { "name": "video_fisheye_cam0", "l2": 5.077863, "mean_abs": 0.105044, "relative": 0.919492 }, { "name": "video_fisheye_cam1", "l2": 5.794818, "mean_abs": 0.123151, "relative": 0.957545 }, { "name": "video_fisheye_cam2", "l2": 5.520625, "mean_abs": 0.1135, "relative": 0.881286 }, { "name": "video_fisheye_cam3", "l2": 5.592882, "mean_abs": 0.112598, "relative": 0.936461 }, { "name": "video_stereo_left", "l2": 5.057069, "mean_abs": 0.112415, "relative": 0.821784 }, { "name": "video_stereo_right", "l2": 5.914867, "mean_abs": 0.132451, "relative": 0.833872 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.988715, "mean_abs": 0.079682, "relative": 0.711731 }, { "name": "caption_objects_interaction_text", "l2": 1.91718, "mean_abs": 0.018568, "relative": 0.880323 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999980926513672" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9999808073043823" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9621915221214294" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999980926513672" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 952, "start_frame": 4760, "end_frame": 4779, "center_frame": 4769, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.088216, "mean_abs": 0.214948, "relative": 0.850984 }, { "name": "hand_right_joints", "l2": 10.663124, "mean_abs": 0.253917, "relative": 0.840539 }, { "name": "body_joints", "l2": 5.358161, "mean_abs": 0.079255, "relative": 0.723723 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.030179, "mean_abs": 0.004029, "relative": 0.071219 }, { "name": "camera_rotation_matrix", "l2": 3.006812, "mean_abs": 0.228543, "relative": 0.923343 }, { "name": "imu_accel_gyro", "l2": 17.119545, "mean_abs": 0.895018, "relative": 0.923658 }, { "name": "depth_confidence", "l2": 13.824684, "mean_abs": 0.242212, "relative": 0.000459 }, { "name": "video_fisheye_cam0", "l2": 5.089902, "mean_abs": 0.106711, "relative": 0.921672 }, { "name": "video_fisheye_cam1", "l2": 5.697859, "mean_abs": 0.123727, "relative": 0.941523 }, { "name": "video_fisheye_cam2", "l2": 5.541004, "mean_abs": 0.115607, "relative": 0.884539 }, { "name": "video_fisheye_cam3", "l2": 5.600387, "mean_abs": 0.114643, "relative": 0.937717 }, { "name": "video_stereo_left", "l2": 5.13948, "mean_abs": 0.118194, "relative": 0.835176 }, { "name": "video_stereo_right", "l2": 5.973885, "mean_abs": 0.13834, "relative": 0.842192 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.973547, "mean_abs": 0.08115, "relative": 0.706303 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012723, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.999990701675415" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9995761513710022" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999974966049194" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 953, "start_frame": 4765, "end_frame": 4784, "center_frame": 4774, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.120994, "mean_abs": 0.216623, "relative": 0.853749 }, { "name": "hand_right_joints", "l2": 10.693751, "mean_abs": 0.256256, "relative": 0.842954 }, { "name": "body_joints", "l2": 5.386755, "mean_abs": 0.0803, "relative": 0.727585 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.041019, "mean_abs": 0.005212, "relative": 0.096799 }, { "name": "camera_rotation_matrix", "l2": 3.010771, "mean_abs": 0.23043, "relative": 0.924559 }, { "name": "imu_accel_gyro", "l2": 17.145508, "mean_abs": 0.907219, "relative": 0.925059 }, { "name": "depth_confidence", "l2": 13.070197, "mean_abs": 0.237338, "relative": 0.000434 }, { "name": "video_fisheye_cam0", "l2": 5.098852, "mean_abs": 0.107307, "relative": 0.923293 }, { "name": "video_fisheye_cam1", "l2": 5.647803, "mean_abs": 0.124021, "relative": 0.933252 }, { "name": "video_fisheye_cam2", "l2": 5.549073, "mean_abs": 0.116482, "relative": 0.885827 }, { "name": "video_fisheye_cam3", "l2": 5.618597, "mean_abs": 0.115565, "relative": 0.940766 }, { "name": "video_stereo_left", "l2": 5.180396, "mean_abs": 0.120066, "relative": 0.841825 }, { "name": "video_stereo_right", "l2": 6.007691, "mean_abs": 0.140268, "relative": 0.846958 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.997043, "mean_abs": 0.080628, "relative": 0.714712 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012723, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9999579191207886" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9988012313842773" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.999990701675415" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 954, "start_frame": 4770, "end_frame": 4789, "center_frame": 4779, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.153837, "mean_abs": 0.217401, "relative": 0.856519 }, { "name": "hand_right_joints", "l2": 10.707982, "mean_abs": 0.25601, "relative": 0.844075 }, { "name": "body_joints", "l2": 5.397603, "mean_abs": 0.081085, "relative": 0.72905 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.053191, "mean_abs": 0.007577, "relative": 0.125523 }, { "name": "camera_rotation_matrix", "l2": 3.012454, "mean_abs": 0.230376, "relative": 0.925076 }, { "name": "imu_accel_gyro", "l2": 17.274759, "mean_abs": 0.951668, "relative": 0.932033 }, { "name": "depth_confidence", "l2": 13.022814, "mean_abs": 0.236432, "relative": 0.000432 }, { "name": "video_fisheye_cam0", "l2": 5.103447, "mean_abs": 0.107583, "relative": 0.924125 }, { "name": "video_fisheye_cam1", "l2": 5.562532, "mean_abs": 0.122298, "relative": 0.919161 }, { "name": "video_fisheye_cam2", "l2": 5.54779, "mean_abs": 0.116794, "relative": 0.885622 }, { "name": "video_fisheye_cam3", "l2": 5.617853, "mean_abs": 0.115752, "relative": 0.940642 }, { "name": "video_stereo_left", "l2": 5.203677, "mean_abs": 0.120541, "relative": 0.845608 }, { "name": "video_stereo_right", "l2": 6.001122, "mean_abs": 0.140164, "relative": 0.846032 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.461535, "mean_abs": 0.103766, "relative": 0.880946 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012723, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9998252987861633" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.996025562286377" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9996464252471924" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 955, "start_frame": 4775, "end_frame": 4794, "center_frame": 4784, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.163959, "mean_abs": 0.217144, "relative": 0.857373 }, { "name": "hand_right_joints", "l2": 10.701938, "mean_abs": 0.254519, "relative": 0.843599 }, { "name": "body_joints", "l2": 5.402743, "mean_abs": 0.080891, "relative": 0.729744 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.062411, "mean_abs": 0.009399, "relative": 0.147281 }, { "name": "camera_rotation_matrix", "l2": 3.01186, "mean_abs": 0.228428, "relative": 0.924893 }, { "name": "imu_accel_gyro", "l2": 17.28076, "mean_abs": 0.971923, "relative": 0.932356 }, { "name": "depth_confidence", "l2": 12.696122, "mean_abs": 0.23252, "relative": 0.000421 }, { "name": "video_fisheye_cam0", "l2": 5.110283, "mean_abs": 0.107501, "relative": 0.925363 }, { "name": "video_fisheye_cam1", "l2": 5.512475, "mean_abs": 0.120932, "relative": 0.91089 }, { "name": "video_fisheye_cam2", "l2": 5.543606, "mean_abs": 0.116333, "relative": 0.884954 }, { "name": "video_fisheye_cam3", "l2": 5.622602, "mean_abs": 0.115543, "relative": 0.941437 }, { "name": "video_stereo_left", "l2": 5.209067, "mean_abs": 0.119825, "relative": 0.846484 }, { "name": "video_stereo_right", "l2": 6.006652, "mean_abs": 0.139072, "relative": 0.846812 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.468695, "mean_abs": 0.105029, "relative": 0.883509 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012723, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9991575479507446" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9896037578582764" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9960676431655884" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|digital scale|water bottle|coffee mug", "correct": 0, "confidence": "" } } }, { "window_index": 956, "start_frame": 4780, "end_frame": 4799, "center_frame": 4789, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.185973, "mean_abs": 0.211199, "relative": 0.85923 }, { "name": "hand_right_joints", "l2": 10.684119, "mean_abs": 0.249014, "relative": 0.842194 }, { "name": "body_joints", "l2": 5.420709, "mean_abs": 0.079934, "relative": 0.732171 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.04984, "mean_abs": 0.007658, "relative": 0.117615 }, { "name": "camera_rotation_matrix", "l2": 3.001859, "mean_abs": 0.219477, "relative": 0.921822 }, { "name": "imu_accel_gyro", "l2": 17.094477, "mean_abs": 0.910615, "relative": 0.922306 }, { "name": "depth_confidence", "l2": 12.503126, "mean_abs": 0.224554, "relative": 0.000415 }, { "name": "video_fisheye_cam0", "l2": 5.115098, "mean_abs": 0.1065, "relative": 0.926235 }, { "name": "video_fisheye_cam1", "l2": 5.385711, "mean_abs": 0.115354, "relative": 0.889943 }, { "name": "video_fisheye_cam2", "l2": 5.524526, "mean_abs": 0.114255, "relative": 0.881908 }, { "name": "video_fisheye_cam3", "l2": 5.607409, "mean_abs": 0.11334, "relative": 0.938893 }, { "name": "video_stereo_left", "l2": 5.176565, "mean_abs": 0.115043, "relative": 0.841202 }, { "name": "video_stereo_right", "l2": 5.977687, "mean_abs": 0.133842, "relative": 0.842728 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.407678, "mean_abs": 0.096624, "relative": 0.861672 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.012723, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9965736865997314" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.963616132736206" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999959468841553" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.882043719291687" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "white mug|coffee equipment|small bottle|weighing scale", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 957, "start_frame": 4785, "end_frame": 4804, "center_frame": 4794, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale", "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.206836, "mean_abs": 0.210862, "relative": 0.86099 }, { "name": "hand_right_joints", "l2": 10.680539, "mean_abs": 0.247647, "relative": 0.841912 }, { "name": "body_joints", "l2": 5.415532, "mean_abs": 0.079692, "relative": 0.731472 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.033189, "mean_abs": 0.00509, "relative": 0.07832 }, { "name": "camera_rotation_matrix", "l2": 3.000016, "mean_abs": 0.214948, "relative": 0.921256 }, { "name": "imu_accel_gyro", "l2": 17.058121, "mean_abs": 0.89505, "relative": 0.920344 }, { "name": "depth_confidence", "l2": 12.41253, "mean_abs": 0.221008, "relative": 0.000412 }, { "name": "video_fisheye_cam0", "l2": 5.122777, "mean_abs": 0.106086, "relative": 0.927625 }, { "name": "video_fisheye_cam1", "l2": 5.341812, "mean_abs": 0.112713, "relative": 0.882689 }, { "name": "video_fisheye_cam2", "l2": 5.516877, "mean_abs": 0.113112, "relative": 0.880687 }, { "name": "video_fisheye_cam3", "l2": 5.605552, "mean_abs": 0.112466, "relative": 0.938582 }, { "name": "video_stereo_left", "l2": 5.17786, "mean_abs": 0.112948, "relative": 0.841412 }, { "name": "video_stereo_right", "l2": 5.975214, "mean_abs": 0.131598, "relative": 0.84238 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.178365, "mean_abs": 0.086232, "relative": 0.779604 }, { "name": "caption_objects_interaction_text", "l2": 1.981493, "mean_abs": 0.021673, "relative": 0.909854 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9930694699287415" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.8999869227409363" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.9771793484687805" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8206517696380615" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|white mug|water bottle|coffee equipment|small bottle|weighing scale|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 958, "start_frame": 4790, "end_frame": 4809, "center_frame": 4799, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale", "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.205623, "mean_abs": 0.21086, "relative": 0.860888 }, { "name": "hand_right_joints", "l2": 10.686698, "mean_abs": 0.247452, "relative": 0.842398 }, { "name": "body_joints", "l2": 5.4013, "mean_abs": 0.07971, "relative": 0.729549 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017716, "mean_abs": 0.002688, "relative": 0.041808 }, { "name": "camera_rotation_matrix", "l2": 3.001046, "mean_abs": 0.213204, "relative": 0.921573 }, { "name": "imu_accel_gyro", "l2": 17.090851, "mean_abs": 0.885032, "relative": 0.92211 }, { "name": "depth_confidence", "l2": 12.367785, "mean_abs": 0.218558, "relative": 0.00041 }, { "name": "video_fisheye_cam0", "l2": 5.126266, "mean_abs": 0.106018, "relative": 0.928257 }, { "name": "video_fisheye_cam1", "l2": 5.329953, "mean_abs": 0.111858, "relative": 0.88073 }, { "name": "video_fisheye_cam2", "l2": 5.508928, "mean_abs": 0.112461, "relative": 0.879418 }, { "name": "video_fisheye_cam3", "l2": 5.601704, "mean_abs": 0.112401, "relative": 0.937938 }, { "name": "video_stereo_left", "l2": 5.168473, "mean_abs": 0.111664, "relative": 0.839887 }, { "name": "video_stereo_right", "l2": 5.979154, "mean_abs": 0.130696, "relative": 0.842935 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.43769, "mean_abs": 0.100873, "relative": 0.872413 }, { "name": "caption_objects_interaction_text", "l2": 1.981493, "mean_abs": 0.022114, "relative": 0.909854 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9422296285629272" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9115092754364014" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.9844224452972412" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9041789770126343" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|white mug|water bottle|coffee equipment|small bottle|weighing scale|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 959, "start_frame": 4795, "end_frame": 4814, "center_frame": 4804, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "coffee equipment", "small bottle", "white mug", "weighing scale", "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.208703, "mean_abs": 0.210464, "relative": 0.861148 }, { "name": "hand_right_joints", "l2": 10.684535, "mean_abs": 0.246968, "relative": 0.842227 }, { "name": "body_joints", "l2": 5.380591, "mean_abs": 0.079079, "relative": 0.726752 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009033, "mean_abs": 0.001498, "relative": 0.021317 }, { "name": "camera_rotation_matrix", "l2": 2.999954, "mean_abs": 0.212762, "relative": 0.921237 }, { "name": "imu_accel_gyro", "l2": 17.08819, "mean_abs": 0.883262, "relative": 0.921967 }, { "name": "depth_confidence", "l2": 12.584442, "mean_abs": 0.220029, "relative": 0.000418 }, { "name": "video_fisheye_cam0", "l2": 5.124955, "mean_abs": 0.106141, "relative": 0.92802 }, { "name": "video_fisheye_cam1", "l2": 5.356501, "mean_abs": 0.112667, "relative": 0.885117 }, { "name": "video_fisheye_cam2", "l2": 5.506119, "mean_abs": 0.112573, "relative": 0.87897 }, { "name": "video_fisheye_cam3", "l2": 5.590735, "mean_abs": 0.11266, "relative": 0.936101 }, { "name": "video_stereo_left", "l2": 5.163399, "mean_abs": 0.111872, "relative": 0.839063 }, { "name": "video_stereo_right", "l2": 6.016497, "mean_abs": 0.131973, "relative": 0.8482 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.435339, "mean_abs": 0.101739, "relative": 0.871571 }, { "name": "caption_objects_interaction_text", "l2": 1.981493, "mean_abs": 0.021897, "relative": 0.909854 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8037304878234863" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.9393290877342224" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.9973804354667664" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9650363922119141" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|white mug|water bottle|coffee equipment|small bottle|weighing scale|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 960, "start_frame": 4800, "end_frame": 4819, "center_frame": 4809, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.21472, "mean_abs": 0.209637, "relative": 0.861655 }, { "name": "hand_right_joints", "l2": 10.684999, "mean_abs": 0.246063, "relative": 0.842264 }, { "name": "body_joints", "l2": 5.361835, "mean_abs": 0.078477, "relative": 0.724219 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013235, "mean_abs": 0.00209, "relative": 0.031233 }, { "name": "camera_rotation_matrix", "l2": 2.99998, "mean_abs": 0.213044, "relative": 0.921245 }, { "name": "imu_accel_gyro", "l2": 17.09692, "mean_abs": 0.891395, "relative": 0.922438 }, { "name": "depth_confidence", "l2": 12.616581, "mean_abs": 0.220343, "relative": 0.000419 }, { "name": "video_fisheye_cam0", "l2": 5.125484, "mean_abs": 0.10624, "relative": 0.928116 }, { "name": "video_fisheye_cam1", "l2": 5.369013, "mean_abs": 0.113216, "relative": 0.887184 }, { "name": "video_fisheye_cam2", "l2": 5.504949, "mean_abs": 0.112588, "relative": 0.878783 }, { "name": "video_fisheye_cam3", "l2": 5.594363, "mean_abs": 0.112851, "relative": 0.936709 }, { "name": "video_stereo_left", "l2": 5.165765, "mean_abs": 0.112298, "relative": 0.839447 }, { "name": "video_stereo_right", "l2": 6.037691, "mean_abs": 0.133114, "relative": 0.851188 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.421412, "mean_abs": 0.100598, "relative": 0.866587 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8388046622276306" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.833260715007782" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999681711196899" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9901487827301025" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 961, "start_frame": 4805, "end_frame": 4824, "center_frame": 4814, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.202473, "mean_abs": 0.209994, "relative": 0.860622 }, { "name": "hand_right_joints", "l2": 10.693203, "mean_abs": 0.246885, "relative": 0.84291 }, { "name": "body_joints", "l2": 5.358876, "mean_abs": 0.078618, "relative": 0.723819 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013761, "mean_abs": 0.002099, "relative": 0.032475 }, { "name": "camera_rotation_matrix", "l2": 2.999782, "mean_abs": 0.212094, "relative": 0.921184 }, { "name": "imu_accel_gyro", "l2": 17.07514, "mean_abs": 0.887792, "relative": 0.921263 }, { "name": "depth_confidence", "l2": 12.687454, "mean_abs": 0.219175, "relative": 0.000421 }, { "name": "video_fisheye_cam0", "l2": 5.126054, "mean_abs": 0.106174, "relative": 0.928219 }, { "name": "video_fisheye_cam1", "l2": 5.377033, "mean_abs": 0.113141, "relative": 0.888509 }, { "name": "video_fisheye_cam2", "l2": 5.497712, "mean_abs": 0.11228, "relative": 0.877628 }, { "name": "video_fisheye_cam3", "l2": 5.587669, "mean_abs": 0.113179, "relative": 0.935588 }, { "name": "video_stereo_left", "l2": 5.159167, "mean_abs": 0.111803, "relative": 0.838375 }, { "name": "video_stereo_right", "l2": 6.04596, "mean_abs": 0.13262, "relative": 0.852354 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.160344, "mean_abs": 0.090018, "relative": 0.773155 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.6672843098640442" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.7204363942146301" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9989945292472839" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8987060189247131" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 962, "start_frame": 4810, "end_frame": 4829, "center_frame": 4819, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.199207, "mean_abs": 0.210508, "relative": 0.860347 }, { "name": "hand_right_joints", "l2": 10.669287, "mean_abs": 0.247153, "relative": 0.841025 }, { "name": "body_joints", "l2": 5.365949, "mean_abs": 0.078927, "relative": 0.724774 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016555, "mean_abs": 0.002491, "relative": 0.039067 }, { "name": "camera_rotation_matrix", "l2": 3.000135, "mean_abs": 0.211912, "relative": 0.921293 }, { "name": "imu_accel_gyro", "l2": 17.074928, "mean_abs": 0.884358, "relative": 0.921251 }, { "name": "depth_confidence", "l2": 12.655128, "mean_abs": 0.218924, "relative": 0.00042 }, { "name": "video_fisheye_cam0", "l2": 5.124693, "mean_abs": 0.105975, "relative": 0.927972 }, { "name": "video_fisheye_cam1", "l2": 5.392325, "mean_abs": 0.113174, "relative": 0.891036 }, { "name": "video_fisheye_cam2", "l2": 5.49272, "mean_abs": 0.112069, "relative": 0.876831 }, { "name": "video_fisheye_cam3", "l2": 5.57457, "mean_abs": 0.112406, "relative": 0.933394 }, { "name": "video_stereo_left", "l2": 5.154113, "mean_abs": 0.11156, "relative": 0.837554 }, { "name": "video_stereo_right", "l2": 6.06826, "mean_abs": 0.132893, "relative": 0.855497 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.144419, "mean_abs": 0.087179, "relative": 0.767455 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.6682916283607483" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.6137091517448425" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.999829888343811" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.7827909588813782" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 963, "start_frame": 4815, "end_frame": 4834, "center_frame": 4824, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.193157, "mean_abs": 0.211523, "relative": 0.859836 }, { "name": "hand_right_joints", "l2": 10.669554, "mean_abs": 0.24774, "relative": 0.841046 }, { "name": "body_joints", "l2": 5.371871, "mean_abs": 0.079241, "relative": 0.725574 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018152, "mean_abs": 0.002506, "relative": 0.042836 }, { "name": "camera_rotation_matrix", "l2": 2.999968, "mean_abs": 0.211663, "relative": 0.921241 }, { "name": "imu_accel_gyro", "l2": 17.090326, "mean_abs": 0.890703, "relative": 0.922082 }, { "name": "depth_confidence", "l2": 12.558547, "mean_abs": 0.218036, "relative": 0.000417 }, { "name": "video_fisheye_cam0", "l2": 5.128179, "mean_abs": 0.106206, "relative": 0.928604 }, { "name": "video_fisheye_cam1", "l2": 5.394929, "mean_abs": 0.113752, "relative": 0.891467 }, { "name": "video_fisheye_cam2", "l2": 5.491223, "mean_abs": 0.11218, "relative": 0.876592 }, { "name": "video_fisheye_cam3", "l2": 5.569016, "mean_abs": 0.113419, "relative": 0.932464 }, { "name": "video_stereo_left", "l2": 5.157183, "mean_abs": 0.111969, "relative": 0.838052 }, { "name": "video_stereo_right", "l2": 6.08741, "mean_abs": 0.134057, "relative": 0.858197 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.097351, "mean_abs": 0.084608, "relative": 0.75061 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.638683021068573" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.7667366862297058" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9972297549247742" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.7491406202316284" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 964, "start_frame": 4820, "end_frame": 4839, "center_frame": 4829, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.191344, "mean_abs": 0.212487, "relative": 0.859683 }, { "name": "hand_right_joints", "l2": 10.649647, "mean_abs": 0.249485, "relative": 0.839477 }, { "name": "body_joints", "l2": 5.390895, "mean_abs": 0.08031, "relative": 0.728144 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020871, "mean_abs": 0.00248, "relative": 0.049252 }, { "name": "camera_rotation_matrix", "l2": 2.999676, "mean_abs": 0.211552, "relative": 0.921152 }, { "name": "imu_accel_gyro", "l2": 17.062838, "mean_abs": 0.91428, "relative": 0.920599 }, { "name": "depth_confidence", "l2": 12.645836, "mean_abs": 0.219397, "relative": 0.00042 }, { "name": "video_fisheye_cam0", "l2": 5.129831, "mean_abs": 0.106037, "relative": 0.928903 }, { "name": "video_fisheye_cam1", "l2": 5.379403, "mean_abs": 0.112901, "relative": 0.888901 }, { "name": "video_fisheye_cam2", "l2": 5.48887, "mean_abs": 0.111906, "relative": 0.876216 }, { "name": "video_fisheye_cam3", "l2": 5.537362, "mean_abs": 0.112193, "relative": 0.927164 }, { "name": "video_stereo_left", "l2": 5.157744, "mean_abs": 0.111574, "relative": 0.838144 }, { "name": "video_stereo_right", "l2": 6.079394, "mean_abs": 0.133094, "relative": 0.857067 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.293745, "mean_abs": 0.099769, "relative": 0.820897 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.5276715755462646" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.6227928400039673" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.999983549118042" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.6980162262916565" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 965, "start_frame": 4825, "end_frame": 4844, "center_frame": 4834, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.159458, "mean_abs": 0.211446, "relative": 0.856994 }, { "name": "hand_right_joints", "l2": 10.701193, "mean_abs": 0.255529, "relative": 0.84354 }, { "name": "body_joints", "l2": 5.429, "mean_abs": 0.081495, "relative": 0.733291 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.038509, "mean_abs": 0.005133, "relative": 0.090877 }, { "name": "camera_rotation_matrix", "l2": 3.000952, "mean_abs": 0.214541, "relative": 0.921544 }, { "name": "imu_accel_gyro", "l2": 16.988358, "mean_abs": 0.941564, "relative": 0.91658 }, { "name": "depth_confidence", "l2": 12.603239, "mean_abs": 0.221364, "relative": 0.000418 }, { "name": "video_fisheye_cam0", "l2": 5.137503, "mean_abs": 0.107274, "relative": 0.930292 }, { "name": "video_fisheye_cam1", "l2": 5.399255, "mean_abs": 0.115415, "relative": 0.892181 }, { "name": "video_fisheye_cam2", "l2": 5.47457, "mean_abs": 0.11355, "relative": 0.873934 }, { "name": "video_fisheye_cam3", "l2": 5.537348, "mean_abs": 0.113404, "relative": 0.927162 }, { "name": "video_stereo_left", "l2": 5.176053, "mean_abs": 0.115496, "relative": 0.841119 }, { "name": "video_stereo_right", "l2": 6.123428, "mean_abs": 0.138264, "relative": 0.863275 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.474297, "mean_abs": 0.109862, "relative": 0.885514 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.48482465744018555" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.5460715889930725" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.5217021703720093" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 966, "start_frame": 4830, "end_frame": 4849, "center_frame": 4839, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.130571, "mean_abs": 0.211262, "relative": 0.854557 }, { "name": "hand_right_joints", "l2": 10.728157, "mean_abs": 0.257055, "relative": 0.845666 }, { "name": "body_joints", "l2": 5.435156, "mean_abs": 0.081486, "relative": 0.734122 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.041923, "mean_abs": 0.005826, "relative": 0.098932 }, { "name": "camera_rotation_matrix", "l2": 3.001657, "mean_abs": 0.214926, "relative": 0.92176 }, { "name": "imu_accel_gyro", "l2": 17.007498, "mean_abs": 0.934632, "relative": 0.917613 }, { "name": "depth_confidence", "l2": 12.614373, "mean_abs": 0.221789, "relative": 0.000419 }, { "name": "video_fisheye_cam0", "l2": 5.13306, "mean_abs": 0.10737, "relative": 0.929487 }, { "name": "video_fisheye_cam1", "l2": 5.404023, "mean_abs": 0.115918, "relative": 0.892969 }, { "name": "video_fisheye_cam2", "l2": 5.461801, "mean_abs": 0.113574, "relative": 0.871895 }, { "name": "video_fisheye_cam3", "l2": 5.533001, "mean_abs": 0.113831, "relative": 0.926434 }, { "name": "video_stereo_left", "l2": 5.170906, "mean_abs": 0.116154, "relative": 0.840282 }, { "name": "video_stereo_right", "l2": 6.128644, "mean_abs": 0.138922, "relative": 0.86401 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.217393, "mean_abs": 0.09523, "relative": 0.793572 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.6003479361534119" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.6762330532073975" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.5684183835983276" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 967, "start_frame": 4835, "end_frame": 4854, "center_frame": 4844, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.113182, "mean_abs": 0.210984, "relative": 0.85309 }, { "name": "hand_right_joints", "l2": 10.733407, "mean_abs": 0.254521, "relative": 0.84608 }, { "name": "body_joints", "l2": 5.428827, "mean_abs": 0.080525, "relative": 0.733267 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.039656, "mean_abs": 0.005825, "relative": 0.093583 }, { "name": "camera_rotation_matrix", "l2": 3.001525, "mean_abs": 0.213827, "relative": 0.92172 }, { "name": "imu_accel_gyro", "l2": 17.07798, "mean_abs": 0.951161, "relative": 0.921416 }, { "name": "depth_confidence", "l2": 12.567231, "mean_abs": 0.220211, "relative": 0.000417 }, { "name": "video_fisheye_cam0", "l2": 5.131246, "mean_abs": 0.107266, "relative": 0.929159 }, { "name": "video_fisheye_cam1", "l2": 5.399905, "mean_abs": 0.115324, "relative": 0.892289 }, { "name": "video_fisheye_cam2", "l2": 5.456878, "mean_abs": 0.11334, "relative": 0.871109 }, { "name": "video_fisheye_cam3", "l2": 5.518205, "mean_abs": 0.113737, "relative": 0.923957 }, { "name": "video_stereo_left", "l2": 5.16198, "mean_abs": 0.1154, "relative": 0.838832 }, { "name": "video_stereo_right", "l2": 6.119576, "mean_abs": 0.138006, "relative": 0.862732 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.295589, "mean_abs": 0.101943, "relative": 0.821557 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.40354663133621216" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.8653438687324524" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.5237605571746826" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "kettle|coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 968, "start_frame": 4840, "end_frame": 4859, "center_frame": 4849, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.109855, "mean_abs": 0.212713, "relative": 0.852809 }, { "name": "hand_right_joints", "l2": 10.616168, "mean_abs": 0.25197, "relative": 0.836838 }, { "name": "body_joints", "l2": 5.436241, "mean_abs": 0.081005, "relative": 0.734269 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.027695, "mean_abs": 0.004148, "relative": 0.065356 }, { "name": "camera_rotation_matrix", "l2": 3.003484, "mean_abs": 0.216104, "relative": 0.922321 }, { "name": "imu_accel_gyro", "l2": 17.115807, "mean_abs": 0.99203, "relative": 0.923457 }, { "name": "depth_confidence", "l2": 12.507475, "mean_abs": 0.221654, "relative": 0.000415 }, { "name": "video_fisheye_cam0", "l2": 5.139988, "mean_abs": 0.107551, "relative": 0.930742 }, { "name": "video_fisheye_cam1", "l2": 5.427052, "mean_abs": 0.115875, "relative": 0.896775 }, { "name": "video_fisheye_cam2", "l2": 5.489019, "mean_abs": 0.11382, "relative": 0.87624 }, { "name": "video_fisheye_cam3", "l2": 5.495162, "mean_abs": 0.112071, "relative": 0.920099 }, { "name": "video_stereo_left", "l2": 5.196802, "mean_abs": 0.116918, "relative": 0.844491 }, { "name": "video_stereo_right", "l2": 6.193903, "mean_abs": 0.140988, "relative": 0.87321 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.335092, "mean_abs": 0.103277, "relative": 0.835694 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.40534141659736633" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.7651394605636597" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.6312804818153381" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 969, "start_frame": 4845, "end_frame": 4864, "center_frame": 4854, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.128821, "mean_abs": 0.216737, "relative": 0.854409 }, { "name": "hand_right_joints", "l2": 10.592303, "mean_abs": 0.262521, "relative": 0.834957 }, { "name": "body_joints", "l2": 5.439816, "mean_abs": 0.082877, "relative": 0.734751 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.023585, "mean_abs": 0.00361, "relative": 0.055658 }, { "name": "camera_rotation_matrix", "l2": 3.00952, "mean_abs": 0.225604, "relative": 0.924175 }, { "name": "imu_accel_gyro", "l2": 17.302265, "mean_abs": 1.029653, "relative": 0.933517 }, { "name": "depth_confidence", "l2": 12.473935, "mean_abs": 0.22623, "relative": 0.000414 }, { "name": "video_fisheye_cam0", "l2": 5.140265, "mean_abs": 0.10898, "relative": 0.930792 }, { "name": "video_fisheye_cam1", "l2": 5.415379, "mean_abs": 0.11802, "relative": 0.894846 }, { "name": "video_fisheye_cam2", "l2": 5.525136, "mean_abs": 0.11602, "relative": 0.882006 }, { "name": "video_fisheye_cam3", "l2": 5.449841, "mean_abs": 0.112259, "relative": 0.91251 }, { "name": "video_stereo_left", "l2": 5.216505, "mean_abs": 0.11988, "relative": 0.847692 }, { "name": "video_stereo_right", "l2": 6.216692, "mean_abs": 0.144777, "relative": 0.876423 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.28337, "mean_abs": 0.102488, "relative": 0.817184 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.5524398684501648" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.6884707808494568" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.6709341406822205" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 970, "start_frame": 4850, "end_frame": 4869, "center_frame": 4859, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.136192, "mean_abs": 0.217592, "relative": 0.855031 }, { "name": "hand_right_joints", "l2": 10.535342, "mean_abs": 0.260962, "relative": 0.830467 }, { "name": "body_joints", "l2": 5.459601, "mean_abs": 0.08396, "relative": 0.737424 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.022713, "mean_abs": 0.003429, "relative": 0.0536 }, { "name": "camera_rotation_matrix", "l2": 3.005555, "mean_abs": 0.22475, "relative": 0.922957 }, { "name": "imu_accel_gyro", "l2": 17.298109, "mean_abs": 1.02301, "relative": 0.933293 }, { "name": "depth_confidence", "l2": 13.115161, "mean_abs": 0.228457, "relative": 0.000435 }, { "name": "video_fisheye_cam0", "l2": 5.136819, "mean_abs": 0.108511, "relative": 0.930168 }, { "name": "video_fisheye_cam1", "l2": 5.388457, "mean_abs": 0.117153, "relative": 0.890397 }, { "name": "video_fisheye_cam2", "l2": 5.540298, "mean_abs": 0.115847, "relative": 0.884426 }, { "name": "video_fisheye_cam3", "l2": 5.432555, "mean_abs": 0.111485, "relative": 0.909616 }, { "name": "video_stereo_left", "l2": 5.201108, "mean_abs": 0.118891, "relative": 0.84519 }, { "name": "video_stereo_right", "l2": 6.189869, "mean_abs": 0.143483, "relative": 0.872642 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.247788, "mean_abs": 0.099133, "relative": 0.804449 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.29971522092819214" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.7166640758514404" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.36488616466522217" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 971, "start_frame": 4855, "end_frame": 4874, "center_frame": 4864, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.155417, "mean_abs": 0.218511, "relative": 0.856653 }, { "name": "hand_right_joints", "l2": 10.468633, "mean_abs": 0.255993, "relative": 0.825208 }, { "name": "body_joints", "l2": 5.454906, "mean_abs": 0.083343, "relative": 0.73679 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.031255, "mean_abs": 0.00511, "relative": 0.073757 }, { "name": "camera_rotation_matrix", "l2": 3.002406, "mean_abs": 0.22257, "relative": 0.92199 }, { "name": "imu_accel_gyro", "l2": 17.202732, "mean_abs": 1.027156, "relative": 0.928147 }, { "name": "depth_confidence", "l2": 13.13381, "mean_abs": 0.226598, "relative": 0.000436 }, { "name": "video_fisheye_cam0", "l2": 5.137053, "mean_abs": 0.108003, "relative": 0.930211 }, { "name": "video_fisheye_cam1", "l2": 5.373881, "mean_abs": 0.115575, "relative": 0.887988 }, { "name": "video_fisheye_cam2", "l2": 5.561668, "mean_abs": 0.115081, "relative": 0.887838 }, { "name": "video_fisheye_cam3", "l2": 5.450681, "mean_abs": 0.111177, "relative": 0.912651 }, { "name": "video_stereo_left", "l2": 5.188233, "mean_abs": 0.116624, "relative": 0.843098 }, { "name": "video_stereo_right", "l2": 6.167047, "mean_abs": 0.140533, "relative": 0.869424 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.461789, "mean_abs": 0.109102, "relative": 0.881037 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.5432841777801514" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.6918377876281738" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.895662248134613" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 972, "start_frame": 4860, "end_frame": 4879, "center_frame": 4869, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.174807, "mean_abs": 0.217779, "relative": 0.858288 }, { "name": "hand_right_joints", "l2": 10.40529, "mean_abs": 0.248024, "relative": 0.820215 }, { "name": "body_joints", "l2": 5.469088, "mean_abs": 0.083796, "relative": 0.738705 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.048866, "mean_abs": 0.006666, "relative": 0.115316 }, { "name": "camera_rotation_matrix", "l2": 3.000562, "mean_abs": 0.221138, "relative": 0.921424 }, { "name": "imu_accel_gyro", "l2": 17.241993, "mean_abs": 1.03936, "relative": 0.930265 }, { "name": "depth_confidence", "l2": 13.175485, "mean_abs": 0.224632, "relative": 0.000437 }, { "name": "video_fisheye_cam0", "l2": 5.123617, "mean_abs": 0.107177, "relative": 0.927777 }, { "name": "video_fisheye_cam1", "l2": 5.365204, "mean_abs": 0.114122, "relative": 0.886555 }, { "name": "video_fisheye_cam2", "l2": 5.582476, "mean_abs": 0.114648, "relative": 0.891159 }, { "name": "video_fisheye_cam3", "l2": 5.448249, "mean_abs": 0.110607, "relative": 0.912244 }, { "name": "video_stereo_left", "l2": 5.181542, "mean_abs": 0.115188, "relative": 0.842011 }, { "name": "video_stereo_right", "l2": 6.192136, "mean_abs": 0.139339, "relative": 0.872961 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.21132, "mean_abs": 0.089971, "relative": 0.791398 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9190503358840942" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.6107038259506226" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.999969482421875" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9291398525238037" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 973, "start_frame": 4865, "end_frame": 4884, "center_frame": 4874, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.172586, "mean_abs": 0.219709, "relative": 0.858101 }, { "name": "hand_right_joints", "l2": 10.482815, "mean_abs": 0.248318, "relative": 0.826326 }, { "name": "body_joints", "l2": 5.486438, "mean_abs": 0.082796, "relative": 0.741049 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.047307, "mean_abs": 0.006127, "relative": 0.111639 }, { "name": "camera_rotation_matrix", "l2": 3.0011, "mean_abs": 0.220123, "relative": 0.921589 }, { "name": "imu_accel_gyro", "l2": 17.014042, "mean_abs": 0.957168, "relative": 0.917966 }, { "name": "depth_confidence", "l2": 13.264666, "mean_abs": 0.225125, "relative": 0.00044 }, { "name": "video_fisheye_cam0", "l2": 5.118627, "mean_abs": 0.106897, "relative": 0.926874 }, { "name": "video_fisheye_cam1", "l2": 5.373926, "mean_abs": 0.114164, "relative": 0.887996 }, { "name": "video_fisheye_cam2", "l2": 5.569, "mean_abs": 0.114408, "relative": 0.889008 }, { "name": "video_fisheye_cam3", "l2": 5.490717, "mean_abs": 0.110602, "relative": 0.919354 }, { "name": "video_stereo_left", "l2": 5.179008, "mean_abs": 0.115009, "relative": 0.841599 }, { "name": "video_stereo_right", "l2": 6.195315, "mean_abs": 0.139304, "relative": 0.873409 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.085587, "mean_abs": 0.080997, "relative": 0.7464 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8465105295181274" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.6128640174865723" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999697208404541" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8150931596755981" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 974, "start_frame": 4870, "end_frame": 4889, "center_frame": 4879, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.163501, "mean_abs": 0.219746, "relative": 0.857335 }, { "name": "hand_right_joints", "l2": 10.459578, "mean_abs": 0.251743, "relative": 0.824495 }, { "name": "body_joints", "l2": 5.546026, "mean_abs": 0.08332, "relative": 0.749097 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.025549, "mean_abs": 0.003774, "relative": 0.060291 }, { "name": "camera_rotation_matrix", "l2": 3.001629, "mean_abs": 0.217633, "relative": 0.921752 }, { "name": "imu_accel_gyro", "l2": 17.039358, "mean_abs": 0.942068, "relative": 0.919332 }, { "name": "depth_confidence", "l2": 12.521047, "mean_abs": 0.220375, "relative": 0.000416 }, { "name": "video_fisheye_cam0", "l2": 5.115439, "mean_abs": 0.106321, "relative": 0.926297 }, { "name": "video_fisheye_cam1", "l2": 5.391134, "mean_abs": 0.113999, "relative": 0.890839 }, { "name": "video_fisheye_cam2", "l2": 5.549933, "mean_abs": 0.113623, "relative": 0.885964 }, { "name": "video_fisheye_cam3", "l2": 5.508743, "mean_abs": 0.110958, "relative": 0.922373 }, { "name": "video_stereo_left", "l2": 5.171552, "mean_abs": 0.113701, "relative": 0.840387 }, { "name": "video_stereo_right", "l2": 6.199455, "mean_abs": 0.138043, "relative": 0.873993 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.199429, "mean_abs": 0.088999, "relative": 0.787143 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8300129175186157" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.569987952709198" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999908208847046" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.7099605798721313" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 975, "start_frame": 4875, "end_frame": 4894, "center_frame": 4884, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.155818, "mean_abs": 0.21978, "relative": 0.856687 }, { "name": "hand_right_joints", "l2": 10.439783, "mean_abs": 0.252141, "relative": 0.822934 }, { "name": "body_joints", "l2": 5.585535, "mean_abs": 0.083016, "relative": 0.754434 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.032747, "mean_abs": 0.0048, "relative": 0.077277 }, { "name": "camera_rotation_matrix", "l2": 3.000593, "mean_abs": 0.216596, "relative": 0.921433 }, { "name": "imu_accel_gyro", "l2": 17.042542, "mean_abs": 0.945946, "relative": 0.919504 }, { "name": "depth_confidence", "l2": 12.595127, "mean_abs": 0.220261, "relative": 0.000418 }, { "name": "video_fisheye_cam0", "l2": 5.11545, "mean_abs": 0.10624, "relative": 0.926299 }, { "name": "video_fisheye_cam1", "l2": 5.410985, "mean_abs": 0.114214, "relative": 0.89412 }, { "name": "video_fisheye_cam2", "l2": 5.54567, "mean_abs": 0.113602, "relative": 0.885284 }, { "name": "video_fisheye_cam3", "l2": 5.529714, "mean_abs": 0.111099, "relative": 0.925884 }, { "name": "video_stereo_left", "l2": 5.183441, "mean_abs": 0.113858, "relative": 0.842319 }, { "name": "video_stereo_right", "l2": 6.221772, "mean_abs": 0.137891, "relative": 0.877139 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.216965, "mean_abs": 0.089092, "relative": 0.793418 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8345176577568054" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.6781909465789795" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999960660934448" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.5660410523414612" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 976, "start_frame": 4880, "end_frame": 4899, "center_frame": 4889, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.145681, "mean_abs": 0.218683, "relative": 0.855832 }, { "name": "hand_right_joints", "l2": 10.425457, "mean_abs": 0.251981, "relative": 0.821805 }, { "name": "body_joints", "l2": 5.600272, "mean_abs": 0.082859, "relative": 0.756424 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.039689, "mean_abs": 0.005528, "relative": 0.093662 }, { "name": "camera_rotation_matrix", "l2": 2.99955, "mean_abs": 0.216033, "relative": 0.921113 }, { "name": "imu_accel_gyro", "l2": 17.029182, "mean_abs": 0.930439, "relative": 0.918783 }, { "name": "depth_confidence", "l2": 12.41604, "mean_abs": 0.219036, "relative": 0.000412 }, { "name": "video_fisheye_cam0", "l2": 5.124415, "mean_abs": 0.106556, "relative": 0.927922 }, { "name": "video_fisheye_cam1", "l2": 5.446023, "mean_abs": 0.115561, "relative": 0.899909 }, { "name": "video_fisheye_cam2", "l2": 5.547099, "mean_abs": 0.113851, "relative": 0.885512 }, { "name": "video_fisheye_cam3", "l2": 5.561712, "mean_abs": 0.112298, "relative": 0.931242 }, { "name": "video_stereo_left", "l2": 5.225946, "mean_abs": 0.115248, "relative": 0.849227 }, { "name": "video_stereo_right", "l2": 6.239317, "mean_abs": 0.138897, "relative": 0.879613 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.125886, "mean_abs": 0.083955, "relative": 0.760823 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8224484920501709" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.5443931221961975" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999967813491821" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.6843630075454712" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 977, "start_frame": 4885, "end_frame": 4904, "center_frame": 4894, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.101404, "mean_abs": 0.214857, "relative": 0.852097 }, { "name": "hand_right_joints", "l2": 10.413763, "mean_abs": 0.249479, "relative": 0.820883 }, { "name": "body_joints", "l2": 5.622716, "mean_abs": 0.082723, "relative": 0.759456 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.040162, "mean_abs": 0.004796, "relative": 0.094776 }, { "name": "camera_rotation_matrix", "l2": 3.000379, "mean_abs": 0.215828, "relative": 0.921368 }, { "name": "imu_accel_gyro", "l2": 17.11375, "mean_abs": 0.930269, "relative": 0.923346 }, { "name": "depth_confidence", "l2": 12.728041, "mean_abs": 0.220661, "relative": 0.000422 }, { "name": "video_fisheye_cam0", "l2": 5.132961, "mean_abs": 0.106753, "relative": 0.92947 }, { "name": "video_fisheye_cam1", "l2": 5.472436, "mean_abs": 0.116041, "relative": 0.904274 }, { "name": "video_fisheye_cam2", "l2": 5.550292, "mean_abs": 0.113957, "relative": 0.886022 }, { "name": "video_fisheye_cam3", "l2": 5.589678, "mean_abs": 0.112884, "relative": 0.935924 }, { "name": "video_stereo_left", "l2": 5.238741, "mean_abs": 0.115595, "relative": 0.851306 }, { "name": "video_stereo_right", "l2": 6.264785, "mean_abs": 0.139704, "relative": 0.883203 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.135337, "mean_abs": 0.087513, "relative": 0.764205 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.7610405683517456" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.5589603185653687" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999868869781494" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.7757101655006409" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 978, "start_frame": 4890, "end_frame": 4909, "center_frame": 4899, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.07287, "mean_abs": 0.212021, "relative": 0.84969 }, { "name": "hand_right_joints", "l2": 10.382048, "mean_abs": 0.247493, "relative": 0.818383 }, { "name": "body_joints", "l2": 5.61962, "mean_abs": 0.082481, "relative": 0.759038 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.045626, "mean_abs": 0.005615, "relative": 0.107671 }, { "name": "camera_rotation_matrix", "l2": 3.001046, "mean_abs": 0.215912, "relative": 0.921573 }, { "name": "imu_accel_gyro", "l2": 17.052217, "mean_abs": 0.921544, "relative": 0.920026 }, { "name": "depth_confidence", "l2": 12.757634, "mean_abs": 0.221773, "relative": 0.000423 }, { "name": "video_fisheye_cam0", "l2": 5.145759, "mean_abs": 0.107152, "relative": 0.931787 }, { "name": "video_fisheye_cam1", "l2": 5.494965, "mean_abs": 0.116681, "relative": 0.907997 }, { "name": "video_fisheye_cam2", "l2": 5.549926, "mean_abs": 0.114172, "relative": 0.885963 }, { "name": "video_fisheye_cam3", "l2": 5.623765, "mean_abs": 0.114164, "relative": 0.941632 }, { "name": "video_stereo_left", "l2": 5.244499, "mean_abs": 0.115727, "relative": 0.852241 }, { "name": "video_stereo_right", "l2": 6.295699, "mean_abs": 0.140474, "relative": 0.887562 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.131714, "mean_abs": 0.087372, "relative": 0.762909 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8329975605010986" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.5939263105392456" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999867677688599" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8416683673858643" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 979, "start_frame": 4895, "end_frame": 4914, "center_frame": 4904, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.062499, "mean_abs": 0.21227, "relative": 0.848815 }, { "name": "hand_right_joints", "l2": 10.387721, "mean_abs": 0.246908, "relative": 0.81883 }, { "name": "body_joints", "l2": 5.603739, "mean_abs": 0.082123, "relative": 0.756893 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.061246, "mean_abs": 0.007, "relative": 0.144531 }, { "name": "camera_rotation_matrix", "l2": 3.000235, "mean_abs": 0.21489, "relative": 0.921323 }, { "name": "imu_accel_gyro", "l2": 17.052334, "mean_abs": 0.942303, "relative": 0.920032 }, { "name": "depth_confidence", "l2": 12.70548, "mean_abs": 0.221366, "relative": 0.000422 }, { "name": "video_fisheye_cam0", "l2": 5.158033, "mean_abs": 0.107398, "relative": 0.934009 }, { "name": "video_fisheye_cam1", "l2": 5.508918, "mean_abs": 0.116766, "relative": 0.910302 }, { "name": "video_fisheye_cam2", "l2": 5.541347, "mean_abs": 0.114213, "relative": 0.884594 }, { "name": "video_fisheye_cam3", "l2": 5.623587, "mean_abs": 0.114092, "relative": 0.941602 }, { "name": "video_stereo_left", "l2": 5.249048, "mean_abs": 0.115429, "relative": 0.852981 }, { "name": "video_stereo_right", "l2": 6.291211, "mean_abs": 0.140283, "relative": 0.886929 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.479552, "mean_abs": 0.104889, "relative": 0.887394 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.7936522364616394" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.5720028877258301" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999655485153198" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.653057873249054" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|mug", "correct": 0, "confidence": "" } } }, { "window_index": 980, "start_frame": 4900, "end_frame": 4919, "center_frame": 4909, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.043287, "mean_abs": 0.210395, "relative": 0.847194 }, { "name": "hand_right_joints", "l2": 10.372494, "mean_abs": 0.246195, "relative": 0.81763 }, { "name": "body_joints", "l2": 5.574811, "mean_abs": 0.082169, "relative": 0.752985 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.086642, "mean_abs": 0.010337, "relative": 0.204463 }, { "name": "camera_rotation_matrix", "l2": 3.000604, "mean_abs": 0.213901, "relative": 0.921437 }, { "name": "imu_accel_gyro", "l2": 17.055347, "mean_abs": 0.938294, "relative": 0.920195 }, { "name": "depth_confidence", "l2": 12.747295, "mean_abs": 0.22239, "relative": 0.000423 }, { "name": "video_fisheye_cam0", "l2": 5.169817, "mean_abs": 0.107597, "relative": 0.936143 }, { "name": "video_fisheye_cam1", "l2": 5.535831, "mean_abs": 0.117433, "relative": 0.914749 }, { "name": "video_fisheye_cam2", "l2": 5.541054, "mean_abs": 0.114251, "relative": 0.884547 }, { "name": "video_fisheye_cam3", "l2": 5.629139, "mean_abs": 0.114528, "relative": 0.942531 }, { "name": "video_stereo_left", "l2": 5.266668, "mean_abs": 0.115309, "relative": 0.855844 }, { "name": "video_stereo_right", "l2": 6.284894, "mean_abs": 0.140393, "relative": 0.886038 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.248508, "mean_abs": 0.090926, "relative": 0.804707 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9082568883895874" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.6280051469802856" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999998807907104" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.9283531308174133" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 981, "start_frame": 4905, "end_frame": 4924, "center_frame": 4914, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.037066, "mean_abs": 0.210039, "relative": 0.846669 }, { "name": "hand_right_joints", "l2": 10.356236, "mean_abs": 0.245078, "relative": 0.816348 }, { "name": "body_joints", "l2": 5.558897, "mean_abs": 0.082471, "relative": 0.750836 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.096899, "mean_abs": 0.012182, "relative": 0.228667 }, { "name": "camera_rotation_matrix", "l2": 3.002252, "mean_abs": 0.214385, "relative": 0.921943 }, { "name": "imu_accel_gyro", "l2": 17.149405, "mean_abs": 0.965325, "relative": 0.925269 }, { "name": "depth_confidence", "l2": 12.497265, "mean_abs": 0.222431, "relative": 0.000415 }, { "name": "video_fisheye_cam0", "l2": 5.171999, "mean_abs": 0.10773, "relative": 0.936539 }, { "name": "video_fisheye_cam1", "l2": 5.554399, "mean_abs": 0.118596, "relative": 0.917818 }, { "name": "video_fisheye_cam2", "l2": 5.528338, "mean_abs": 0.114503, "relative": 0.882517 }, { "name": "video_fisheye_cam3", "l2": 5.616717, "mean_abs": 0.115115, "relative": 0.940452 }, { "name": "video_stereo_left", "l2": 5.256072, "mean_abs": 0.11602, "relative": 0.854122 }, { "name": "video_stereo_right", "l2": 6.284734, "mean_abs": 0.141697, "relative": 0.886016 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.265039, "mean_abs": 0.092163, "relative": 0.810624 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.7107468843460083" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.5905346870422363" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.8171787858009338" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|bottle|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 982, "start_frame": 4910, "end_frame": 4929, "center_frame": 4919, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.035838, "mean_abs": 0.209858, "relative": 0.846566 }, { "name": "hand_right_joints", "l2": 10.353883, "mean_abs": 0.243148, "relative": 0.816163 }, { "name": "body_joints", "l2": 5.56142, "mean_abs": 0.082425, "relative": 0.751176 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.081872, "mean_abs": 0.01094, "relative": 0.193205 }, { "name": "camera_rotation_matrix", "l2": 3.002932, "mean_abs": 0.213459, "relative": 0.922152 }, { "name": "imu_accel_gyro", "l2": 17.289299, "mean_abs": 0.986538, "relative": 0.932817 }, { "name": "depth_confidence", "l2": 12.900332, "mean_abs": 0.22337, "relative": 0.000428 }, { "name": "video_fisheye_cam0", "l2": 5.175725, "mean_abs": 0.107617, "relative": 0.937213 }, { "name": "video_fisheye_cam1", "l2": 5.571677, "mean_abs": 0.118901, "relative": 0.920673 }, { "name": "video_fisheye_cam2", "l2": 5.515169, "mean_abs": 0.114045, "relative": 0.880415 }, { "name": "video_fisheye_cam3", "l2": 5.59567, "mean_abs": 0.114906, "relative": 0.936927 }, { "name": "video_stereo_left", "l2": 5.247674, "mean_abs": 0.116237, "relative": 0.852758 }, { "name": "video_stereo_right", "l2": 6.279363, "mean_abs": 0.141938, "relative": 0.885258 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.277477, "mean_abs": 0.09127, "relative": 0.815075 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.5210253000259399" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.5790156722068787" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999998807907104" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.6346155405044556" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|bottle|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 983, "start_frame": 4915, "end_frame": 4934, "center_frame": 4924, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.02037, "mean_abs": 0.207417, "relative": 0.845261 }, { "name": "hand_right_joints", "l2": 10.364639, "mean_abs": 0.240728, "relative": 0.817011 }, { "name": "body_joints", "l2": 5.580991, "mean_abs": 0.082345, "relative": 0.75382 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.036475, "mean_abs": 0.004487, "relative": 0.086077 }, { "name": "camera_rotation_matrix", "l2": 3.001555, "mean_abs": 0.212197, "relative": 0.921729 }, { "name": "imu_accel_gyro", "l2": 17.287088, "mean_abs": 0.965395, "relative": 0.932698 }, { "name": "depth_confidence", "l2": 12.901099, "mean_abs": 0.223715, "relative": 0.000428 }, { "name": "video_fisheye_cam0", "l2": 5.177929, "mean_abs": 0.107412, "relative": 0.937612 }, { "name": "video_fisheye_cam1", "l2": 5.595378, "mean_abs": 0.119465, "relative": 0.924589 }, { "name": "video_fisheye_cam2", "l2": 5.501012, "mean_abs": 0.114046, "relative": 0.878155 }, { "name": "video_fisheye_cam3", "l2": 5.579662, "mean_abs": 0.11438, "relative": 0.934247 }, { "name": "video_stereo_left", "l2": 5.248065, "mean_abs": 0.116862, "relative": 0.852821 }, { "name": "video_stereo_right", "l2": 6.296395, "mean_abs": 0.143076, "relative": 0.88766 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.424901, "mean_abs": 0.099702, "relative": 0.867836 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.4961884021759033" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Handle gooseneck kettle", "correct": 0, "confidence": "0.5943541526794434" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Hold gooseneck kettle", "correct": 0, "confidence": "0.73974609375" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 984, "start_frame": 4920, "end_frame": 4939, "center_frame": 4929, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.995502, "mean_abs": 0.208883, "relative": 0.843163 }, { "name": "hand_right_joints", "l2": 10.451067, "mean_abs": 0.250541, "relative": 0.823824 }, { "name": "body_joints", "l2": 5.68399, "mean_abs": 0.086204, "relative": 0.767732 }, { "name": "body_contacts", "l2": 3.464234, "mean_abs": 0.081713, "relative": 0.999663 }, { "name": "camera_translation", "l2": 0.09603, "mean_abs": 0.013004, "relative": 0.226618 }, { "name": "camera_rotation_matrix", "l2": 3.003486, "mean_abs": 0.211717, "relative": 0.922322 }, { "name": "imu_accel_gyro", "l2": 17.245245, "mean_abs": 0.969424, "relative": 0.93044 }, { "name": "depth_confidence", "l2": 12.823435, "mean_abs": 0.22367, "relative": 0.000426 }, { "name": "video_fisheye_cam0", "l2": 5.18546, "mean_abs": 0.107831, "relative": 0.938976 }, { "name": "video_fisheye_cam1", "l2": 5.594472, "mean_abs": 0.119182, "relative": 0.924439 }, { "name": "video_fisheye_cam2", "l2": 5.476459, "mean_abs": 0.114013, "relative": 0.874235 }, { "name": "video_fisheye_cam3", "l2": 5.61735, "mean_abs": 0.115184, "relative": 0.940557 }, { "name": "video_stereo_left", "l2": 5.24403, "mean_abs": 0.118802, "relative": 0.852165 }, { "name": "video_stereo_right", "l2": 6.290477, "mean_abs": 0.144484, "relative": 0.886825 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.998562, "mean_abs": 0.078377, "relative": 0.715255 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.9666850566864014" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.8124080896377563" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999997615814209" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.9856481552124023" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 985, "start_frame": 4925, "end_frame": 4944, "center_frame": 4934, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.93494, "mean_abs": 0.21249, "relative": 0.838055 }, { "name": "hand_right_joints", "l2": 10.82746, "mean_abs": 0.278826, "relative": 0.853494 }, { "name": "body_joints", "l2": 5.943273, "mean_abs": 0.092995, "relative": 0.802753 }, { "name": "body_contacts", "l2": 3.464209, "mean_abs": 0.081721, "relative": 0.999656 }, { "name": "camera_translation", "l2": 0.217874, "mean_abs": 0.029032, "relative": 0.514152 }, { "name": "camera_rotation_matrix", "l2": 3.005044, "mean_abs": 0.212177, "relative": 0.9228 }, { "name": "imu_accel_gyro", "l2": 17.12645, "mean_abs": 0.995755, "relative": 0.924031 }, { "name": "depth_confidence", "l2": 12.835996, "mean_abs": 0.223186, "relative": 0.000426 }, { "name": "video_fisheye_cam0", "l2": 5.187805, "mean_abs": 0.108362, "relative": 0.9394 }, { "name": "video_fisheye_cam1", "l2": 5.55197, "mean_abs": 0.119956, "relative": 0.917416 }, { "name": "video_fisheye_cam2", "l2": 5.441104, "mean_abs": 0.11421, "relative": 0.868591 }, { "name": "video_fisheye_cam3", "l2": 5.643736, "mean_abs": 0.117364, "relative": 0.944976 }, { "name": "video_stereo_left", "l2": 5.229431, "mean_abs": 0.120715, "relative": 0.849793 }, { "name": "video_stereo_right", "l2": 6.268933, "mean_abs": 0.145808, "relative": 0.883788 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.989357, "mean_abs": 0.077909, "relative": 0.711961 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.9816434979438782" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "0.941862165927887" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.9982447624206543" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|scale|gooseneck kettle|water bottle|coffee mug|white cup", "correct": 0, "confidence": "" } } }, { "window_index": 986, "start_frame": 4930, "end_frame": 4949, "center_frame": 4939, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.900598, "mean_abs": 0.217639, "relative": 0.835158 }, { "name": "hand_right_joints", "l2": 11.230263, "mean_abs": 0.296467, "relative": 0.885245 }, { "name": "body_joints", "l2": 6.107731, "mean_abs": 0.099613, "relative": 0.824966 }, { "name": "body_contacts", "l2": 3.465402, "mean_abs": 0.084767, "relative": 1.0 }, { "name": "camera_translation", "l2": 0.334795, "mean_abs": 0.045125, "relative": 0.790068 }, { "name": "camera_rotation_matrix", "l2": 3.014086, "mean_abs": 0.216405, "relative": 0.925577 }, { "name": "imu_accel_gyro", "l2": 17.106632, "mean_abs": 1.07644, "relative": 0.922962 }, { "name": "depth_confidence", "l2": 11.877414, "mean_abs": 0.21851, "relative": 0.000394 }, { "name": "video_fisheye_cam0", "l2": 5.202053, "mean_abs": 0.109566, "relative": 0.94198 }, { "name": "video_fisheye_cam1", "l2": 5.491441, "mean_abs": 0.121042, "relative": 0.907414 }, { "name": "video_fisheye_cam2", "l2": 5.436595, "mean_abs": 0.114983, "relative": 0.867872 }, { "name": "video_fisheye_cam3", "l2": 5.676638, "mean_abs": 0.119542, "relative": 0.950485 }, { "name": "video_stereo_left", "l2": 5.23824, "mean_abs": 0.122448, "relative": 0.851224 }, { "name": "video_stereo_right", "l2": 6.261093, "mean_abs": 0.147894, "relative": 0.882683 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.149495, "mean_abs": 0.090496, "relative": 0.769272 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.9931026697158813" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9913145899772644" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Pick up white bottle", "correct": 0, "confidence": "1.0" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|bottle|gooseneck kettle|digital scale|table|coffee filter|glass carafe|coffee scale|wooden scoop|mug", "correct": 0, "confidence": "" } } }, { "window_index": 987, "start_frame": 4935, "end_frame": 4954, "center_frame": 4944, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.909907, "mean_abs": 0.225588, "relative": 0.835943 }, { "name": "hand_right_joints", "l2": 11.57648, "mean_abs": 0.307064, "relative": 0.912536 }, { "name": "body_joints", "l2": 6.153392, "mean_abs": 0.100578, "relative": 0.831134 }, { "name": "body_contacts", "l2": 3.325772, "mean_abs": 0.078275, "relative": 0.959707 }, { "name": "camera_translation", "l2": 0.378627, "mean_abs": 0.052029, "relative": 0.893506 }, { "name": "camera_rotation_matrix", "l2": 3.038581, "mean_abs": 0.226761, "relative": 0.933099 }, { "name": "imu_accel_gyro", "l2": 17.599777, "mean_abs": 1.231897, "relative": 0.949569 }, { "name": "depth_confidence", "l2": 11.612531, "mean_abs": 0.21649, "relative": 0.000385 }, { "name": "video_fisheye_cam0", "l2": 5.223042, "mean_abs": 0.111814, "relative": 0.945781 }, { "name": "video_fisheye_cam1", "l2": 5.4314, "mean_abs": 0.122452, "relative": 0.897493 }, { "name": "video_fisheye_cam2", "l2": 5.455667, "mean_abs": 0.116393, "relative": 0.870916 }, { "name": "video_fisheye_cam3", "l2": 5.67011, "mean_abs": 0.121168, "relative": 0.949391 }, { "name": "video_stereo_left", "l2": 5.298054, "mean_abs": 0.125254, "relative": 0.860944 }, { "name": "video_stereo_right", "l2": 6.321256, "mean_abs": 0.151418, "relative": 0.891164 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.166947, "mean_abs": 0.090108, "relative": 0.775518 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Position kettle to pour", "correct": 0, "confidence": "1.0" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9999996423721313" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Position kettle to pour", "correct": 0, "confidence": "1.0" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "gooseneck kettle|digital scale|table|dripper|coffee filter|glass carafe|wooden scoop|closed coffee container|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 988, "start_frame": 4940, "end_frame": 4959, "center_frame": 4949, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.934715, "mean_abs": 0.230998, "relative": 0.838036 }, { "name": "hand_right_joints", "l2": 11.885172, "mean_abs": 0.314063, "relative": 0.93687 }, { "name": "body_joints", "l2": 6.177915, "mean_abs": 0.097956, "relative": 0.834446 }, { "name": "body_contacts", "l2": 3.32577, "mean_abs": 0.078295, "relative": 0.959707 }, { "name": "camera_translation", "l2": 0.322986, "mean_abs": 0.045277, "relative": 0.7622 }, { "name": "camera_rotation_matrix", "l2": 3.038897, "mean_abs": 0.229192, "relative": 0.933196 }, { "name": "imu_accel_gyro", "l2": 17.40167, "mean_abs": 1.236495, "relative": 0.93888 }, { "name": "depth_confidence", "l2": 11.207504, "mean_abs": 0.212684, "relative": 0.000372 }, { "name": "video_fisheye_cam0", "l2": 5.235318, "mean_abs": 0.112613, "relative": 0.948004 }, { "name": "video_fisheye_cam1", "l2": 5.324853, "mean_abs": 0.120384, "relative": 0.879887 }, { "name": "video_fisheye_cam2", "l2": 5.487627, "mean_abs": 0.11723, "relative": 0.876018 }, { "name": "video_fisheye_cam3", "l2": 5.635613, "mean_abs": 0.120675, "relative": 0.943615 }, { "name": "video_stereo_left", "l2": 5.28651, "mean_abs": 0.125338, "relative": 0.859068 }, { "name": "video_stereo_right", "l2": 6.254805, "mean_abs": 0.149668, "relative": 0.881796 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.122764, "mean_abs": 0.087828, "relative": 0.759705 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014063, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Position kettle to pour", "correct": 0, "confidence": "1.0" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9999998807907104" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Position kettle to pour", "correct": 0, "confidence": "1.0" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "gooseneck kettle|digital scale|table|dripper|coffee filter|glass carafe|wooden scoop|closed coffee container|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 989, "start_frame": 4945, "end_frame": 4964, "center_frame": 4954, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.960426, "mean_abs": 0.231368, "relative": 0.840204 }, { "name": "hand_right_joints", "l2": 12.13858, "mean_abs": 0.312806, "relative": 0.956845 }, { "name": "body_joints", "l2": 6.097405, "mean_abs": 0.093672, "relative": 0.823572 }, { "name": "body_contacts", "l2": 3.325785, "mean_abs": 0.078296, "relative": 0.959711 }, { "name": "camera_translation", "l2": 0.208959, "mean_abs": 0.028489, "relative": 0.493113 }, { "name": "camera_rotation_matrix", "l2": 3.021585, "mean_abs": 0.223146, "relative": 0.92788 }, { "name": "imu_accel_gyro", "l2": 17.336903, "mean_abs": 1.206374, "relative": 0.935386 }, { "name": "depth_confidence", "l2": 10.877168, "mean_abs": 0.207535, "relative": 0.000361 }, { "name": "video_fisheye_cam0", "l2": 5.240192, "mean_abs": 0.111856, "relative": 0.948887 }, { "name": "video_fisheye_cam1", "l2": 5.200595, "mean_abs": 0.116655, "relative": 0.859354 }, { "name": "video_fisheye_cam2", "l2": 5.499839, "mean_abs": 0.116349, "relative": 0.877967 }, { "name": "video_fisheye_cam3", "l2": 5.572639, "mean_abs": 0.118116, "relative": 0.933071 }, { "name": "video_stereo_left", "l2": 5.237951, "mean_abs": 0.122068, "relative": 0.851177 }, { "name": "video_stereo_right", "l2": 6.158538, "mean_abs": 0.145781, "relative": 0.868225 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.073383, "mean_abs": 0.082762, "relative": 0.742033 }, { "name": "caption_objects_interaction_text", "l2": 1.900721, "mean_abs": 0.019049, "relative": 0.872765 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Position kettle to pour", "correct": 0, "confidence": "1.0" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9999997615814209" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Position kettle to pour", "correct": 0, "confidence": "1.0" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "gooseneck kettle|digital scale|table|dripper|coffee filter|glass carafe|wooden scoop|closed coffee container|water bottle", "correct": 0, "confidence": "" } } }, { "window_index": 990, "start_frame": 4950, "end_frame": 4969, "center_frame": 4959, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.983652, "mean_abs": 0.228757, "relative": 0.842164 }, { "name": "hand_right_joints", "l2": 12.353632, "mean_abs": 0.304253, "relative": 0.973797 }, { "name": "body_joints", "l2": 5.966191, "mean_abs": 0.089408, "relative": 0.805849 }, { "name": "body_contacts", "l2": 3.464388, "mean_abs": 0.08172, "relative": 0.999707 }, { "name": "camera_translation", "l2": 0.123062, "mean_abs": 0.019011, "relative": 0.290409 }, { "name": "camera_rotation_matrix", "l2": 3.008069, "mean_abs": 0.215183, "relative": 0.923729 }, { "name": "imu_accel_gyro", "l2": 17.350433, "mean_abs": 1.156275, "relative": 0.936116 }, { "name": "depth_confidence", "l2": 10.601166, "mean_abs": 0.201289, "relative": 0.000352 }, { "name": "video_fisheye_cam0", "l2": 5.251605, "mean_abs": 0.110765, "relative": 0.950953 }, { "name": "video_fisheye_cam1", "l2": 5.099525, "mean_abs": 0.111756, "relative": 0.842653 }, { "name": "video_fisheye_cam2", "l2": 5.539223, "mean_abs": 0.115166, "relative": 0.884255 }, { "name": "video_fisheye_cam3", "l2": 5.533161, "mean_abs": 0.115711, "relative": 0.926461 }, { "name": "video_stereo_left", "l2": 5.249288, "mean_abs": 0.118988, "relative": 0.85302 }, { "name": "video_stereo_right", "l2": 6.110577, "mean_abs": 0.141399, "relative": 0.861463 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.236207, "mean_abs": 0.094498, "relative": 0.800305 }, { "name": "caption_objects_interaction_text", "l2": 1.900721, "mean_abs": 0.019182, "relative": 0.872765 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.48497480154037476" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.941844642162323" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.9106757640838623" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9978758096694946" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "scale|dripper", "correct": 0, "confidence": "" } } }, { "window_index": 991, "start_frame": 4955, "end_frame": 4974, "center_frame": 4964, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.043891, "mean_abs": 0.227035, "relative": 0.847245 }, { "name": "hand_right_joints", "l2": 12.38505, "mean_abs": 0.298532, "relative": 0.976273 }, { "name": "body_joints", "l2": 5.803541, "mean_abs": 0.083305, "relative": 0.78388 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.151308, "mean_abs": 0.019365, "relative": 0.357065 }, { "name": "camera_rotation_matrix", "l2": 3.002975, "mean_abs": 0.208902, "relative": 0.922165 }, { "name": "imu_accel_gyro", "l2": 17.400566, "mean_abs": 1.166327, "relative": 0.93882 }, { "name": "depth_confidence", "l2": 10.537827, "mean_abs": 0.198864, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.268039, "mean_abs": 0.109662, "relative": 0.953929 }, { "name": "video_fisheye_cam1", "l2": 5.073222, "mean_abs": 0.10903, "relative": 0.838307 }, { "name": "video_fisheye_cam2", "l2": 5.584197, "mean_abs": 0.114115, "relative": 0.891434 }, { "name": "video_fisheye_cam3", "l2": 5.51811, "mean_abs": 0.114617, "relative": 0.923941 }, { "name": "video_stereo_left", "l2": 5.299768, "mean_abs": 0.117345, "relative": 0.861223 }, { "name": "video_stereo_right", "l2": 6.164523, "mean_abs": 0.139237, "relative": 0.869068 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.207845, "mean_abs": 0.09264, "relative": 0.790155 }, { "name": "caption_objects_interaction_text", "l2": 1.900721, "mean_abs": 0.019154, "relative": 0.872765 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.5992944836616516" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.6315953135490417" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.9731902480125427" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9651735424995422" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "scale|dripper", "correct": 0, "confidence": "" } } }, { "window_index": 992, "start_frame": 4960, "end_frame": 4979, "center_frame": 4969, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.07399, "mean_abs": 0.228039, "relative": 0.849784 }, { "name": "hand_right_joints", "l2": 12.23607, "mean_abs": 0.304179, "relative": 0.96453 }, { "name": "body_joints", "l2": 5.84657, "mean_abs": 0.084555, "relative": 0.789691 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.18209, "mean_abs": 0.021985, "relative": 0.429707 }, { "name": "camera_rotation_matrix", "l2": 3.005244, "mean_abs": 0.210855, "relative": 0.922862 }, { "name": "imu_accel_gyro", "l2": 17.443302, "mean_abs": 1.195381, "relative": 0.941126 }, { "name": "depth_confidence", "l2": 10.448615, "mean_abs": 0.197899, "relative": 0.000347 }, { "name": "video_fisheye_cam0", "l2": 5.282058, "mean_abs": 0.110603, "relative": 0.956468 }, { "name": "video_fisheye_cam1", "l2": 5.092758, "mean_abs": 0.110955, "relative": 0.841535 }, { "name": "video_fisheye_cam2", "l2": 5.614115, "mean_abs": 0.11515, "relative": 0.89621 }, { "name": "video_fisheye_cam3", "l2": 5.5994, "mean_abs": 0.116908, "relative": 0.937552 }, { "name": "video_stereo_left", "l2": 5.394604, "mean_abs": 0.120243, "relative": 0.876634 }, { "name": "video_stereo_right", "l2": 6.24715, "mean_abs": 0.142323, "relative": 0.880717 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.478845, "mean_abs": 0.106582, "relative": 0.887141 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014661, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9233120083808899" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.40814462304115295" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999308586120605" }, "next_action": { "true": "Wait/Prepare for pouring", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.988637387752533" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "kettle|scale|dripper", "correct": 0, "confidence": "" } } }, { "window_index": 993, "start_frame": 4965, "end_frame": 4984, "center_frame": 4974, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.056147, "mean_abs": 0.229354, "relative": 0.848279 }, { "name": "hand_right_joints", "l2": 12.050174, "mean_abs": 0.30854, "relative": 0.949876 }, { "name": "body_joints", "l2": 5.89767, "mean_abs": 0.086275, "relative": 0.796593 }, { "name": "body_contacts", "l2": 3.464573, "mean_abs": 0.081719, "relative": 0.999761 }, { "name": "camera_translation", "l2": 0.196316, "mean_abs": 0.02699, "relative": 0.463277 }, { "name": "camera_rotation_matrix", "l2": 3.01167, "mean_abs": 0.215375, "relative": 0.924835 }, { "name": "imu_accel_gyro", "l2": 17.086943, "mean_abs": 1.147772, "relative": 0.921899 }, { "name": "depth_confidence", "l2": 10.355989, "mean_abs": 0.200186, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.294479, "mean_abs": 0.111738, "relative": 0.958717 }, { "name": "video_fisheye_cam1", "l2": 5.181426, "mean_abs": 0.11587, "relative": 0.856187 }, { "name": "video_fisheye_cam2", "l2": 5.649765, "mean_abs": 0.118481, "relative": 0.901901 }, { "name": "video_fisheye_cam3", "l2": 5.701356, "mean_abs": 0.119964, "relative": 0.954623 }, { "name": "video_stereo_left", "l2": 5.605504, "mean_abs": 0.127478, "relative": 0.910905 }, { "name": "video_stereo_right", "l2": 6.504674, "mean_abs": 0.151246, "relative": 0.917023 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.295016, "mean_abs": 0.097811, "relative": 0.821352 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014661, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9438009858131409" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.7562705278396606" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999998807907104" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9789952635765076" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "scale|gooseneck kettle|dripper", "correct": 0, "confidence": "" } } }, { "window_index": 994, "start_frame": 4970, "end_frame": 4989, "center_frame": 4979, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.047019, "mean_abs": 0.226945, "relative": 0.847509 }, { "name": "hand_right_joints", "l2": 12.017894, "mean_abs": 0.314126, "relative": 0.947332 }, { "name": "body_joints", "l2": 5.954252, "mean_abs": 0.087506, "relative": 0.804236 }, { "name": "body_contacts", "l2": 3.464572, "mean_abs": 0.081704, "relative": 0.999761 }, { "name": "camera_translation", "l2": 0.228839, "mean_abs": 0.032139, "relative": 0.540028 }, { "name": "camera_rotation_matrix", "l2": 3.018591, "mean_abs": 0.217233, "relative": 0.92696 }, { "name": "imu_accel_gyro", "l2": 17.1392, "mean_abs": 1.140105, "relative": 0.924719 }, { "name": "depth_confidence", "l2": 10.323949, "mean_abs": 0.197473, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.300563, "mean_abs": 0.112613, "relative": 0.959819 }, { "name": "video_fisheye_cam1", "l2": 5.288485, "mean_abs": 0.119675, "relative": 0.873877 }, { "name": "video_fisheye_cam2", "l2": 5.685575, "mean_abs": 0.120761, "relative": 0.907617 }, { "name": "video_fisheye_cam3", "l2": 5.795887, "mean_abs": 0.12241, "relative": 0.970451 }, { "name": "video_stereo_left", "l2": 5.744362, "mean_abs": 0.132157, "relative": 0.93347 }, { "name": "video_stereo_right", "l2": 6.698051, "mean_abs": 0.156651, "relative": 0.944285 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.449879, "mean_abs": 0.107772, "relative": 0.876775 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014661, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.6903070211410522" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.6166724562644958" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.7701170444488525" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 995, "start_frame": 4975, "end_frame": 4994, "center_frame": 4984, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.028895, "mean_abs": 0.222812, "relative": 0.84598 }, { "name": "hand_right_joints", "l2": 11.948133, "mean_abs": 0.314373, "relative": 0.941833 }, { "name": "body_joints", "l2": 6.0302, "mean_abs": 0.088021, "relative": 0.814494 }, { "name": "body_contacts", "l2": 3.464572, "mean_abs": 0.081704, "relative": 0.999761 }, { "name": "camera_translation", "l2": 0.234839, "mean_abs": 0.033181, "relative": 0.554188 }, { "name": "camera_rotation_matrix", "l2": 3.017139, "mean_abs": 0.214538, "relative": 0.926515 }, { "name": "imu_accel_gyro", "l2": 17.159279, "mean_abs": 1.120077, "relative": 0.925802 }, { "name": "depth_confidence", "l2": 10.185309, "mean_abs": 0.194103, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.298948, "mean_abs": 0.112594, "relative": 0.959526 }, { "name": "video_fisheye_cam1", "l2": 5.316763, "mean_abs": 0.120184, "relative": 0.87855 }, { "name": "video_fisheye_cam2", "l2": 5.702392, "mean_abs": 0.121423, "relative": 0.910302 }, { "name": "video_fisheye_cam3", "l2": 5.830888, "mean_abs": 0.122425, "relative": 0.976312 }, { "name": "video_stereo_left", "l2": 5.799541, "mean_abs": 0.133715, "relative": 0.942437 }, { "name": "video_stereo_right", "l2": 6.779993, "mean_abs": 0.158393, "relative": 0.955837 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.517509, "mean_abs": 0.106892, "relative": 0.900979 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014661, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.9340563416481018" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.4530346095561981" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.581606388092041" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper|mug", "correct": 0, "confidence": "" } } }, { "window_index": 996, "start_frame": 4980, "end_frame": 4999, "center_frame": 4989, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.016732, "mean_abs": 0.219139, "relative": 0.844954 }, { "name": "hand_right_joints", "l2": 11.752806, "mean_abs": 0.308767, "relative": 0.926436 }, { "name": "body_joints", "l2": 6.105004, "mean_abs": 0.086429, "relative": 0.824598 }, { "name": "body_contacts", "l2": 3.464572, "mean_abs": 0.081704, "relative": 0.999761 }, { "name": "camera_translation", "l2": 0.200608, "mean_abs": 0.029026, "relative": 0.473406 }, { "name": "camera_rotation_matrix", "l2": 3.012641, "mean_abs": 0.209485, "relative": 0.925133 }, { "name": "imu_accel_gyro", "l2": 17.175756, "mean_abs": 1.114842, "relative": 0.926691 }, { "name": "depth_confidence", "l2": 10.053959, "mean_abs": 0.189912, "relative": 0.000334 }, { "name": "video_fisheye_cam0", "l2": 5.293969, "mean_abs": 0.112093, "relative": 0.958625 }, { "name": "video_fisheye_cam1", "l2": 5.318938, "mean_abs": 0.119129, "relative": 0.87891 }, { "name": "video_fisheye_cam2", "l2": 5.699265, "mean_abs": 0.120461, "relative": 0.909803 }, { "name": "video_fisheye_cam3", "l2": 5.839766, "mean_abs": 0.121847, "relative": 0.977798 }, { "name": "video_stereo_left", "l2": 5.845384, "mean_abs": 0.132734, "relative": 0.949887 }, { "name": "video_stereo_right", "l2": 6.830504, "mean_abs": 0.157714, "relative": 0.962958 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.421172, "mean_abs": 0.10393, "relative": 0.866501 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.014661, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.9863518476486206" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7479261159896851" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.8374666571617126" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "glass carafe|water bottle|white coffee cup|digital scale with dripper|metal pitcher", "predicted": "coffee dripper", "correct": 0, "confidence": "" } } }, { "window_index": 997, "start_frame": 4985, "end_frame": 5004, "center_frame": 4994, "action": "Wait/Prepare for pouring", "subtask": "Prepare for pouring", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe", "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.003341, "mean_abs": 0.215164, "relative": 0.843824 }, { "name": "hand_right_joints", "l2": 11.776004, "mean_abs": 0.311267, "relative": 0.928264 }, { "name": "body_joints", "l2": 6.166604, "mean_abs": 0.088033, "relative": 0.832918 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.116523, "mean_abs": 0.017746, "relative": 0.274977 }, { "name": "camera_rotation_matrix", "l2": 3.003351, "mean_abs": 0.200361, "relative": 0.92228 }, { "name": "imu_accel_gyro", "l2": 17.15374, "mean_abs": 1.091389, "relative": 0.925503 }, { "name": "depth_confidence", "l2": 9.94834, "mean_abs": 0.185359, "relative": 0.00033 }, { "name": "video_fisheye_cam0", "l2": 5.280931, "mean_abs": 0.110486, "relative": 0.956264 }, { "name": "video_fisheye_cam1", "l2": 5.317735, "mean_abs": 0.115755, "relative": 0.878711 }, { "name": "video_fisheye_cam2", "l2": 5.681481, "mean_abs": 0.117493, "relative": 0.906964 }, { "name": "video_fisheye_cam3", "l2": 5.833198, "mean_abs": 0.120054, "relative": 0.976699 }, { "name": "video_stereo_left", "l2": 5.850017, "mean_abs": 0.127824, "relative": 0.950639 }, { "name": "video_stereo_right", "l2": 6.898036, "mean_abs": 0.152988, "relative": 0.972478 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.469829, "mean_abs": 0.104682, "relative": 0.883915 }, { "name": "caption_objects_interaction_text", "l2": 2.101556, "mean_abs": 0.024132, "relative": 0.964984 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Wait/Prepare for pouring", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.9919301867485046" }, "timeline_subtask": { "true": "Prepare for pouring", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.972639262676239" }, "transition_detection": { "true": "transition", "predicted": "steady", "correct": 0, "confidence": "0.9997796416282654" }, "next_action": { "true": "Pour coffee", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.8546449542045593" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|glass carafe|water bottle|coffee mug|white coffee cup|digital scale with dripper|metal pitcher|carafe", "predicted": "coffee dripper", "correct": 0, "confidence": "" } } }, { "window_index": 998, "start_frame": 4990, "end_frame": 5009, "center_frame": 4999, "action": "", "subtask": "", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe", "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.985271, "mean_abs": 0.213409, "relative": 0.8423 }, { "name": "hand_right_joints", "l2": 11.829371, "mean_abs": 0.312346, "relative": 0.932471 }, { "name": "body_joints", "l2": 6.218741, "mean_abs": 0.089356, "relative": 0.83996 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.03863, "mean_abs": 0.005938, "relative": 0.091162 }, { "name": "camera_rotation_matrix", "l2": 3.000514, "mean_abs": 0.192977, "relative": 0.921409 }, { "name": "imu_accel_gyro", "l2": 17.134232, "mean_abs": 1.050976, "relative": 0.924451 }, { "name": "depth_confidence", "l2": 9.767466, "mean_abs": 0.180649, "relative": 0.000324 }, { "name": "video_fisheye_cam0", "l2": 5.277409, "mean_abs": 0.10925, "relative": 0.955626 }, { "name": "video_fisheye_cam1", "l2": 5.331199, "mean_abs": 0.112873, "relative": 0.880936 }, { "name": "video_fisheye_cam2", "l2": 5.676946, "mean_abs": 0.115064, "relative": 0.90624 }, { "name": "video_fisheye_cam3", "l2": 5.821836, "mean_abs": 0.118047, "relative": 0.974796 }, { "name": "video_stereo_left", "l2": 5.82509, "mean_abs": 0.123497, "relative": 0.946589 }, { "name": "video_stereo_right", "l2": 6.897284, "mean_abs": 0.148498, "relative": 0.972372 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.530928, "mean_abs": 0.109258, "relative": 0.905781 }, { "name": "caption_objects_interaction_text", "l2": 2.101556, "mean_abs": 0.024225, "relative": 0.964984 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": null, "transition_detection": { "true": "transition", "predicted": "steady", "correct": 0, "confidence": "0.9977325201034546" }, "next_action": { "true": "Pour coffee", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.6717064380645752" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|glass carafe|water bottle|coffee mug|white coffee cup|digital scale with dripper|metal pitcher|carafe", "predicted": "coffee dripper", "correct": 0, "confidence": "" } } }, { "window_index": 999, "start_frame": 4995, "end_frame": 5014, "center_frame": 5004, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "white coffee cup", "digital scale with dripper", "water bottle", "metal pitcher", "glass carafe", "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.94593, "mean_abs": 0.212653, "relative": 0.838982 }, { "name": "hand_right_joints", "l2": 11.842458, "mean_abs": 0.311074, "relative": 0.933502 }, { "name": "body_joints", "l2": 6.2838, "mean_abs": 0.089771, "relative": 0.848748 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.040453, "mean_abs": 0.005912, "relative": 0.095463 }, { "name": "camera_rotation_matrix", "l2": 3.000954, "mean_abs": 0.194982, "relative": 0.921544 }, { "name": "imu_accel_gyro", "l2": 17.073849, "mean_abs": 1.034173, "relative": 0.921193 }, { "name": "depth_confidence", "l2": 9.687142, "mean_abs": 0.1803, "relative": 0.000321 }, { "name": "video_fisheye_cam0", "l2": 5.278766, "mean_abs": 0.109842, "relative": 0.955872 }, { "name": "video_fisheye_cam1", "l2": 5.328775, "mean_abs": 0.11431, "relative": 0.880535 }, { "name": "video_fisheye_cam2", "l2": 5.669924, "mean_abs": 0.116057, "relative": 0.905119 }, { "name": "video_fisheye_cam3", "l2": 5.800533, "mean_abs": 0.118377, "relative": 0.971229 }, { "name": "video_stereo_left", "l2": 5.821178, "mean_abs": 0.124658, "relative": 0.945953 }, { "name": "video_stereo_right", "l2": 6.90675, "mean_abs": 0.150225, "relative": 0.973707 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.241565, "mean_abs": 0.089668, "relative": 0.802223 }, { "name": "caption_objects_interaction_text", "l2": 2.101556, "mean_abs": 0.02344, "relative": 0.964984 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.9142341017723083" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8726351857185364" }, "transition_detection": { "true": "transition", "predicted": "steady", "correct": 0, "confidence": "0.9997128844261169" }, "next_action": { "true": "Pour coffee", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.6332637071609497" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|glass carafe|water bottle|coffee mug|white coffee cup|digital scale with dripper|metal pitcher|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|coffee jar", "correct": 0, "confidence": "" } } }, { "window_index": 1000, "start_frame": 5000, "end_frame": 5019, "center_frame": 5009, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.918031, "mean_abs": 0.211863, "relative": 0.836628 }, { "name": "hand_right_joints", "l2": 11.780734, "mean_abs": 0.312599, "relative": 0.928637 }, { "name": "body_joints", "l2": 6.33017, "mean_abs": 0.09021, "relative": 0.855011 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.064301, "mean_abs": 0.009219, "relative": 0.15174 }, { "name": "camera_rotation_matrix", "l2": 3.002419, "mean_abs": 0.197413, "relative": 0.921994 }, { "name": "imu_accel_gyro", "l2": 17.081245, "mean_abs": 1.039107, "relative": 0.921592 }, { "name": "depth_confidence", "l2": 9.800895, "mean_abs": 0.182848, "relative": 0.000325 }, { "name": "video_fisheye_cam0", "l2": 5.279898, "mean_abs": 0.110228, "relative": 0.956077 }, { "name": "video_fisheye_cam1", "l2": 5.323931, "mean_abs": 0.115338, "relative": 0.879735 }, { "name": "video_fisheye_cam2", "l2": 5.662912, "mean_abs": 0.11669, "relative": 0.904 }, { "name": "video_fisheye_cam3", "l2": 5.77461, "mean_abs": 0.117726, "relative": 0.966889 }, { "name": "video_stereo_left", "l2": 5.823034, "mean_abs": 0.126226, "relative": 0.946255 }, { "name": "video_stereo_right", "l2": 6.898094, "mean_abs": 0.151762, "relative": 0.972486 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.098819, "mean_abs": 0.086658, "relative": 0.751136 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.5923903584480286" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6359732151031494" }, "transition_detection": { "true": "transition", "predicted": "steady", "correct": 0, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.36084404587745667" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|coffee jar", "correct": 0, "confidence": "" } } }, { "window_index": 1001, "start_frame": 5005, "end_frame": 5024, "center_frame": 5014, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.895008, "mean_abs": 0.211783, "relative": 0.834686 }, { "name": "hand_right_joints", "l2": 11.712176, "mean_abs": 0.312521, "relative": 0.923233 }, { "name": "body_joints", "l2": 6.359252, "mean_abs": 0.090385, "relative": 0.858939 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.07461, "mean_abs": 0.011002, "relative": 0.176069 }, { "name": "camera_rotation_matrix", "l2": 3.003216, "mean_abs": 0.198624, "relative": 0.922239 }, { "name": "imu_accel_gyro", "l2": 17.083975, "mean_abs": 1.044934, "relative": 0.921739 }, { "name": "depth_confidence", "l2": 9.845317, "mean_abs": 0.183861, "relative": 0.000327 }, { "name": "video_fisheye_cam0", "l2": 5.279157, "mean_abs": 0.110284, "relative": 0.955942 }, { "name": "video_fisheye_cam1", "l2": 5.318136, "mean_abs": 0.115764, "relative": 0.878777 }, { "name": "video_fisheye_cam2", "l2": 5.649886, "mean_abs": 0.116811, "relative": 0.90192 }, { "name": "video_fisheye_cam3", "l2": 5.746951, "mean_abs": 0.117233, "relative": 0.962258 }, { "name": "video_stereo_left", "l2": 5.802546, "mean_abs": 0.126325, "relative": 0.942925 }, { "name": "video_stereo_right", "l2": 6.876049, "mean_abs": 0.151842, "relative": 0.969379 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.053686, "mean_abs": 0.082037, "relative": 0.734984 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.5594492554664612" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Secure coffee container", "correct": 0, "confidence": "0.489734947681427" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.5437143445014954" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|dripper|coffee jar|coffee scoop", "correct": 0, "confidence": "" } } }, { "window_index": 1002, "start_frame": 5010, "end_frame": 5029, "center_frame": 5019, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.863111, "mean_abs": 0.210347, "relative": 0.831995 }, { "name": "hand_right_joints", "l2": 11.639527, "mean_abs": 0.309652, "relative": 0.917506 }, { "name": "body_joints", "l2": 6.376977, "mean_abs": 0.090197, "relative": 0.861333 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.071244, "mean_abs": 0.01049, "relative": 0.168125 }, { "name": "camera_rotation_matrix", "l2": 3.003701, "mean_abs": 0.199232, "relative": 0.922388 }, { "name": "imu_accel_gyro", "l2": 17.049513, "mean_abs": 1.036759, "relative": 0.91988 }, { "name": "depth_confidence", "l2": 9.920809, "mean_abs": 0.184792, "relative": 0.000329 }, { "name": "video_fisheye_cam0", "l2": 5.28038, "mean_abs": 0.11024, "relative": 0.956164 }, { "name": "video_fisheye_cam1", "l2": 5.320014, "mean_abs": 0.116167, "relative": 0.879088 }, { "name": "video_fisheye_cam2", "l2": 5.628098, "mean_abs": 0.116683, "relative": 0.898442 }, { "name": "video_fisheye_cam3", "l2": 5.739974, "mean_abs": 0.117419, "relative": 0.961089 }, { "name": "video_stereo_left", "l2": 5.743073, "mean_abs": 0.12544, "relative": 0.933261 }, { "name": "video_stereo_right", "l2": 6.829587, "mean_abs": 0.151285, "relative": 0.962828 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.062009, "mean_abs": 0.084117, "relative": 0.737962 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.7384365200996399" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Secure coffee container", "correct": 0, "confidence": "0.7505196928977966" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.5097517371177673" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|dripper|coffee jar|coffee scoop", "correct": 0, "confidence": "" } } }, { "window_index": 1003, "start_frame": 5015, "end_frame": 5034, "center_frame": 5024, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.845804, "mean_abs": 0.20998, "relative": 0.830536 }, { "name": "hand_right_joints", "l2": 11.588449, "mean_abs": 0.306966, "relative": 0.91348 }, { "name": "body_joints", "l2": 6.386503, "mean_abs": 0.090018, "relative": 0.86262 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.076878, "mean_abs": 0.01098, "relative": 0.18142 }, { "name": "camera_rotation_matrix", "l2": 3.004685, "mean_abs": 0.20094, "relative": 0.92269 }, { "name": "imu_accel_gyro", "l2": 17.07947, "mean_abs": 1.043481, "relative": 0.921496 }, { "name": "depth_confidence", "l2": 9.963711, "mean_abs": 0.186555, "relative": 0.000331 }, { "name": "video_fisheye_cam0", "l2": 5.281928, "mean_abs": 0.110505, "relative": 0.956444 }, { "name": "video_fisheye_cam1", "l2": 5.331142, "mean_abs": 0.11734, "relative": 0.880926 }, { "name": "video_fisheye_cam2", "l2": 5.612126, "mean_abs": 0.117079, "relative": 0.895893 }, { "name": "video_fisheye_cam3", "l2": 5.712934, "mean_abs": 0.117629, "relative": 0.956562 }, { "name": "video_stereo_left", "l2": 5.664432, "mean_abs": 0.125112, "relative": 0.920481 }, { "name": "video_stereo_right", "l2": 6.783094, "mean_abs": 0.151859, "relative": 0.956274 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.096964, "mean_abs": 0.086302, "relative": 0.750472 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.8374500870704651" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Secure coffee container", "correct": 0, "confidence": "0.8071506023406982" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.6181125640869141" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|dripper|coffee jar|coffee scoop", "correct": 0, "confidence": "" } } }, { "window_index": 1004, "start_frame": 5020, "end_frame": 5039, "center_frame": 5029, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.83558, "mean_abs": 0.210862, "relative": 0.829673 }, { "name": "hand_right_joints", "l2": 11.532908, "mean_abs": 0.306683, "relative": 0.909102 }, { "name": "body_joints", "l2": 6.371875, "mean_abs": 0.090796, "relative": 0.860644 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.094969, "mean_abs": 0.013359, "relative": 0.224114 }, { "name": "camera_rotation_matrix", "l2": 3.007736, "mean_abs": 0.202977, "relative": 0.923627 }, { "name": "imu_accel_gyro", "l2": 17.037104, "mean_abs": 1.062116, "relative": 0.91921 }, { "name": "depth_confidence", "l2": 9.967089, "mean_abs": 0.187732, "relative": 0.000331 }, { "name": "video_fisheye_cam0", "l2": 5.28265, "mean_abs": 0.110755, "relative": 0.956575 }, { "name": "video_fisheye_cam1", "l2": 5.355978, "mean_abs": 0.118455, "relative": 0.88503 }, { "name": "video_fisheye_cam2", "l2": 5.597921, "mean_abs": 0.117461, "relative": 0.893625 }, { "name": "video_fisheye_cam3", "l2": 5.679034, "mean_abs": 0.117792, "relative": 0.950886 }, { "name": "video_stereo_left", "l2": 5.605316, "mean_abs": 0.125215, "relative": 0.910875 }, { "name": "video_stereo_right", "l2": 6.736365, "mean_abs": 0.152354, "relative": 0.949686 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.101593, "mean_abs": 0.087776, "relative": 0.752128 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.8813374638557434" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Secure coffee container", "correct": 0, "confidence": "0.7873004078865051" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.6756475567817688" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|dripper|coffee scoop", "correct": 0, "confidence": "" } } }, { "window_index": 1005, "start_frame": 5025, "end_frame": 5044, "center_frame": 5034, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.847225, "mean_abs": 0.211276, "relative": 0.830655 }, { "name": "hand_right_joints", "l2": 11.414253, "mean_abs": 0.302934, "relative": 0.899749 }, { "name": "body_joints", "l2": 6.346279, "mean_abs": 0.091449, "relative": 0.857187 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.123586, "mean_abs": 0.01651, "relative": 0.291645 }, { "name": "camera_rotation_matrix", "l2": 3.010984, "mean_abs": 0.203199, "relative": 0.924624 }, { "name": "imu_accel_gyro", "l2": 16.988663, "mean_abs": 1.046613, "relative": 0.916597 }, { "name": "depth_confidence", "l2": 9.85639, "mean_abs": 0.187402, "relative": 0.000327 }, { "name": "video_fisheye_cam0", "l2": 5.290117, "mean_abs": 0.111152, "relative": 0.957927 }, { "name": "video_fisheye_cam1", "l2": 5.364401, "mean_abs": 0.118715, "relative": 0.886422 }, { "name": "video_fisheye_cam2", "l2": 5.580152, "mean_abs": 0.117735, "relative": 0.890788 }, { "name": "video_fisheye_cam3", "l2": 5.671467, "mean_abs": 0.118034, "relative": 0.949619 }, { "name": "video_stereo_left", "l2": 5.553844, "mean_abs": 0.125513, "relative": 0.902511 }, { "name": "video_stereo_right", "l2": 6.640542, "mean_abs": 0.151808, "relative": 0.936177 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.093518, "mean_abs": 0.086324, "relative": 0.749238 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.8651036024093628" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Transfer coffee grounds to dripper", "correct": 0, "confidence": "0.4944051504135132" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.8781424164772034" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|dripper|coffee scoop", "correct": 0, "confidence": "" } } }, { "window_index": 1006, "start_frame": 5030, "end_frame": 5049, "center_frame": 5039, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.866649, "mean_abs": 0.212498, "relative": 0.832294 }, { "name": "hand_right_joints", "l2": 11.378813, "mean_abs": 0.299355, "relative": 0.896955 }, { "name": "body_joints", "l2": 6.292439, "mean_abs": 0.090956, "relative": 0.849915 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.149546, "mean_abs": 0.019519, "relative": 0.352908 }, { "name": "camera_rotation_matrix", "l2": 3.009753, "mean_abs": 0.201988, "relative": 0.924246 }, { "name": "imu_accel_gyro", "l2": 16.986784, "mean_abs": 1.056429, "relative": 0.916496 }, { "name": "depth_confidence", "l2": 9.810822, "mean_abs": 0.186677, "relative": 0.000326 }, { "name": "video_fisheye_cam0", "l2": 5.293614, "mean_abs": 0.111071, "relative": 0.95856 }, { "name": "video_fisheye_cam1", "l2": 5.329391, "mean_abs": 0.117499, "relative": 0.880637 }, { "name": "video_fisheye_cam2", "l2": 5.548722, "mean_abs": 0.116907, "relative": 0.885771 }, { "name": "video_fisheye_cam3", "l2": 5.606917, "mean_abs": 0.11747, "relative": 0.938811 }, { "name": "video_stereo_left", "l2": 5.474306, "mean_abs": 0.123933, "relative": 0.889586 }, { "name": "video_stereo_right", "l2": 6.520749, "mean_abs": 0.149947, "relative": 0.919289 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.131186, "mean_abs": 0.087218, "relative": 0.762719 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.8665270209312439" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Transfer coffee grounds to dripper", "correct": 0, "confidence": "0.6343866586685181" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.795142412185669" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|dripper|coffee scoop", "correct": 0, "confidence": "" } } }, { "window_index": 1007, "start_frame": 5035, "end_frame": 5054, "center_frame": 5044, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.914786, "mean_abs": 0.216071, "relative": 0.836355 }, { "name": "hand_right_joints", "l2": 11.443085, "mean_abs": 0.302649, "relative": 0.902021 }, { "name": "body_joints", "l2": 6.252483, "mean_abs": 0.09083, "relative": 0.844518 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.151894, "mean_abs": 0.020761, "relative": 0.358447 }, { "name": "camera_rotation_matrix", "l2": 3.005903, "mean_abs": 0.200701, "relative": 0.923064 }, { "name": "imu_accel_gyro", "l2": 17.062752, "mean_abs": 1.086499, "relative": 0.920594 }, { "name": "depth_confidence", "l2": 9.787696, "mean_abs": 0.185284, "relative": 0.000325 }, { "name": "video_fisheye_cam0", "l2": 5.292398, "mean_abs": 0.110492, "relative": 0.95834 }, { "name": "video_fisheye_cam1", "l2": 5.301606, "mean_abs": 0.115305, "relative": 0.876046 }, { "name": "video_fisheye_cam2", "l2": 5.514974, "mean_abs": 0.115225, "relative": 0.880384 }, { "name": "video_fisheye_cam3", "l2": 5.529877, "mean_abs": 0.114899, "relative": 0.925911 }, { "name": "video_stereo_left", "l2": 5.368443, "mean_abs": 0.120239, "relative": 0.872383 }, { "name": "video_stereo_right", "l2": 6.411623, "mean_abs": 0.146578, "relative": 0.903904 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.062231, "mean_abs": 0.080804, "relative": 0.738041 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.8333640098571777" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Transfer coffee grounds to dripper", "correct": 0, "confidence": "0.6809888482093811" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.8256841897964478" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|dripper|coffee scoop", "correct": 0, "confidence": "" } } }, { "window_index": 1008, "start_frame": 5040, "end_frame": 5059, "center_frame": 5049, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.949874, "mean_abs": 0.219916, "relative": 0.839314 }, { "name": "hand_right_joints", "l2": 11.535238, "mean_abs": 0.30921, "relative": 0.909285 }, { "name": "body_joints", "l2": 6.218095, "mean_abs": 0.089734, "relative": 0.839873 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.130907, "mean_abs": 0.018564, "relative": 0.308922 }, { "name": "camera_rotation_matrix", "l2": 3.003775, "mean_abs": 0.197583, "relative": 0.922411 }, { "name": "imu_accel_gyro", "l2": 17.091372, "mean_abs": 1.113473, "relative": 0.922138 }, { "name": "depth_confidence", "l2": 9.75019, "mean_abs": 0.183609, "relative": 0.000324 }, { "name": "video_fisheye_cam0", "l2": 5.276907, "mean_abs": 0.109848, "relative": 0.955535 }, { "name": "video_fisheye_cam1", "l2": 5.274052, "mean_abs": 0.11293, "relative": 0.871493 }, { "name": "video_fisheye_cam2", "l2": 5.481325, "mean_abs": 0.11277, "relative": 0.875012 }, { "name": "video_fisheye_cam3", "l2": 5.485839, "mean_abs": 0.1137, "relative": 0.918538 }, { "name": "video_stereo_left", "l2": 5.274516, "mean_abs": 0.116184, "relative": 0.857119 }, { "name": "video_stereo_right", "l2": 6.296783, "mean_abs": 0.142295, "relative": 0.887714 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.162496, "mean_abs": 0.089078, "relative": 0.773925 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.8690798282623291" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Transfer coffee grounds to dripper", "correct": 0, "confidence": "0.768146276473999" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.6197906732559204" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|dripper|coffee scoop", "correct": 0, "confidence": "" } } }, { "window_index": 1009, "start_frame": 5045, "end_frame": 5064, "center_frame": 5054, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.976218, "mean_abs": 0.222162, "relative": 0.841537 }, { "name": "hand_right_joints", "l2": 11.650098, "mean_abs": 0.306565, "relative": 0.918339 }, { "name": "body_joints", "l2": 6.163421, "mean_abs": 0.087574, "relative": 0.832488 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.083015, "mean_abs": 0.011861, "relative": 0.195904 }, { "name": "camera_rotation_matrix", "l2": 3.001267, "mean_abs": 0.193206, "relative": 0.92164 }, { "name": "imu_accel_gyro", "l2": 17.155088, "mean_abs": 1.096286, "relative": 0.925576 }, { "name": "depth_confidence", "l2": 9.715292, "mean_abs": 0.182381, "relative": 0.000322 }, { "name": "video_fisheye_cam0", "l2": 5.273139, "mean_abs": 0.109513, "relative": 0.954853 }, { "name": "video_fisheye_cam1", "l2": 5.227947, "mean_abs": 0.111083, "relative": 0.863874 }, { "name": "video_fisheye_cam2", "l2": 5.478676, "mean_abs": 0.111486, "relative": 0.874589 }, { "name": "video_fisheye_cam3", "l2": 5.431867, "mean_abs": 0.112236, "relative": 0.909501 }, { "name": "video_stereo_left", "l2": 5.251493, "mean_abs": 0.114133, "relative": 0.853378 }, { "name": "video_stereo_right", "l2": 6.220473, "mean_abs": 0.13879, "relative": 0.876956 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.112935, "mean_abs": 0.082915, "relative": 0.756188 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.7094391584396362" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Transfer coffee grounds to dripper", "correct": 0, "confidence": "0.6615092158317566" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.5605292320251465" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|dripper|coffee scoop", "correct": 0, "confidence": "" } } }, { "window_index": 1010, "start_frame": 5050, "end_frame": 5069, "center_frame": 5059, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.998821, "mean_abs": 0.222218, "relative": 0.843443 }, { "name": "hand_right_joints", "l2": 11.809845, "mean_abs": 0.302522, "relative": 0.930932 }, { "name": "body_joints", "l2": 6.118846, "mean_abs": 0.086376, "relative": 0.826468 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.03539, "mean_abs": 0.005183, "relative": 0.083516 }, { "name": "camera_rotation_matrix", "l2": 3.000607, "mean_abs": 0.193269, "relative": 0.921438 }, { "name": "imu_accel_gyro", "l2": 17.170067, "mean_abs": 1.092357, "relative": 0.926384 }, { "name": "depth_confidence", "l2": 9.731405, "mean_abs": 0.182321, "relative": 0.000323 }, { "name": "video_fisheye_cam0", "l2": 5.268886, "mean_abs": 0.10929, "relative": 0.954083 }, { "name": "video_fisheye_cam1", "l2": 5.199617, "mean_abs": 0.109857, "relative": 0.859193 }, { "name": "video_fisheye_cam2", "l2": 5.503518, "mean_abs": 0.111856, "relative": 0.878555 }, { "name": "video_fisheye_cam3", "l2": 5.372317, "mean_abs": 0.110401, "relative": 0.89953 }, { "name": "video_stereo_left", "l2": 5.267824, "mean_abs": 0.113968, "relative": 0.856032 }, { "name": "video_stereo_right", "l2": 6.210114, "mean_abs": 0.137417, "relative": 0.875496 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.140792, "mean_abs": 0.082394, "relative": 0.766157 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.8641939759254456" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Transfer coffee grounds to dripper", "correct": 0, "confidence": "0.7283592820167542" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.3737025558948517" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|dripper", "correct": 0, "confidence": "" } } }, { "window_index": 1011, "start_frame": 5055, "end_frame": 5074, "center_frame": 5064, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.004762, "mean_abs": 0.221466, "relative": 0.843944 }, { "name": "hand_right_joints", "l2": 11.733344, "mean_abs": 0.308626, "relative": 0.924901 }, { "name": "body_joints", "l2": 6.103604, "mean_abs": 0.087664, "relative": 0.824409 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.04405, "mean_abs": 0.006787, "relative": 0.103952 }, { "name": "camera_rotation_matrix", "l2": 3.001806, "mean_abs": 0.198691, "relative": 0.921806 }, { "name": "imu_accel_gyro", "l2": 17.17893, "mean_abs": 1.124132, "relative": 0.926862 }, { "name": "depth_confidence", "l2": 9.883115, "mean_abs": 0.186505, "relative": 0.000328 }, { "name": "video_fisheye_cam0", "l2": 5.26755, "mean_abs": 0.10959, "relative": 0.953841 }, { "name": "video_fisheye_cam1", "l2": 5.206582, "mean_abs": 0.11118, "relative": 0.860344 }, { "name": "video_fisheye_cam2", "l2": 5.560265, "mean_abs": 0.11394, "relative": 0.887614 }, { "name": "video_fisheye_cam3", "l2": 5.330847, "mean_abs": 0.109271, "relative": 0.892586 }, { "name": "video_stereo_left", "l2": 5.327319, "mean_abs": 0.116751, "relative": 0.8657 }, { "name": "video_stereo_right", "l2": 6.258859, "mean_abs": 0.140641, "relative": 0.882368 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.110713, "mean_abs": 0.082856, "relative": 0.755392 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.9720942974090576" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Transfer coffee grounds to dripper", "correct": 0, "confidence": "0.7926531434059143" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999998807907104" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.5026437044143677" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "scale|dripper|coffee scoop", "correct": 0, "confidence": "" } } }, { "window_index": 1012, "start_frame": 5060, "end_frame": 5079, "center_frame": 5069, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.896688, "mean_abs": 0.225493, "relative": 0.834828 }, { "name": "hand_right_joints", "l2": 11.525375, "mean_abs": 0.305451, "relative": 0.908508 }, { "name": "body_joints", "l2": 6.043833, "mean_abs": 0.09393, "relative": 0.816336 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.107148, "mean_abs": 0.015647, "relative": 0.252855 }, { "name": "camera_rotation_matrix", "l2": 3.003091, "mean_abs": 0.202705, "relative": 0.922201 }, { "name": "imu_accel_gyro", "l2": 17.190376, "mean_abs": 1.13048, "relative": 0.92748 }, { "name": "depth_confidence", "l2": 10.017541, "mean_abs": 0.191286, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.273826, "mean_abs": 0.110295, "relative": 0.954977 }, { "name": "video_fisheye_cam1", "l2": 5.225204, "mean_abs": 0.112894, "relative": 0.863421 }, { "name": "video_fisheye_cam2", "l2": 5.601377, "mean_abs": 0.115884, "relative": 0.894176 }, { "name": "video_fisheye_cam3", "l2": 5.378522, "mean_abs": 0.111481, "relative": 0.900569 }, { "name": "video_stereo_left", "l2": 5.373105, "mean_abs": 0.119693, "relative": 0.87314 }, { "name": "video_stereo_right", "l2": 6.293847, "mean_abs": 0.144223, "relative": 0.8873 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.081911, "mean_abs": 0.080328, "relative": 0.745085 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.9865186810493469" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Transfer coffee grounds to dripper", "correct": 0, "confidence": "0.679445207118988" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.9206676483154297" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "scale|dripper|coffee scoop|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1013, "start_frame": 5065, "end_frame": 5084, "center_frame": 5074, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.110493, "mean_abs": 0.244556, "relative": 0.852863 }, { "name": "hand_right_joints", "l2": 11.424691, "mean_abs": 0.301656, "relative": 0.900571 }, { "name": "body_joints", "l2": 6.191575, "mean_abs": 0.102751, "relative": 0.836291 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.159197, "mean_abs": 0.02172, "relative": 0.375684 }, { "name": "camera_rotation_matrix", "l2": 3.003265, "mean_abs": 0.203949, "relative": 0.922254 }, { "name": "imu_accel_gyro", "l2": 16.93692, "mean_abs": 1.141621, "relative": 0.913805 }, { "name": "depth_confidence", "l2": 10.116007, "mean_abs": 0.194126, "relative": 0.000336 }, { "name": "video_fisheye_cam0", "l2": 5.275107, "mean_abs": 0.110456, "relative": 0.955209 }, { "name": "video_fisheye_cam1", "l2": 5.235598, "mean_abs": 0.113418, "relative": 0.865138 }, { "name": "video_fisheye_cam2", "l2": 5.62867, "mean_abs": 0.116521, "relative": 0.898533 }, { "name": "video_fisheye_cam3", "l2": 5.390899, "mean_abs": 0.112111, "relative": 0.902641 }, { "name": "video_stereo_left", "l2": 5.38315, "mean_abs": 0.121281, "relative": 0.874773 }, { "name": "video_stereo_right", "l2": 6.291988, "mean_abs": 0.14648, "relative": 0.887038 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.144498, "mean_abs": 0.082864, "relative": 0.767484 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.5520269870758057" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Prepare coffee equipment and scoop grounds", "correct": 0, "confidence": "0.9247586727142334" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Transfer coffee to dripper", "correct": 0, "confidence": "0.42440110445022583" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "scale|dripper|coffee scoop|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1014, "start_frame": 5070, "end_frame": 5089, "center_frame": 5079, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.582583, "mean_abs": 0.263018, "relative": 0.892686 }, { "name": "hand_right_joints", "l2": 11.36319, "mean_abs": 0.297919, "relative": 0.895723 }, { "name": "body_joints", "l2": 6.327158, "mean_abs": 0.106794, "relative": 0.854604 }, { "name": "body_contacts", "l2": 3.46471, "mean_abs": 0.081716, "relative": 0.9998 }, { "name": "camera_translation", "l2": 0.206399, "mean_abs": 0.025464, "relative": 0.487073 }, { "name": "camera_rotation_matrix", "l2": 3.001119, "mean_abs": 0.202495, "relative": 0.921595 }, { "name": "imu_accel_gyro", "l2": 16.867853, "mean_abs": 1.121728, "relative": 0.910079 }, { "name": "depth_confidence", "l2": 10.188525, "mean_abs": 0.194723, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.29227, "mean_abs": 0.111181, "relative": 0.958317 }, { "name": "video_fisheye_cam1", "l2": 5.251781, "mean_abs": 0.113722, "relative": 0.867812 }, { "name": "video_fisheye_cam2", "l2": 5.647807, "mean_abs": 0.116169, "relative": 0.901588 }, { "name": "video_fisheye_cam3", "l2": 5.434693, "mean_abs": 0.113851, "relative": 0.909974 }, { "name": "video_stereo_left", "l2": 5.368701, "mean_abs": 0.120436, "relative": 0.872425 }, { "name": "video_stereo_right", "l2": 6.305674, "mean_abs": 0.146229, "relative": 0.888968 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.074853, "mean_abs": 0.081545, "relative": 0.742559 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9708778858184814" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Prepare coffee equipment and scoop grounds", "correct": 0, "confidence": "0.9212464690208435" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.5309319496154785" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|scale|dripper|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1015, "start_frame": 5075, "end_frame": 5094, "center_frame": 5084, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.807799, "mean_abs": 0.266244, "relative": 0.911684 }, { "name": "hand_right_joints", "l2": 11.256933, "mean_abs": 0.28801, "relative": 0.887347 }, { "name": "body_joints", "l2": 6.284255, "mean_abs": 0.104858, "relative": 0.848809 }, { "name": "body_contacts", "l2": 3.46471, "mean_abs": 0.081716, "relative": 0.9998 }, { "name": "camera_translation", "l2": 0.217359, "mean_abs": 0.031303, "relative": 0.512937 }, { "name": "camera_rotation_matrix", "l2": 3.006255, "mean_abs": 0.208548, "relative": 0.923172 }, { "name": "imu_accel_gyro", "l2": 17.269794, "mean_abs": 1.167867, "relative": 0.931765 }, { "name": "depth_confidence", "l2": 10.135638, "mean_abs": 0.195151, "relative": 0.000336 }, { "name": "video_fisheye_cam0", "l2": 5.300006, "mean_abs": 0.112342, "relative": 0.959718 }, { "name": "video_fisheye_cam1", "l2": 5.256557, "mean_abs": 0.114841, "relative": 0.868602 }, { "name": "video_fisheye_cam2", "l2": 5.626662, "mean_abs": 0.11767, "relative": 0.898213 }, { "name": "video_fisheye_cam3", "l2": 5.517188, "mean_abs": 0.11738, "relative": 0.923787 }, { "name": "video_stereo_left", "l2": 5.358668, "mean_abs": 0.120605, "relative": 0.870794 }, { "name": "video_stereo_right", "l2": 6.295299, "mean_abs": 0.145997, "relative": 0.887505 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.027584, "mean_abs": 0.080495, "relative": 0.725642 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9635362029075623" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Prepare coffee equipment and scoop grounds", "correct": 0, "confidence": "0.9279413223266602" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.49623507261276245" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|scale|dripper|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1016, "start_frame": 5080, "end_frame": 5099, "center_frame": 5089, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.926826, "mean_abs": 0.258188, "relative": 0.921724 }, { "name": "hand_right_joints", "l2": 11.255478, "mean_abs": 0.266047, "relative": 0.887233 }, { "name": "body_joints", "l2": 6.240498, "mean_abs": 0.101725, "relative": 0.842899 }, { "name": "body_contacts", "l2": 3.46471, "mean_abs": 0.081715, "relative": 0.9998 }, { "name": "camera_translation", "l2": 0.210823, "mean_abs": 0.033395, "relative": 0.497512 }, { "name": "camera_rotation_matrix", "l2": 3.015009, "mean_abs": 0.210085, "relative": 0.92586 }, { "name": "imu_accel_gyro", "l2": 17.253092, "mean_abs": 1.126461, "relative": 0.930864 }, { "name": "depth_confidence", "l2": 9.993052, "mean_abs": 0.193797, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.274002, "mean_abs": 0.112115, "relative": 0.955009 }, { "name": "video_fisheye_cam1", "l2": 5.252312, "mean_abs": 0.11515, "relative": 0.8679 }, { "name": "video_fisheye_cam2", "l2": 5.587867, "mean_abs": 0.118172, "relative": 0.89202 }, { "name": "video_fisheye_cam3", "l2": 5.54054, "mean_abs": 0.11781, "relative": 0.927697 }, { "name": "video_stereo_left", "l2": 5.333582, "mean_abs": 0.120885, "relative": 0.866718 }, { "name": "video_stereo_right", "l2": 6.235074, "mean_abs": 0.145067, "relative": 0.879015 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.014413, "mean_abs": 0.079879, "relative": 0.720928 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.4945189356803894" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Prepare coffee equipment and scoop grounds", "correct": 0, "confidence": "0.8607233166694641" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.41759663820266724" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|bottle|dripper|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1017, "start_frame": 5085, "end_frame": 5104, "center_frame": 5094, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.316845, "mean_abs": 0.249273, "relative": 0.954624 }, { "name": "hand_right_joints", "l2": 11.361213, "mean_abs": 0.255552, "relative": 0.895567 }, { "name": "body_joints", "l2": 6.328641, "mean_abs": 0.09727, "relative": 0.854804 }, { "name": "body_contacts", "l2": 3.46471, "mean_abs": 0.081715, "relative": 0.9998 }, { "name": "camera_translation", "l2": 0.15825, "mean_abs": 0.025089, "relative": 0.373449 }, { "name": "camera_rotation_matrix", "l2": 3.009797, "mean_abs": 0.205779, "relative": 0.92426 }, { "name": "imu_accel_gyro", "l2": 17.391428, "mean_abs": 1.113994, "relative": 0.938327 }, { "name": "depth_confidence", "l2": 9.856077, "mean_abs": 0.189176, "relative": 0.000327 }, { "name": "video_fisheye_cam0", "l2": 5.262763, "mean_abs": 0.111288, "relative": 0.952974 }, { "name": "video_fisheye_cam1", "l2": 5.241762, "mean_abs": 0.11374, "relative": 0.866157 }, { "name": "video_fisheye_cam2", "l2": 5.536219, "mean_abs": 0.115988, "relative": 0.883775 }, { "name": "video_fisheye_cam3", "l2": 5.563827, "mean_abs": 0.11647, "relative": 0.931596 }, { "name": "video_stereo_left", "l2": 5.252239, "mean_abs": 0.117171, "relative": 0.853499 }, { "name": "video_stereo_right", "l2": 6.167145, "mean_abs": 0.141954, "relative": 0.869438 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.009084, "mean_abs": 0.081068, "relative": 0.719021 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.9166053533554077" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.7806551456451416" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.9049640893936157" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|bottle|dripper|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1018, "start_frame": 5090, "end_frame": 5109, "center_frame": 5099, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.596701, "mean_abs": 0.246923, "relative": 0.978231 }, { "name": "hand_right_joints", "l2": 11.394958, "mean_abs": 0.255018, "relative": 0.898228 }, { "name": "body_joints", "l2": 6.325882, "mean_abs": 0.094897, "relative": 0.854432 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.075854, "mean_abs": 0.010936, "relative": 0.179004 }, { "name": "camera_rotation_matrix", "l2": 3.004769, "mean_abs": 0.199423, "relative": 0.922716 }, { "name": "imu_accel_gyro", "l2": 17.506989, "mean_abs": 1.092965, "relative": 0.944562 }, { "name": "depth_confidence", "l2": 9.850473, "mean_abs": 0.185711, "relative": 0.000327 }, { "name": "video_fisheye_cam0", "l2": 5.243809, "mean_abs": 0.110015, "relative": 0.949542 }, { "name": "video_fisheye_cam1", "l2": 5.214322, "mean_abs": 0.111889, "relative": 0.861623 }, { "name": "video_fisheye_cam2", "l2": 5.479005, "mean_abs": 0.113459, "relative": 0.874642 }, { "name": "video_fisheye_cam3", "l2": 5.579104, "mean_abs": 0.115995, "relative": 0.934154 }, { "name": "video_stereo_left", "l2": 5.180482, "mean_abs": 0.113926, "relative": 0.841839 }, { "name": "video_stereo_right", "l2": 6.090116, "mean_abs": 0.137993, "relative": 0.858579 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.124372, "mean_abs": 0.088445, "relative": 0.760281 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.9497877359390259" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.9605011940002441" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.9401044845581055" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|bottle|dripper|coffee filter", "correct": 0, "confidence": "" } } }, { "window_index": 1019, "start_frame": 5095, "end_frame": 5114, "center_frame": 5104, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.578143, "mean_abs": 0.240325, "relative": 0.976666 }, { "name": "hand_right_joints", "l2": 11.361524, "mean_abs": 0.248814, "relative": 0.895592 }, { "name": "body_joints", "l2": 6.315565, "mean_abs": 0.09581, "relative": 0.853038 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.023215, "mean_abs": 0.003438, "relative": 0.054784 }, { "name": "camera_rotation_matrix", "l2": 3.000327, "mean_abs": 0.190265, "relative": 0.921352 }, { "name": "imu_accel_gyro", "l2": 17.198875, "mean_abs": 1.068349, "relative": 0.927939 }, { "name": "depth_confidence", "l2": 9.760631, "mean_abs": 0.181072, "relative": 0.000324 }, { "name": "video_fisheye_cam0", "l2": 5.211554, "mean_abs": 0.107961, "relative": 0.943701 }, { "name": "video_fisheye_cam1", "l2": 5.190973, "mean_abs": 0.110096, "relative": 0.857764 }, { "name": "video_fisheye_cam2", "l2": 5.401338, "mean_abs": 0.109842, "relative": 0.862243 }, { "name": "video_fisheye_cam3", "l2": 5.585505, "mean_abs": 0.114667, "relative": 0.935225 }, { "name": "video_stereo_left", "l2": 5.079997, "mean_abs": 0.108723, "relative": 0.82551 }, { "name": "video_stereo_right", "l2": 5.995883, "mean_abs": 0.133313, "relative": 0.845294 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.222177, "mean_abs": 0.091612, "relative": 0.795284 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.9502944350242615" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.9771428108215332" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.8994784355163574" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|coffee dripper|scale|bottle|dripper|coffee filter", "correct": 0, "confidence": "" } } }, { "window_index": 1020, "start_frame": 5100, "end_frame": 5119, "center_frame": 5109, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.507866, "mean_abs": 0.236568, "relative": 0.970738 }, { "name": "hand_right_joints", "l2": 11.348314, "mean_abs": 0.245838, "relative": 0.894551 }, { "name": "body_joints", "l2": 6.288231, "mean_abs": 0.094863, "relative": 0.849346 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013684, "mean_abs": 0.002209, "relative": 0.032292 }, { "name": "camera_rotation_matrix", "l2": 2.999971, "mean_abs": 0.186605, "relative": 0.921242 }, { "name": "imu_accel_gyro", "l2": 17.161274, "mean_abs": 1.055377, "relative": 0.92591 }, { "name": "depth_confidence", "l2": 11.517874, "mean_abs": 0.185596, "relative": 0.000382 }, { "name": "video_fisheye_cam0", "l2": 5.196759, "mean_abs": 0.106995, "relative": 0.941022 }, { "name": "video_fisheye_cam1", "l2": 5.195908, "mean_abs": 0.109135, "relative": 0.85858 }, { "name": "video_fisheye_cam2", "l2": 5.391241, "mean_abs": 0.108514, "relative": 0.860631 }, { "name": "video_fisheye_cam3", "l2": 5.594152, "mean_abs": 0.115042, "relative": 0.936673 }, { "name": "video_stereo_left", "l2": 5.070258, "mean_abs": 0.1068, "relative": 0.823927 }, { "name": "video_stereo_right", "l2": 5.981109, "mean_abs": 0.131302, "relative": 0.843211 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.264921, "mean_abs": 0.091802, "relative": 0.810581 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.9347808361053467" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.9797072410583496" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.9422776699066162" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|coffee dripper|scale|bottle|dripper|coffee filter", "correct": 0, "confidence": "" } } }, { "window_index": 1021, "start_frame": 5105, "end_frame": 5124, "center_frame": 5114, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 11.208226, "mean_abs": 0.236523, "relative": 0.945462 }, { "name": "hand_right_joints", "l2": 11.333435, "mean_abs": 0.244518, "relative": 0.893378 }, { "name": "body_joints", "l2": 6.309919, "mean_abs": 0.096402, "relative": 0.852276 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.039773, "mean_abs": 0.005857, "relative": 0.093859 }, { "name": "camera_rotation_matrix", "l2": 3.00012, "mean_abs": 0.189233, "relative": 0.921288 }, { "name": "imu_accel_gyro", "l2": 17.219013, "mean_abs": 1.042812, "relative": 0.929025 }, { "name": "depth_confidence", "l2": 11.633878, "mean_abs": 0.189688, "relative": 0.000386 }, { "name": "video_fisheye_cam0", "l2": 5.19821, "mean_abs": 0.107266, "relative": 0.941285 }, { "name": "video_fisheye_cam1", "l2": 5.218109, "mean_abs": 0.110069, "relative": 0.862248 }, { "name": "video_fisheye_cam2", "l2": 5.435489, "mean_abs": 0.110189, "relative": 0.867695 }, { "name": "video_fisheye_cam3", "l2": 5.522449, "mean_abs": 0.112989, "relative": 0.924668 }, { "name": "video_stereo_left", "l2": 5.105643, "mean_abs": 0.108697, "relative": 0.829677 }, { "name": "video_stereo_right", "l2": 6.009132, "mean_abs": 0.13313, "relative": 0.847162 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.235849, "mean_abs": 0.090824, "relative": 0.800177 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Hold coffee carafe", "correct": 0, "confidence": "0.7165603041648865" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.6599328517913818" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle", "correct": 0, "confidence": "0.816390335559845" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|coffee dripper|scale|bottle|coffee filter|coffee scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1022, "start_frame": 5110, "end_frame": 5129, "center_frame": 5119, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.906225, "mean_abs": 0.236091, "relative": 0.919987 }, { "name": "hand_right_joints", "l2": 11.320634, "mean_abs": 0.248158, "relative": 0.892369 }, { "name": "body_joints", "l2": 6.362029, "mean_abs": 0.099895, "relative": 0.859314 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.088345, "mean_abs": 0.013174, "relative": 0.208482 }, { "name": "camera_rotation_matrix", "l2": 3.003348, "mean_abs": 0.197757, "relative": 0.922279 }, { "name": "imu_accel_gyro", "l2": 17.251062, "mean_abs": 1.095707, "relative": 0.930754 }, { "name": "depth_confidence", "l2": 11.753034, "mean_abs": 0.196651, "relative": 0.00039 }, { "name": "video_fisheye_cam0", "l2": 5.208578, "mean_abs": 0.10888, "relative": 0.943162 }, { "name": "video_fisheye_cam1", "l2": 5.263819, "mean_abs": 0.11277, "relative": 0.869802 }, { "name": "video_fisheye_cam2", "l2": 5.513045, "mean_abs": 0.114058, "relative": 0.880076 }, { "name": "video_fisheye_cam3", "l2": 5.482807, "mean_abs": 0.113507, "relative": 0.91803 }, { "name": "video_stereo_left", "l2": 5.192993, "mean_abs": 0.114228, "relative": 0.843872 }, { "name": "video_stereo_right", "l2": 6.116295, "mean_abs": 0.139189, "relative": 0.862269 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.435709, "mean_abs": 0.099966, "relative": 0.871704 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Hold coffee carafe", "correct": 0, "confidence": "0.9422505497932434" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Prepare coffee equipment and scoop grounds", "correct": 0, "confidence": "0.9304724931716919" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.8562440276145935" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|coffee dripper|scale|bottle|coffee scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1023, "start_frame": 5115, "end_frame": 5134, "center_frame": 5124, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.720885, "mean_abs": 0.235596, "relative": 0.904352 }, { "name": "hand_right_joints", "l2": 11.272177, "mean_abs": 0.259392, "relative": 0.888549 }, { "name": "body_joints", "l2": 6.366507, "mean_abs": 0.100359, "relative": 0.859919 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.156868, "mean_abs": 0.023584, "relative": 0.370187 }, { "name": "camera_rotation_matrix", "l2": 3.008338, "mean_abs": 0.205769, "relative": 0.923812 }, { "name": "imu_accel_gyro", "l2": 17.443865, "mean_abs": 1.144511, "relative": 0.941157 }, { "name": "depth_confidence", "l2": 12.558981, "mean_abs": 0.206702, "relative": 0.000417 }, { "name": "video_fisheye_cam0", "l2": 5.223904, "mean_abs": 0.110158, "relative": 0.945937 }, { "name": "video_fisheye_cam1", "l2": 5.316758, "mean_abs": 0.115336, "relative": 0.878549 }, { "name": "video_fisheye_cam2", "l2": 5.588329, "mean_abs": 0.117488, "relative": 0.892094 }, { "name": "video_fisheye_cam3", "l2": 5.442732, "mean_abs": 0.113803, "relative": 0.91132 }, { "name": "video_stereo_left", "l2": 5.275243, "mean_abs": 0.118654, "relative": 0.857238 }, { "name": "video_stereo_right", "l2": 6.210959, "mean_abs": 0.144303, "relative": 0.875615 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.123652, "mean_abs": 0.083139, "relative": 0.760023 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Hold coffee carafe", "correct": 0, "confidence": "0.8514485955238342" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Prepare coffee equipment and scoop grounds", "correct": 0, "confidence": "0.7899371385574341" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999998807907104" }, "next_action": { "true": "Pour coffee", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.9842697381973267" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1024, "start_frame": 5120, "end_frame": 5139, "center_frame": 5129, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.298969, "mean_abs": 0.230788, "relative": 0.868762 }, { "name": "hand_right_joints", "l2": 11.09692, "mean_abs": 0.268384, "relative": 0.874734 }, { "name": "body_joints", "l2": 6.343932, "mean_abs": 0.100765, "relative": 0.85687 }, { "name": "body_contacts", "l2": 3.464548, "mean_abs": 0.081741, "relative": 0.999754 }, { "name": "camera_translation", "l2": 0.222365, "mean_abs": 0.033215, "relative": 0.524751 }, { "name": "camera_rotation_matrix", "l2": 3.009311, "mean_abs": 0.209387, "relative": 0.924111 }, { "name": "imu_accel_gyro", "l2": 17.268044, "mean_abs": 1.157575, "relative": 0.93167 }, { "name": "depth_confidence", "l2": 11.040924, "mean_abs": 0.203975, "relative": 0.000366 }, { "name": "video_fisheye_cam0", "l2": 5.224992, "mean_abs": 0.110222, "relative": 0.946134 }, { "name": "video_fisheye_cam1", "l2": 5.381909, "mean_abs": 0.116949, "relative": 0.889315 }, { "name": "video_fisheye_cam2", "l2": 5.657968, "mean_abs": 0.11912, "relative": 0.90321 }, { "name": "video_fisheye_cam3", "l2": 5.410855, "mean_abs": 0.113567, "relative": 0.905982 }, { "name": "video_stereo_left", "l2": 5.321915, "mean_abs": 0.120493, "relative": 0.864822 }, { "name": "video_stereo_right", "l2": 6.270208, "mean_abs": 0.145717, "relative": 0.883968 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.002398, "mean_abs": 0.078035, "relative": 0.716628 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.6639490127563477" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Prepare coffee equipment and scoop grounds", "correct": 0, "confidence": "0.8932585716247559" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999998807907104" }, "next_action": { "true": "Pour coffee", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.9840779304504395" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1025, "start_frame": 5125, "end_frame": 5144, "center_frame": 5134, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.946459, "mean_abs": 0.225137, "relative": 0.839026 }, { "name": "hand_right_joints", "l2": 10.915402, "mean_abs": 0.279039, "relative": 0.860426 }, { "name": "body_joints", "l2": 6.339865, "mean_abs": 0.101468, "relative": 0.85632 }, { "name": "body_contacts", "l2": 3.458094, "mean_abs": 0.084978, "relative": 0.997891 }, { "name": "camera_translation", "l2": 0.25789, "mean_abs": 0.037975, "relative": 0.608583 }, { "name": "camera_rotation_matrix", "l2": 3.009217, "mean_abs": 0.209931, "relative": 0.924082 }, { "name": "imu_accel_gyro", "l2": 17.379351, "mean_abs": 1.189435, "relative": 0.937676 }, { "name": "depth_confidence", "l2": 10.452424, "mean_abs": 0.20193, "relative": 0.000347 }, { "name": "video_fisheye_cam0", "l2": 5.218285, "mean_abs": 0.109885, "relative": 0.94492 }, { "name": "video_fisheye_cam1", "l2": 5.466238, "mean_abs": 0.119435, "relative": 0.90325 }, { "name": "video_fisheye_cam2", "l2": 5.703269, "mean_abs": 0.119324, "relative": 0.910442 }, { "name": "video_fisheye_cam3", "l2": 5.370389, "mean_abs": 0.112563, "relative": 0.899207 }, { "name": "video_stereo_left", "l2": 5.353099, "mean_abs": 0.121784, "relative": 0.869889 }, { "name": "video_stereo_right", "l2": 6.411255, "mean_abs": 0.149575, "relative": 0.903852 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.990273, "mean_abs": 0.077484, "relative": 0.712289 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9943434000015259" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour milk into coffee", "correct": 0, "confidence": "1.0" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle away", "correct": 0, "confidence": "0.6651692986488342" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|digital scale|coffee filter|glass carafe|white mug|wooden scoop|coffee mug|mug|carafe|milk bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1026, "start_frame": 5130, "end_frame": 5149, "center_frame": 5139, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.787419, "mean_abs": 0.225613, "relative": 0.825611 }, { "name": "hand_right_joints", "l2": 10.859257, "mean_abs": 0.288659, "relative": 0.856 }, { "name": "body_joints", "l2": 6.26758, "mean_abs": 0.102156, "relative": 0.846557 }, { "name": "body_contacts", "l2": 3.295051, "mean_abs": 0.078379, "relative": 0.950842 }, { "name": "camera_translation", "l2": 0.291783, "mean_abs": 0.043451, "relative": 0.688568 }, { "name": "camera_rotation_matrix", "l2": 3.007363, "mean_abs": 0.212006, "relative": 0.923512 }, { "name": "imu_accel_gyro", "l2": 17.368132, "mean_abs": 1.176264, "relative": 0.93707 }, { "name": "depth_confidence", "l2": 10.635536, "mean_abs": 0.204215, "relative": 0.000353 }, { "name": "video_fisheye_cam0", "l2": 5.207269, "mean_abs": 0.110095, "relative": 0.942925 }, { "name": "video_fisheye_cam1", "l2": 5.533599, "mean_abs": 0.121169, "relative": 0.914381 }, { "name": "video_fisheye_cam2", "l2": 5.723979, "mean_abs": 0.119916, "relative": 0.913748 }, { "name": "video_fisheye_cam3", "l2": 5.356579, "mean_abs": 0.112702, "relative": 0.896895 }, { "name": "video_stereo_left", "l2": 5.432339, "mean_abs": 0.125795, "relative": 0.882766 }, { "name": "video_stereo_right", "l2": 6.538274, "mean_abs": 0.153885, "relative": 0.921759 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.034685, "mean_abs": 0.081179, "relative": 0.728183 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Move kettle away", "correct": 0, "confidence": "0.9999996423721313" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour coffee", "correct": 1, "confidence": "1.0" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.561688244342804" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|digital scale|coffee filter|glass carafe|coffee scale|white mug|wooden scoop|closed coffee container|coffee mug|milk bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1027, "start_frame": 5135, "end_frame": 5154, "center_frame": 5144, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.938686, "mean_abs": 0.236717, "relative": 0.838371 }, { "name": "hand_right_joints", "l2": 10.664884, "mean_abs": 0.287464, "relative": 0.840678 }, { "name": "body_joints", "l2": 6.15786, "mean_abs": 0.100132, "relative": 0.831737 }, { "name": "body_contacts", "l2": 3.303752, "mean_abs": 0.079126, "relative": 0.953353 }, { "name": "camera_translation", "l2": 0.302592, "mean_abs": 0.04458, "relative": 0.714076 }, { "name": "camera_rotation_matrix", "l2": 3.011199, "mean_abs": 0.216147, "relative": 0.92469 }, { "name": "imu_accel_gyro", "l2": 17.169615, "mean_abs": 1.140161, "relative": 0.92636 }, { "name": "depth_confidence", "l2": 10.712818, "mean_abs": 0.205084, "relative": 0.000356 }, { "name": "video_fisheye_cam0", "l2": 5.191029, "mean_abs": 0.1105, "relative": 0.939984 }, { "name": "video_fisheye_cam1", "l2": 5.574136, "mean_abs": 0.122357, "relative": 0.921079 }, { "name": "video_fisheye_cam2", "l2": 5.745203, "mean_abs": 0.120916, "relative": 0.917136 }, { "name": "video_fisheye_cam3", "l2": 5.317333, "mean_abs": 0.111863, "relative": 0.890323 }, { "name": "video_stereo_left", "l2": 5.51679, "mean_abs": 0.128809, "relative": 0.896489 }, { "name": "video_stereo_right", "l2": 6.670541, "mean_abs": 0.158339, "relative": 0.940406 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.082289, "mean_abs": 0.085045, "relative": 0.74522 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9999984502792358" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour coffee", "correct": 1, "confidence": "1.0" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.9999996423721313" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|digital scale|coffee filter|glass carafe|coffee scale|wooden scoop|closed coffee container|coffee mug|milk bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1028, "start_frame": 5140, "end_frame": 5159, "center_frame": 5149, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.088595, "mean_abs": 0.245871, "relative": 0.851016 }, { "name": "hand_right_joints", "l2": 10.456785, "mean_abs": 0.282288, "relative": 0.824274 }, { "name": "body_joints", "l2": 6.048359, "mean_abs": 0.098216, "relative": 0.816947 }, { "name": "body_contacts", "l2": 3.303219, "mean_abs": 0.079137, "relative": 0.953199 }, { "name": "camera_translation", "l2": 0.26862, "mean_abs": 0.037951, "relative": 0.633905 }, { "name": "camera_rotation_matrix", "l2": 3.012092, "mean_abs": 0.2156, "relative": 0.924964 }, { "name": "imu_accel_gyro", "l2": 16.983, "mean_abs": 1.0588, "relative": 0.916291 }, { "name": "depth_confidence", "l2": 10.740235, "mean_abs": 0.203376, "relative": 0.000356 }, { "name": "video_fisheye_cam0", "l2": 5.166207, "mean_abs": 0.11035, "relative": 0.93549 }, { "name": "video_fisheye_cam1", "l2": 5.582137, "mean_abs": 0.122514, "relative": 0.922401 }, { "name": "video_fisheye_cam2", "l2": 5.740058, "mean_abs": 0.121421, "relative": 0.916315 }, { "name": "video_fisheye_cam3", "l2": 5.305442, "mean_abs": 0.111329, "relative": 0.888332 }, { "name": "video_stereo_left", "l2": 5.599506, "mean_abs": 0.130794, "relative": 0.909931 }, { "name": "video_stereo_right", "l2": 6.776058, "mean_abs": 0.160023, "relative": 0.955282 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.102032, "mean_abs": 0.086612, "relative": 0.752286 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9999712705612183" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour coffee", "correct": 1, "confidence": "1.0" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.9999998807907104" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|digital scale|coffee filter|glass carafe|coffee scale|wooden scoop|closed coffee container|coffee mug|milk bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1029, "start_frame": 5145, "end_frame": 5164, "center_frame": 5154, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.158941, "mean_abs": 0.247339, "relative": 0.85695 }, { "name": "hand_right_joints", "l2": 10.265509, "mean_abs": 0.270893, "relative": 0.809197 }, { "name": "body_joints", "l2": 5.983666, "mean_abs": 0.096825, "relative": 0.808209 }, { "name": "body_contacts", "l2": 3.457593, "mean_abs": 0.086112, "relative": 0.997747 }, { "name": "camera_translation", "l2": 0.198495, "mean_abs": 0.026702, "relative": 0.468421 }, { "name": "camera_rotation_matrix", "l2": 3.004235, "mean_abs": 0.210091, "relative": 0.922552 }, { "name": "imu_accel_gyro", "l2": 16.937628, "mean_abs": 1.070791, "relative": 0.913843 }, { "name": "depth_confidence", "l2": 10.718085, "mean_abs": 0.199679, "relative": 0.000356 }, { "name": "video_fisheye_cam0", "l2": 5.134096, "mean_abs": 0.108874, "relative": 0.929675 }, { "name": "video_fisheye_cam1", "l2": 5.603237, "mean_abs": 0.121352, "relative": 0.925888 }, { "name": "video_fisheye_cam2", "l2": 5.698231, "mean_abs": 0.119683, "relative": 0.909638 }, { "name": "video_fisheye_cam3", "l2": 5.346959, "mean_abs": 0.111772, "relative": 0.895284 }, { "name": "video_stereo_left", "l2": 5.556012, "mean_abs": 0.12817, "relative": 0.902863 }, { "name": "video_stereo_right", "l2": 6.828979, "mean_abs": 0.159242, "relative": 0.962743 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.235379, "mean_abs": 0.089398, "relative": 0.800009 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Lift gooseneck kettle", "correct": 0, "confidence": "1.0" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour coffee", "correct": 1, "confidence": "1.0" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "1.0" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|digital scale|coffee filter|glass carafe|coffee scale|wooden scoop|closed coffee container|coffee mug|coffee equipment|small bottle|milk bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1030, "start_frame": 5150, "end_frame": 5169, "center_frame": 5159, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.296796, "mean_abs": 0.248673, "relative": 0.868579 }, { "name": "hand_right_joints", "l2": 10.315609, "mean_abs": 0.272668, "relative": 0.813146 }, { "name": "body_joints", "l2": 5.909173, "mean_abs": 0.09255, "relative": 0.798147 }, { "name": "body_contacts", "l2": 3.465272, "mean_abs": 0.084741, "relative": 0.999963 }, { "name": "camera_translation", "l2": 0.09256, "mean_abs": 0.01215, "relative": 0.218428 }, { "name": "camera_rotation_matrix", "l2": 2.999667, "mean_abs": 0.206663, "relative": 0.921149 }, { "name": "imu_accel_gyro", "l2": 17.012314, "mean_abs": 1.076571, "relative": 0.917873 }, { "name": "depth_confidence", "l2": 10.686304, "mean_abs": 0.19803, "relative": 0.000355 }, { "name": "video_fisheye_cam0", "l2": 5.103792, "mean_abs": 0.107061, "relative": 0.924188 }, { "name": "video_fisheye_cam1", "l2": 5.586452, "mean_abs": 0.119708, "relative": 0.923114 }, { "name": "video_fisheye_cam2", "l2": 5.688673, "mean_abs": 0.118154, "relative": 0.908112 }, { "name": "video_fisheye_cam3", "l2": 5.443203, "mean_abs": 0.113584, "relative": 0.911399 }, { "name": "video_stereo_left", "l2": 5.567446, "mean_abs": 0.126213, "relative": 0.904721 }, { "name": "video_stereo_right", "l2": 6.840509, "mean_abs": 0.156126, "relative": 0.964368 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.195879, "mean_abs": 0.088782, "relative": 0.785872 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.9757156372070312" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour coffee", "correct": 1, "confidence": "0.9892581105232239" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.813234806060791" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|digital scale|table|coffee filter|glass carafe|wooden scoop|closed coffee container|coffee mug|coffee equipment|small bottle|milk bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1031, "start_frame": 5155, "end_frame": 5174, "center_frame": 5164, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.437869, "mean_abs": 0.253664, "relative": 0.880479 }, { "name": "hand_right_joints", "l2": 10.338916, "mean_abs": 0.280625, "relative": 0.814983 }, { "name": "body_joints", "l2": 5.923757, "mean_abs": 0.0928, "relative": 0.800117 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.087335, "mean_abs": 0.011998, "relative": 0.206099 }, { "name": "camera_rotation_matrix", "l2": 3.002824, "mean_abs": 0.208163, "relative": 0.922119 }, { "name": "imu_accel_gyro", "l2": 17.102432, "mean_abs": 1.100143, "relative": 0.922735 }, { "name": "depth_confidence", "l2": 10.612138, "mean_abs": 0.196757, "relative": 0.000352 }, { "name": "video_fisheye_cam0", "l2": 5.105183, "mean_abs": 0.107201, "relative": 0.924439 }, { "name": "video_fisheye_cam1", "l2": 5.574277, "mean_abs": 0.119865, "relative": 0.921102 }, { "name": "video_fisheye_cam2", "l2": 5.682472, "mean_abs": 0.118011, "relative": 0.907122 }, { "name": "video_fisheye_cam3", "l2": 5.54809, "mean_abs": 0.117132, "relative": 0.928961 }, { "name": "video_stereo_left", "l2": 5.58571, "mean_abs": 0.124705, "relative": 0.907689 }, { "name": "video_stereo_right", "l2": 6.873284, "mean_abs": 0.15571, "relative": 0.968989 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.198011, "mean_abs": 0.088753, "relative": 0.786635 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.7183989882469177" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.854143500328064" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.30626800656318665" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1032, "start_frame": 5160, "end_frame": 5179, "center_frame": 5169, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.459478, "mean_abs": 0.257152, "relative": 0.882302 }, { "name": "hand_right_joints", "l2": 10.30488, "mean_abs": 0.284025, "relative": 0.8123 }, { "name": "body_joints", "l2": 5.961876, "mean_abs": 0.094398, "relative": 0.805266 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.116463, "mean_abs": 0.014609, "relative": 0.274837 }, { "name": "camera_rotation_matrix", "l2": 3.003103, "mean_abs": 0.207815, "relative": 0.922204 }, { "name": "imu_accel_gyro", "l2": 17.151258, "mean_abs": 1.078417, "relative": 0.925369 }, { "name": "depth_confidence", "l2": 10.546371, "mean_abs": 0.194793, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.110972, "mean_abs": 0.107609, "relative": 0.925488 }, { "name": "video_fisheye_cam1", "l2": 5.597098, "mean_abs": 0.119965, "relative": 0.924873 }, { "name": "video_fisheye_cam2", "l2": 5.681051, "mean_abs": 0.118213, "relative": 0.906895 }, { "name": "video_fisheye_cam3", "l2": 5.621729, "mean_abs": 0.11708, "relative": 0.941291 }, { "name": "video_stereo_left", "l2": 5.60107, "mean_abs": 0.124594, "relative": 0.910185 }, { "name": "video_stereo_right", "l2": 6.933857, "mean_abs": 0.156236, "relative": 0.977528 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.267764, "mean_abs": 0.093583, "relative": 0.811599 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.7217246890068054" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.7608070969581604" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.29437217116355896" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1033, "start_frame": 5165, "end_frame": 5184, "center_frame": 5174, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.445679, "mean_abs": 0.256372, "relative": 0.881138 }, { "name": "hand_right_joints", "l2": 10.212999, "mean_abs": 0.279915, "relative": 0.805058 }, { "name": "body_joints", "l2": 6.007888, "mean_abs": 0.094003, "relative": 0.811481 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.089611, "mean_abs": 0.012121, "relative": 0.211469 }, { "name": "camera_rotation_matrix", "l2": 3.001201, "mean_abs": 0.203704, "relative": 0.92162 }, { "name": "imu_accel_gyro", "l2": 17.060493, "mean_abs": 1.056607, "relative": 0.920472 }, { "name": "depth_confidence", "l2": 10.499024, "mean_abs": 0.192749, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.117777, "mean_abs": 0.106885, "relative": 0.92672 }, { "name": "video_fisheye_cam1", "l2": 5.603076, "mean_abs": 0.118843, "relative": 0.925861 }, { "name": "video_fisheye_cam2", "l2": 5.665847, "mean_abs": 0.117169, "relative": 0.904468 }, { "name": "video_fisheye_cam3", "l2": 5.699393, "mean_abs": 0.117118, "relative": 0.954295 }, { "name": "video_stereo_left", "l2": 5.605036, "mean_abs": 0.123471, "relative": 0.910829 }, { "name": "video_stereo_right", "l2": 6.93516, "mean_abs": 0.153723, "relative": 0.977712 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.120677, "mean_abs": 0.087395, "relative": 0.758958 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.41114774346351624" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.3951317071914673" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999994039535522" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6223069429397583" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1034, "start_frame": 5170, "end_frame": 5189, "center_frame": 5179, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.439236, "mean_abs": 0.254347, "relative": 0.880594 }, { "name": "hand_right_joints", "l2": 10.137465, "mean_abs": 0.275255, "relative": 0.799103 }, { "name": "body_joints", "l2": 6.04935, "mean_abs": 0.092851, "relative": 0.817081 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.048016, "mean_abs": 0.006801, "relative": 0.113311 }, { "name": "camera_rotation_matrix", "l2": 3.000802, "mean_abs": 0.199404, "relative": 0.921498 }, { "name": "imu_accel_gyro", "l2": 17.047424, "mean_abs": 1.021844, "relative": 0.919767 }, { "name": "depth_confidence", "l2": 10.421362, "mean_abs": 0.189269, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.122111, "mean_abs": 0.106144, "relative": 0.927505 }, { "name": "video_fisheye_cam1", "l2": 5.588083, "mean_abs": 0.117148, "relative": 0.923384 }, { "name": "video_fisheye_cam2", "l2": 5.641216, "mean_abs": 0.115801, "relative": 0.900536 }, { "name": "video_fisheye_cam3", "l2": 5.709561, "mean_abs": 0.116184, "relative": 0.955997 }, { "name": "video_stereo_left", "l2": 5.605466, "mean_abs": 0.121895, "relative": 0.910899 }, { "name": "video_stereo_right", "l2": 6.92381, "mean_abs": 0.150371, "relative": 0.976112 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.304624, "mean_abs": 0.092395, "relative": 0.82479 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5209978222846985" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.4907296299934387" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999980926513672" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8411495685577393" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1035, "start_frame": 5175, "end_frame": 5194, "center_frame": 5184, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.443578, "mean_abs": 0.25328, "relative": 0.88096 }, { "name": "hand_right_joints", "l2": 10.068984, "mean_abs": 0.272277, "relative": 0.793705 }, { "name": "body_joints", "l2": 6.082952, "mean_abs": 0.091335, "relative": 0.821619 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021117, "mean_abs": 0.003105, "relative": 0.049833 }, { "name": "camera_rotation_matrix", "l2": 3.000009, "mean_abs": 0.198834, "relative": 0.921254 }, { "name": "imu_accel_gyro", "l2": 17.053457, "mean_abs": 1.02306, "relative": 0.920093 }, { "name": "depth_confidence", "l2": 10.384648, "mean_abs": 0.188637, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.115339, "mean_abs": 0.105832, "relative": 0.926278 }, { "name": "video_fisheye_cam1", "l2": 5.551649, "mean_abs": 0.116177, "relative": 0.917363 }, { "name": "video_fisheye_cam2", "l2": 5.630361, "mean_abs": 0.115384, "relative": 0.898803 }, { "name": "video_fisheye_cam3", "l2": 5.659861, "mean_abs": 0.1148, "relative": 0.947675 }, { "name": "video_stereo_left", "l2": 5.603792, "mean_abs": 0.121464, "relative": 0.910627 }, { "name": "video_stereo_right", "l2": 6.887991, "mean_abs": 0.148941, "relative": 0.971062 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.145132, "mean_abs": 0.088936, "relative": 0.767711 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6426913738250732" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.30375537276268005" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999982118606567" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8454254269599915" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1036, "start_frame": 5180, "end_frame": 5199, "center_frame": 5189, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.44064, "mean_abs": 0.253147, "relative": 0.880713 }, { "name": "hand_right_joints", "l2": 10.021583, "mean_abs": 0.270904, "relative": 0.789969 }, { "name": "body_joints", "l2": 6.107556, "mean_abs": 0.09091, "relative": 0.824943 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010971, "mean_abs": 0.00169, "relative": 0.025889 }, { "name": "camera_rotation_matrix", "l2": 3.000232, "mean_abs": 0.199033, "relative": 0.921322 }, { "name": "imu_accel_gyro", "l2": 17.03857, "mean_abs": 1.008196, "relative": 0.91929 }, { "name": "depth_confidence", "l2": 10.394781, "mean_abs": 0.189174, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.114523, "mean_abs": 0.105871, "relative": 0.926131 }, { "name": "video_fisheye_cam1", "l2": 5.545686, "mean_abs": 0.116, "relative": 0.916378 }, { "name": "video_fisheye_cam2", "l2": 5.625035, "mean_abs": 0.115319, "relative": 0.897953 }, { "name": "video_fisheye_cam3", "l2": 5.663636, "mean_abs": 0.115414, "relative": 0.948308 }, { "name": "video_stereo_left", "l2": 5.607852, "mean_abs": 0.121477, "relative": 0.911287 }, { "name": "video_stereo_right", "l2": 6.890983, "mean_abs": 0.148952, "relative": 0.971484 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.072991, "mean_abs": 0.083277, "relative": 0.741892 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010506, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6571552753448486" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.29451286792755127" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999947547912598" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8644084930419922" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1037, "start_frame": 5185, "end_frame": 5204, "center_frame": 5194, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.427595, "mean_abs": 0.252569, "relative": 0.879612 }, { "name": "hand_right_joints", "l2": 9.993095, "mean_abs": 0.270282, "relative": 0.787723 }, { "name": "body_joints", "l2": 6.130493, "mean_abs": 0.090546, "relative": 0.828041 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012319, "mean_abs": 0.001871, "relative": 0.029071 }, { "name": "camera_rotation_matrix", "l2": 3.000052, "mean_abs": 0.199111, "relative": 0.921267 }, { "name": "imu_accel_gyro", "l2": 17.034855, "mean_abs": 1.004491, "relative": 0.919089 }, { "name": "depth_confidence", "l2": 10.383751, "mean_abs": 0.189007, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.112652, "mean_abs": 0.105914, "relative": 0.925792 }, { "name": "video_fisheye_cam1", "l2": 5.528618, "mean_abs": 0.115325, "relative": 0.913557 }, { "name": "video_fisheye_cam2", "l2": 5.616985, "mean_abs": 0.115212, "relative": 0.896668 }, { "name": "video_fisheye_cam3", "l2": 5.657229, "mean_abs": 0.115116, "relative": 0.947235 }, { "name": "video_stereo_left", "l2": 5.603524, "mean_abs": 0.121461, "relative": 0.910584 }, { "name": "video_stereo_right", "l2": 6.877248, "mean_abs": 0.148164, "relative": 0.969548 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.015195, "mean_abs": 0.08166, "relative": 0.721208 }, { "name": "caption_objects_interaction_text", "l2": 1.828079, "mean_abs": 0.013999, "relative": 0.83941 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9318910241127014" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.28990304470062256" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999396800994873" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9659320712089539" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1038, "start_frame": 5190, "end_frame": 5209, "center_frame": 5199, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.404929, "mean_abs": 0.251105, "relative": 0.8777 }, { "name": "hand_right_joints", "l2": 9.937674, "mean_abs": 0.268132, "relative": 0.783355 }, { "name": "body_joints", "l2": 6.147984, "mean_abs": 0.090732, "relative": 0.830403 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013644, "mean_abs": 0.002043, "relative": 0.032199 }, { "name": "camera_rotation_matrix", "l2": 3.000161, "mean_abs": 0.198887, "relative": 0.921301 }, { "name": "imu_accel_gyro", "l2": 17.050526, "mean_abs": 1.010204, "relative": 0.919935 }, { "name": "depth_confidence", "l2": 10.370642, "mean_abs": 0.188138, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.113014, "mean_abs": 0.105856, "relative": 0.925858 }, { "name": "video_fisheye_cam1", "l2": 5.523945, "mean_abs": 0.115111, "relative": 0.912785 }, { "name": "video_fisheye_cam2", "l2": 5.613209, "mean_abs": 0.115001, "relative": 0.896065 }, { "name": "video_fisheye_cam3", "l2": 5.649154, "mean_abs": 0.114618, "relative": 0.945883 }, { "name": "video_stereo_left", "l2": 5.604964, "mean_abs": 0.121222, "relative": 0.910818 }, { "name": "video_stereo_right", "l2": 6.879145, "mean_abs": 0.147883, "relative": 0.969815 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.121939, "mean_abs": 0.085952, "relative": 0.75941 }, { "name": "caption_objects_interaction_text", "l2": 1.828079, "mean_abs": 0.01414, "relative": 0.83941 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.946835458278656" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.3146528899669647" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9994280934333801" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9539498686790466" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1039, "start_frame": 5195, "end_frame": 5214, "center_frame": 5204, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.386288, "mean_abs": 0.249292, "relative": 0.876128 }, { "name": "hand_right_joints", "l2": 9.893341, "mean_abs": 0.265477, "relative": 0.77986 }, { "name": "body_joints", "l2": 6.163981, "mean_abs": 0.090264, "relative": 0.832564 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010942, "mean_abs": 0.00172, "relative": 0.025822 }, { "name": "camera_rotation_matrix", "l2": 2.999998, "mean_abs": 0.198386, "relative": 0.921251 }, { "name": "imu_accel_gyro", "l2": 17.05681, "mean_abs": 1.014578, "relative": 0.920274 }, { "name": "depth_confidence", "l2": 10.397491, "mean_abs": 0.188941, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.124094, "mean_abs": 0.105884, "relative": 0.927864 }, { "name": "video_fisheye_cam1", "l2": 5.541658, "mean_abs": 0.115117, "relative": 0.915712 }, { "name": "video_fisheye_cam2", "l2": 5.613585, "mean_abs": 0.114852, "relative": 0.896125 }, { "name": "video_fisheye_cam3", "l2": 5.663212, "mean_abs": 0.114698, "relative": 0.948237 }, { "name": "video_stereo_left", "l2": 5.615895, "mean_abs": 0.12101, "relative": 0.912594 }, { "name": "video_stereo_right", "l2": 6.918857, "mean_abs": 0.148191, "relative": 0.975414 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.057906, "mean_abs": 0.081746, "relative": 0.736494 }, { "name": "caption_objects_interaction_text", "l2": 1.828079, "mean_abs": 0.014009, "relative": 0.83941 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9574429988861084" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.3126557767391205" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9990620017051697" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9175865054130554" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1040, "start_frame": 5200, "end_frame": 5219, "center_frame": 5209, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.380814, "mean_abs": 0.248003, "relative": 0.875666 }, { "name": "hand_right_joints", "l2": 9.849945, "mean_abs": 0.263763, "relative": 0.776439 }, { "name": "body_joints", "l2": 6.171444, "mean_abs": 0.089567, "relative": 0.833572 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008754, "mean_abs": 0.001413, "relative": 0.020658 }, { "name": "camera_rotation_matrix", "l2": 3.000029, "mean_abs": 0.198276, "relative": 0.92126 }, { "name": "imu_accel_gyro", "l2": 17.053528, "mean_abs": 1.010938, "relative": 0.920097 }, { "name": "depth_confidence", "l2": 10.34633, "mean_abs": 0.188363, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.12631, "mean_abs": 0.105894, "relative": 0.928265 }, { "name": "video_fisheye_cam1", "l2": 5.54394, "mean_abs": 0.115146, "relative": 0.916089 }, { "name": "video_fisheye_cam2", "l2": 5.61113, "mean_abs": 0.114814, "relative": 0.895733 }, { "name": "video_fisheye_cam3", "l2": 5.65275, "mean_abs": 0.113994, "relative": 0.946485 }, { "name": "video_stereo_left", "l2": 5.621013, "mean_abs": 0.12113, "relative": 0.913426 }, { "name": "video_stereo_right", "l2": 6.937631, "mean_abs": 0.148491, "relative": 0.97806 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.989278, "mean_abs": 0.078363, "relative": 0.711933 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9510948657989502" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6079044938087463" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9997900128364563" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8682889342308044" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1041, "start_frame": 5205, "end_frame": 5224, "center_frame": 5214, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.367294, "mean_abs": 0.245771, "relative": 0.874525 }, { "name": "hand_right_joints", "l2": 9.83267, "mean_abs": 0.261793, "relative": 0.775077 }, { "name": "body_joints", "l2": 6.165586, "mean_abs": 0.087927, "relative": 0.832781 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008923, "mean_abs": 0.001455, "relative": 0.021057 }, { "name": "camera_rotation_matrix", "l2": 3.000119, "mean_abs": 0.198631, "relative": 0.921288 }, { "name": "imu_accel_gyro", "l2": 17.066198, "mean_abs": 1.014258, "relative": 0.92078 }, { "name": "depth_confidence", "l2": 10.331429, "mean_abs": 0.188334, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.129441, "mean_abs": 0.106007, "relative": 0.928832 }, { "name": "video_fisheye_cam1", "l2": 5.553654, "mean_abs": 0.115372, "relative": 0.917694 }, { "name": "video_fisheye_cam2", "l2": 5.610107, "mean_abs": 0.115055, "relative": 0.89557 }, { "name": "video_fisheye_cam3", "l2": 5.669133, "mean_abs": 0.114356, "relative": 0.949228 }, { "name": "video_stereo_left", "l2": 5.632555, "mean_abs": 0.121508, "relative": 0.915301 }, { "name": "video_stereo_right", "l2": 6.96683, "mean_abs": 0.149141, "relative": 0.982177 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.068682, "mean_abs": 0.084762, "relative": 0.74035 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9244909286499023" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6202170252799988" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.993901252746582" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7778785228729248" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1042, "start_frame": 5210, "end_frame": 5229, "center_frame": 5219, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.357018, "mean_abs": 0.243781, "relative": 0.873659 }, { "name": "hand_right_joints", "l2": 9.822715, "mean_abs": 0.260033, "relative": 0.774293 }, { "name": "body_joints", "l2": 6.161147, "mean_abs": 0.086746, "relative": 0.832181 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012011, "mean_abs": 0.001874, "relative": 0.028345 }, { "name": "camera_rotation_matrix", "l2": 3.00015, "mean_abs": 0.199312, "relative": 0.921297 }, { "name": "imu_accel_gyro", "l2": 17.048223, "mean_abs": 1.018813, "relative": 0.91981 }, { "name": "depth_confidence", "l2": 10.313481, "mean_abs": 0.188094, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.133297, "mean_abs": 0.106263, "relative": 0.92953 }, { "name": "video_fisheye_cam1", "l2": 5.553103, "mean_abs": 0.11575, "relative": 0.917603 }, { "name": "video_fisheye_cam2", "l2": 5.610323, "mean_abs": 0.115459, "relative": 0.895605 }, { "name": "video_fisheye_cam3", "l2": 5.677729, "mean_abs": 0.114717, "relative": 0.950667 }, { "name": "video_stereo_left", "l2": 5.645068, "mean_abs": 0.122173, "relative": 0.917335 }, { "name": "video_stereo_right", "l2": 6.98195, "mean_abs": 0.149857, "relative": 0.984308 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.056293, "mean_abs": 0.07909, "relative": 0.735916 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9325146675109863" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6726285815238953" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9654051661491394" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7930545210838318" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1043, "start_frame": 5215, "end_frame": 5234, "center_frame": 5224, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.354485, "mean_abs": 0.243678, "relative": 0.873445 }, { "name": "hand_right_joints", "l2": 9.803987, "mean_abs": 0.258854, "relative": 0.772816 }, { "name": "body_joints", "l2": 6.145976, "mean_abs": 0.086473, "relative": 0.830132 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081684, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.02026, "mean_abs": 0.003087, "relative": 0.047811 }, { "name": "camera_rotation_matrix", "l2": 3.000242, "mean_abs": 0.200091, "relative": 0.921326 }, { "name": "imu_accel_gyro", "l2": 17.039797, "mean_abs": 1.027161, "relative": 0.919356 }, { "name": "depth_confidence", "l2": 10.296949, "mean_abs": 0.187883, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.134974, "mean_abs": 0.106497, "relative": 0.929834 }, { "name": "video_fisheye_cam1", "l2": 5.540941, "mean_abs": 0.115865, "relative": 0.915594 }, { "name": "video_fisheye_cam2", "l2": 5.609387, "mean_abs": 0.115744, "relative": 0.895455 }, { "name": "video_fisheye_cam3", "l2": 5.67371, "mean_abs": 0.114847, "relative": 0.949994 }, { "name": "video_stereo_left", "l2": 5.655874, "mean_abs": 0.122875, "relative": 0.919091 }, { "name": "video_stereo_right", "l2": 6.982699, "mean_abs": 0.150302, "relative": 0.984414 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.07676, "mean_abs": 0.082267, "relative": 0.743241 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9264324903488159" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6294464468955994" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9088281989097595" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.786862850189209" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1044, "start_frame": 5220, "end_frame": 5239, "center_frame": 5229, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.332777, "mean_abs": 0.242839, "relative": 0.871614 }, { "name": "hand_right_joints", "l2": 9.824981, "mean_abs": 0.258533, "relative": 0.774471 }, { "name": "body_joints", "l2": 6.134087, "mean_abs": 0.086077, "relative": 0.828526 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021494, "mean_abs": 0.00327, "relative": 0.050723 }, { "name": "camera_rotation_matrix", "l2": 3.000511, "mean_abs": 0.200693, "relative": 0.921408 }, { "name": "imu_accel_gyro", "l2": 17.043318, "mean_abs": 1.027969, "relative": 0.919546 }, { "name": "depth_confidence", "l2": 10.331409, "mean_abs": 0.187787, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.136966, "mean_abs": 0.106586, "relative": 0.930195 }, { "name": "video_fisheye_cam1", "l2": 5.539077, "mean_abs": 0.116105, "relative": 0.915286 }, { "name": "video_fisheye_cam2", "l2": 5.610186, "mean_abs": 0.115987, "relative": 0.895583 }, { "name": "video_fisheye_cam3", "l2": 5.682464, "mean_abs": 0.115217, "relative": 0.95146 }, { "name": "video_stereo_left", "l2": 5.671762, "mean_abs": 0.123318, "relative": 0.921673 }, { "name": "video_stereo_right", "l2": 7.000936, "mean_abs": 0.150658, "relative": 0.986985 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.099504, "mean_abs": 0.083987, "relative": 0.751381 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9321900010108948" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6062692999839783" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.841119110584259" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8243280649185181" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1045, "start_frame": 5225, "end_frame": 5244, "center_frame": 5234, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.349012, "mean_abs": 0.243582, "relative": 0.872983 }, { "name": "hand_right_joints", "l2": 9.824013, "mean_abs": 0.258266, "relative": 0.774395 }, { "name": "body_joints", "l2": 6.127208, "mean_abs": 0.085925, "relative": 0.827597 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018318, "mean_abs": 0.002816, "relative": 0.043227 }, { "name": "camera_rotation_matrix", "l2": 3.000405, "mean_abs": 0.200431, "relative": 0.921376 }, { "name": "imu_accel_gyro", "l2": 17.064667, "mean_abs": 1.033912, "relative": 0.920698 }, { "name": "depth_confidence", "l2": 10.315906, "mean_abs": 0.187118, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.134714, "mean_abs": 0.10656, "relative": 0.929787 }, { "name": "video_fisheye_cam1", "l2": 5.533385, "mean_abs": 0.115947, "relative": 0.914345 }, { "name": "video_fisheye_cam2", "l2": 5.607499, "mean_abs": 0.115895, "relative": 0.895154 }, { "name": "video_fisheye_cam3", "l2": 5.681576, "mean_abs": 0.115434, "relative": 0.951311 }, { "name": "video_stereo_left", "l2": 5.683946, "mean_abs": 0.123268, "relative": 0.923653 }, { "name": "video_stereo_right", "l2": 7.011633, "mean_abs": 0.150618, "relative": 0.988493 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.109374, "mean_abs": 0.084924, "relative": 0.754913 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9436421394348145" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6079227328300476" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9626895189285278" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.862740695476532" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|bottle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1046, "start_frame": 5230, "end_frame": 5249, "center_frame": 5239, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.351114, "mean_abs": 0.243175, "relative": 0.873161 }, { "name": "hand_right_joints", "l2": 9.803174, "mean_abs": 0.258142, "relative": 0.772752 }, { "name": "body_joints", "l2": 6.1222, "mean_abs": 0.086042, "relative": 0.826921 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081685, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.016097, "mean_abs": 0.002319, "relative": 0.037987 }, { "name": "camera_rotation_matrix", "l2": 3.000217, "mean_abs": 0.19996, "relative": 0.921318 }, { "name": "imu_accel_gyro", "l2": 17.049665, "mean_abs": 1.036164, "relative": 0.919888 }, { "name": "depth_confidence", "l2": 10.314993, "mean_abs": 0.186707, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.132891, "mean_abs": 0.106417, "relative": 0.929457 }, { "name": "video_fisheye_cam1", "l2": 5.518222, "mean_abs": 0.115291, "relative": 0.91184 }, { "name": "video_fisheye_cam2", "l2": 5.605158, "mean_abs": 0.115628, "relative": 0.89478 }, { "name": "video_fisheye_cam3", "l2": 5.671386, "mean_abs": 0.11495, "relative": 0.949605 }, { "name": "video_stereo_left", "l2": 5.694185, "mean_abs": 0.123114, "relative": 0.925316 }, { "name": "video_stereo_right", "l2": 7.013878, "mean_abs": 0.150255, "relative": 0.98881 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.077188, "mean_abs": 0.084078, "relative": 0.743394 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9561206698417664" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7806938886642456" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9852851033210754" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8772063851356506" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1047, "start_frame": 5235, "end_frame": 5254, "center_frame": 5244, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.353366, "mean_abs": 0.242775, "relative": 0.873351 }, { "name": "hand_right_joints", "l2": 9.781068, "mean_abs": 0.257453, "relative": 0.77101 }, { "name": "body_joints", "l2": 6.116924, "mean_abs": 0.086192, "relative": 0.826208 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010864, "mean_abs": 0.001595, "relative": 0.025637 }, { "name": "camera_rotation_matrix", "l2": 3.000182, "mean_abs": 0.199858, "relative": 0.921307 }, { "name": "imu_accel_gyro", "l2": 17.020609, "mean_abs": 1.033752, "relative": 0.91832 }, { "name": "depth_confidence", "l2": 10.312437, "mean_abs": 0.186628, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.146866, "mean_abs": 0.106881, "relative": 0.931987 }, { "name": "video_fisheye_cam1", "l2": 5.522872, "mean_abs": 0.11535, "relative": 0.912608 }, { "name": "video_fisheye_cam2", "l2": 5.603583, "mean_abs": 0.115498, "relative": 0.894529 }, { "name": "video_fisheye_cam3", "l2": 5.673168, "mean_abs": 0.115009, "relative": 0.949904 }, { "name": "video_stereo_left", "l2": 5.703619, "mean_abs": 0.123126, "relative": 0.926849 }, { "name": "video_stereo_right", "l2": 7.040266, "mean_abs": 0.150598, "relative": 0.99253 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.066093, "mean_abs": 0.082511, "relative": 0.739424 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.956155002117157" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8003977537155151" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9899033308029175" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8860884308815002" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1048, "start_frame": 5240, "end_frame": 5259, "center_frame": 5249, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.38205, "mean_abs": 0.243597, "relative": 0.87577 }, { "name": "hand_right_joints", "l2": 9.76367, "mean_abs": 0.256919, "relative": 0.769638 }, { "name": "body_joints", "l2": 6.119327, "mean_abs": 0.086758, "relative": 0.826532 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.015167, "mean_abs": 0.002244, "relative": 0.035793 }, { "name": "camera_rotation_matrix", "l2": 3.000277, "mean_abs": 0.200256, "relative": 0.921336 }, { "name": "imu_accel_gyro", "l2": 17.033834, "mean_abs": 1.036058, "relative": 0.919034 }, { "name": "depth_confidence", "l2": 10.277674, "mean_abs": 0.185486, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.153459, "mean_abs": 0.107437, "relative": 0.933181 }, { "name": "video_fisheye_cam1", "l2": 5.518421, "mean_abs": 0.115346, "relative": 0.911873 }, { "name": "video_fisheye_cam2", "l2": 5.604711, "mean_abs": 0.115623, "relative": 0.894709 }, { "name": "video_fisheye_cam3", "l2": 5.669032, "mean_abs": 0.115122, "relative": 0.949211 }, { "name": "video_stereo_left", "l2": 5.718983, "mean_abs": 0.123642, "relative": 0.929346 }, { "name": "video_stereo_right", "l2": 7.048932, "mean_abs": 0.150923, "relative": 0.993752 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.087312, "mean_abs": 0.082756, "relative": 0.747018 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9595621228218079" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8625264763832092" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9684026837348938" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8566864132881165" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1049, "start_frame": 5245, "end_frame": 5264, "center_frame": 5254, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.404362, "mean_abs": 0.245202, "relative": 0.877652 }, { "name": "hand_right_joints", "l2": 9.731283, "mean_abs": 0.25585, "relative": 0.767085 }, { "name": "body_joints", "l2": 6.118363, "mean_abs": 0.087229, "relative": 0.826402 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.022497, "mean_abs": 0.003194, "relative": 0.053089 }, { "name": "camera_rotation_matrix", "l2": 3.000506, "mean_abs": 0.201819, "relative": 0.921407 }, { "name": "imu_accel_gyro", "l2": 17.041676, "mean_abs": 1.049088, "relative": 0.919457 }, { "name": "depth_confidence", "l2": 10.275496, "mean_abs": 0.185441, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.16362, "mean_abs": 0.107645, "relative": 0.935021 }, { "name": "video_fisheye_cam1", "l2": 5.521325, "mean_abs": 0.115936, "relative": 0.912352 }, { "name": "video_fisheye_cam2", "l2": 5.607628, "mean_abs": 0.116114, "relative": 0.895174 }, { "name": "video_fisheye_cam3", "l2": 5.671489, "mean_abs": 0.115458, "relative": 0.949622 }, { "name": "video_stereo_left", "l2": 5.741484, "mean_abs": 0.12458, "relative": 0.933003 }, { "name": "video_stereo_right", "l2": 7.071153, "mean_abs": 0.152101, "relative": 0.996884 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.093371, "mean_abs": 0.085543, "relative": 0.749186 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9276778697967529" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.870999813079834" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.8572291731834412" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8109790682792664" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1050, "start_frame": 5250, "end_frame": 5269, "center_frame": 5259, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.419634, "mean_abs": 0.245471, "relative": 0.878941 }, { "name": "hand_right_joints", "l2": 9.687919, "mean_abs": 0.254448, "relative": 0.763667 }, { "name": "body_joints", "l2": 6.103831, "mean_abs": 0.087401, "relative": 0.824439 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.027332, "mean_abs": 0.003814, "relative": 0.064501 }, { "name": "camera_rotation_matrix", "l2": 3.000538, "mean_abs": 0.20273, "relative": 0.921416 }, { "name": "imu_accel_gyro", "l2": 17.035755, "mean_abs": 1.052755, "relative": 0.919138 }, { "name": "depth_confidence", "l2": 10.256258, "mean_abs": 0.184825, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.169508, "mean_abs": 0.107648, "relative": 0.936087 }, { "name": "video_fisheye_cam1", "l2": 5.516972, "mean_abs": 0.116294, "relative": 0.911633 }, { "name": "video_fisheye_cam2", "l2": 5.609025, "mean_abs": 0.116373, "relative": 0.895397 }, { "name": "video_fisheye_cam3", "l2": 5.658373, "mean_abs": 0.115766, "relative": 0.947426 }, { "name": "video_stereo_left", "l2": 5.751332, "mean_abs": 0.125095, "relative": 0.934603 }, { "name": "video_stereo_right", "l2": 7.076053, "mean_abs": 0.152472, "relative": 0.997575 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.043708, "mean_abs": 0.081853, "relative": 0.731413 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9393765330314636" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.887285590171814" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.8631205558776855" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.79525226354599" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1051, "start_frame": 5255, "end_frame": 5274, "center_frame": 5264, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.436685, "mean_abs": 0.246118, "relative": 0.880379 }, { "name": "hand_right_joints", "l2": 9.640462, "mean_abs": 0.25305, "relative": 0.759926 }, { "name": "body_joints", "l2": 6.08586, "mean_abs": 0.087887, "relative": 0.822012 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.022631, "mean_abs": 0.003318, "relative": 0.053407 }, { "name": "camera_rotation_matrix", "l2": 3.000545, "mean_abs": 0.202843, "relative": 0.921419 }, { "name": "imu_accel_gyro", "l2": 17.023544, "mean_abs": 1.052385, "relative": 0.918479 }, { "name": "depth_confidence", "l2": 10.246074, "mean_abs": 0.184997, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.179079, "mean_abs": 0.107626, "relative": 0.93782 }, { "name": "video_fisheye_cam1", "l2": 5.506785, "mean_abs": 0.116323, "relative": 0.90995 }, { "name": "video_fisheye_cam2", "l2": 5.60855, "mean_abs": 0.11626, "relative": 0.895321 }, { "name": "video_fisheye_cam3", "l2": 5.631431, "mean_abs": 0.115151, "relative": 0.942915 }, { "name": "video_stereo_left", "l2": 5.756205, "mean_abs": 0.125227, "relative": 0.935395 }, { "name": "video_stereo_right", "l2": 7.062797, "mean_abs": 0.152479, "relative": 0.995706 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.085383, "mean_abs": 0.083716, "relative": 0.746327 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9374621510505676" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9742437601089478" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.7359462380409241" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6988240480422974" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1052, "start_frame": 5260, "end_frame": 5279, "center_frame": 5269, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.45478, "mean_abs": 0.24608, "relative": 0.881905 }, { "name": "hand_right_joints", "l2": 9.635947, "mean_abs": 0.252589, "relative": 0.75957 }, { "name": "body_joints", "l2": 6.069253, "mean_abs": 0.087877, "relative": 0.819769 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013655, "mean_abs": 0.001902, "relative": 0.032225 }, { "name": "camera_rotation_matrix", "l2": 3.000285, "mean_abs": 0.20204, "relative": 0.921339 }, { "name": "imu_accel_gyro", "l2": 17.030918, "mean_abs": 1.055939, "relative": 0.918877 }, { "name": "depth_confidence", "l2": 10.248824, "mean_abs": 0.184719, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.175908, "mean_abs": 0.107277, "relative": 0.937246 }, { "name": "video_fisheye_cam1", "l2": 5.481886, "mean_abs": 0.115179, "relative": 0.905835 }, { "name": "video_fisheye_cam2", "l2": 5.605621, "mean_abs": 0.115764, "relative": 0.894854 }, { "name": "video_fisheye_cam3", "l2": 5.602376, "mean_abs": 0.113765, "relative": 0.93805 }, { "name": "video_stereo_left", "l2": 5.759084, "mean_abs": 0.12463, "relative": 0.935863 }, { "name": "video_stereo_right", "l2": 7.032859, "mean_abs": 0.150729, "relative": 0.991486 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.06712, "mean_abs": 0.083165, "relative": 0.739791 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9655011296272278" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9896154403686523" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.5320137143135071" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7636183500289917" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1053, "start_frame": 5265, "end_frame": 5284, "center_frame": 5274, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.454514, "mean_abs": 0.245343, "relative": 0.881883 }, { "name": "hand_right_joints", "l2": 9.64205, "mean_abs": 0.251496, "relative": 0.760051 }, { "name": "body_joints", "l2": 6.050186, "mean_abs": 0.087273, "relative": 0.817194 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006534, "mean_abs": 0.000994, "relative": 0.01542 }, { "name": "camera_rotation_matrix", "l2": 3.000104, "mean_abs": 0.20048, "relative": 0.921283 }, { "name": "imu_accel_gyro", "l2": 17.033054, "mean_abs": 1.053865, "relative": 0.918992 }, { "name": "depth_confidence", "l2": 10.266644, "mean_abs": 0.184425, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.170536, "mean_abs": 0.106839, "relative": 0.936273 }, { "name": "video_fisheye_cam1", "l2": 5.473751, "mean_abs": 0.114141, "relative": 0.904491 }, { "name": "video_fisheye_cam2", "l2": 5.604329, "mean_abs": 0.115022, "relative": 0.894648 }, { "name": "video_fisheye_cam3", "l2": 5.595052, "mean_abs": 0.11326, "relative": 0.936824 }, { "name": "video_stereo_left", "l2": 5.759051, "mean_abs": 0.123537, "relative": 0.935857 }, { "name": "video_stereo_right", "l2": 7.021667, "mean_abs": 0.149539, "relative": 0.989908 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.036245, "mean_abs": 0.078346, "relative": 0.728741 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9857827425003052" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9931135773658752" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.8076270222663879" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8485909700393677" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1054, "start_frame": 5270, "end_frame": 5289, "center_frame": 5279, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.45444, "mean_abs": 0.245174, "relative": 0.881877 }, { "name": "hand_right_joints", "l2": 9.643448, "mean_abs": 0.249456, "relative": 0.760162 }, { "name": "body_joints", "l2": 6.037443, "mean_abs": 0.086787, "relative": 0.815472 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01413, "mean_abs": 0.002093, "relative": 0.033345 }, { "name": "camera_rotation_matrix", "l2": 3.000082, "mean_abs": 0.201312, "relative": 0.921276 }, { "name": "imu_accel_gyro", "l2": 17.040308, "mean_abs": 1.058213, "relative": 0.919383 }, { "name": "depth_confidence", "l2": 10.279537, "mean_abs": 0.184036, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.191987, "mean_abs": 0.107686, "relative": 0.940158 }, { "name": "video_fisheye_cam1", "l2": 5.463868, "mean_abs": 0.114473, "relative": 0.902858 }, { "name": "video_fisheye_cam2", "l2": 5.601169, "mean_abs": 0.115346, "relative": 0.894143 }, { "name": "video_fisheye_cam3", "l2": 5.584556, "mean_abs": 0.11357, "relative": 0.935067 }, { "name": "video_stereo_left", "l2": 5.753353, "mean_abs": 0.12399, "relative": 0.934931 }, { "name": "video_stereo_right", "l2": 6.998295, "mean_abs": 0.149904, "relative": 0.986613 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.085083, "mean_abs": 0.083832, "relative": 0.74622 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9858607053756714" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9921494722366333" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9520547986030579" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8372886776924133" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1055, "start_frame": 5275, "end_frame": 5294, "center_frame": 5284, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.450711, "mean_abs": 0.246916, "relative": 0.881562 }, { "name": "hand_right_joints", "l2": 9.693129, "mean_abs": 0.25582, "relative": 0.764078 }, { "name": "body_joints", "l2": 6.095257, "mean_abs": 0.089227, "relative": 0.823281 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017689, "mean_abs": 0.002579, "relative": 0.041743 }, { "name": "camera_rotation_matrix", "l2": 2.999988, "mean_abs": 0.201501, "relative": 0.921248 }, { "name": "imu_accel_gyro", "l2": 17.042404, "mean_abs": 1.05223, "relative": 0.919496 }, { "name": "depth_confidence", "l2": 10.284295, "mean_abs": 0.18539, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.196185, "mean_abs": 0.107836, "relative": 0.940918 }, { "name": "video_fisheye_cam1", "l2": 5.447802, "mean_abs": 0.114881, "relative": 0.900203 }, { "name": "video_fisheye_cam2", "l2": 5.597672, "mean_abs": 0.115556, "relative": 0.893585 }, { "name": "video_fisheye_cam3", "l2": 5.566414, "mean_abs": 0.11392, "relative": 0.932029 }, { "name": "video_stereo_left", "l2": 5.751789, "mean_abs": 0.124518, "relative": 0.934677 }, { "name": "video_stereo_right", "l2": 6.973277, "mean_abs": 0.151024, "relative": 0.983086 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.249591, "mean_abs": 0.095536, "relative": 0.805095 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9099838733673096" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9837228655815125" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9966501593589783" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7520685791969299" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1056, "start_frame": 5280, "end_frame": 5299, "center_frame": 5289, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.432382, "mean_abs": 0.247275, "relative": 0.880016 }, { "name": "hand_right_joints", "l2": 9.749978, "mean_abs": 0.261417, "relative": 0.768559 }, { "name": "body_joints", "l2": 6.164107, "mean_abs": 0.091344, "relative": 0.832581 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028031, "mean_abs": 0.004001, "relative": 0.06615 }, { "name": "camera_rotation_matrix", "l2": 3.000637, "mean_abs": 0.202951, "relative": 0.921447 }, { "name": "imu_accel_gyro", "l2": 17.042063, "mean_abs": 1.06925, "relative": 0.919478 }, { "name": "depth_confidence", "l2": 10.305667, "mean_abs": 0.18735, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.200287, "mean_abs": 0.108262, "relative": 0.941661 }, { "name": "video_fisheye_cam1", "l2": 5.435342, "mean_abs": 0.115547, "relative": 0.898144 }, { "name": "video_fisheye_cam2", "l2": 5.593931, "mean_abs": 0.116212, "relative": 0.892988 }, { "name": "video_fisheye_cam3", "l2": 5.540216, "mean_abs": 0.113962, "relative": 0.927642 }, { "name": "video_stereo_left", "l2": 5.736097, "mean_abs": 0.12575, "relative": 0.932127 }, { "name": "video_stereo_right", "l2": 6.927145, "mean_abs": 0.152368, "relative": 0.976582 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.176095, "mean_abs": 0.088733, "relative": 0.778792 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9000096321105957" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9741169214248657" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9993910789489746" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.671899676322937" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1057, "start_frame": 5285, "end_frame": 5304, "center_frame": 5294, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.401669, "mean_abs": 0.245515, "relative": 0.877425 }, { "name": "hand_right_joints", "l2": 9.811938, "mean_abs": 0.265287, "relative": 0.773443 }, { "name": "body_joints", "l2": 6.217669, "mean_abs": 0.092561, "relative": 0.839815 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.038629, "mean_abs": 0.005574, "relative": 0.091159 }, { "name": "camera_rotation_matrix", "l2": 3.001752, "mean_abs": 0.204678, "relative": 0.921789 }, { "name": "imu_accel_gyro", "l2": 17.054104, "mean_abs": 1.062675, "relative": 0.920128 }, { "name": "depth_confidence", "l2": 10.316301, "mean_abs": 0.188023, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.211277, "mean_abs": 0.10907, "relative": 0.943651 }, { "name": "video_fisheye_cam1", "l2": 5.41811, "mean_abs": 0.116197, "relative": 0.895297 }, { "name": "video_fisheye_cam2", "l2": 5.592926, "mean_abs": 0.117426, "relative": 0.892827 }, { "name": "video_fisheye_cam3", "l2": 5.505134, "mean_abs": 0.114298, "relative": 0.921768 }, { "name": "video_stereo_left", "l2": 5.714581, "mean_abs": 0.12727, "relative": 0.928631 }, { "name": "video_stereo_right", "l2": 6.850964, "mean_abs": 0.153821, "relative": 0.965842 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.128677, "mean_abs": 0.088255, "relative": 0.761821 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8803130984306335" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9578664302825928" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9997581839561462" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5274430513381958" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1058, "start_frame": 5290, "end_frame": 5309, "center_frame": 5299, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.361047, "mean_abs": 0.246457, "relative": 0.873999 }, { "name": "hand_right_joints", "l2": 9.874454, "mean_abs": 0.267937, "relative": 0.778371 }, { "name": "body_joints", "l2": 6.248177, "mean_abs": 0.093821, "relative": 0.843936 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.072326, "mean_abs": 0.010139, "relative": 0.17068 }, { "name": "camera_rotation_matrix", "l2": 3.007396, "mean_abs": 0.209122, "relative": 0.923522 }, { "name": "imu_accel_gyro", "l2": 17.122114, "mean_abs": 1.101838, "relative": 0.923797 }, { "name": "depth_confidence", "l2": 10.375411, "mean_abs": 0.192167, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.225494, "mean_abs": 0.110271, "relative": 0.946225 }, { "name": "video_fisheye_cam1", "l2": 5.393793, "mean_abs": 0.116911, "relative": 0.891279 }, { "name": "video_fisheye_cam2", "l2": 5.611271, "mean_abs": 0.119663, "relative": 0.895756 }, { "name": "video_fisheye_cam3", "l2": 5.48147, "mean_abs": 0.114556, "relative": 0.917806 }, { "name": "video_stereo_left", "l2": 5.709112, "mean_abs": 0.130924, "relative": 0.927742 }, { "name": "video_stereo_right", "l2": 6.739246, "mean_abs": 0.15521, "relative": 0.950092 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.122179, "mean_abs": 0.087739, "relative": 0.759496 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9293034076690674" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9268026947975159" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999958276748657" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.6570851802825928" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1059, "start_frame": 5295, "end_frame": 5314, "center_frame": 5304, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.31529, "mean_abs": 0.24648, "relative": 0.870139 }, { "name": "hand_right_joints", "l2": 10.095981, "mean_abs": 0.275686, "relative": 0.795833 }, { "name": "body_joints", "l2": 6.237948, "mean_abs": 0.094798, "relative": 0.842555 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.129527, "mean_abs": 0.018063, "relative": 0.305666 }, { "name": "camera_rotation_matrix", "l2": 3.02477, "mean_abs": 0.214685, "relative": 0.928858 }, { "name": "imu_accel_gyro", "l2": 17.101593, "mean_abs": 1.11035, "relative": 0.92269 }, { "name": "depth_confidence", "l2": 10.428179, "mean_abs": 0.196864, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.242484, "mean_abs": 0.111884, "relative": 0.949302 }, { "name": "video_fisheye_cam1", "l2": 5.355768, "mean_abs": 0.118116, "relative": 0.884995 }, { "name": "video_fisheye_cam2", "l2": 5.609205, "mean_abs": 0.121454, "relative": 0.895426 }, { "name": "video_fisheye_cam3", "l2": 5.420972, "mean_abs": 0.114943, "relative": 0.907676 }, { "name": "video_stereo_left", "l2": 5.686411, "mean_abs": 0.133625, "relative": 0.924053 }, { "name": "video_stereo_right", "l2": 6.596065, "mean_abs": 0.155897, "relative": 0.929907 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.118235, "mean_abs": 0.088226, "relative": 0.758084 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.956604540348053" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8842471837997437" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.7473781704902649" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1060, "start_frame": 5300, "end_frame": 5319, "center_frame": 5309, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.270191, "mean_abs": 0.24277, "relative": 0.866334 }, { "name": "hand_right_joints", "l2": 10.36237, "mean_abs": 0.283635, "relative": 0.816832 }, { "name": "body_joints", "l2": 6.186045, "mean_abs": 0.095158, "relative": 0.835544 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.156765, "mean_abs": 0.021869, "relative": 0.369944 }, { "name": "camera_rotation_matrix", "l2": 3.030828, "mean_abs": 0.214783, "relative": 0.930718 }, { "name": "imu_accel_gyro", "l2": 17.184353, "mean_abs": 1.106919, "relative": 0.927155 }, { "name": "depth_confidence", "l2": 10.494113, "mean_abs": 0.199397, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.226601, "mean_abs": 0.112059, "relative": 0.946426 }, { "name": "video_fisheye_cam1", "l2": 5.302505, "mean_abs": 0.11846, "relative": 0.876194 }, { "name": "video_fisheye_cam2", "l2": 5.567806, "mean_abs": 0.120588, "relative": 0.888817 }, { "name": "video_fisheye_cam3", "l2": 5.39172, "mean_abs": 0.114624, "relative": 0.902779 }, { "name": "video_stereo_left", "l2": 5.603753, "mean_abs": 0.132985, "relative": 0.910621 }, { "name": "video_stereo_right", "l2": 6.409083, "mean_abs": 0.152383, "relative": 0.903546 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.129365, "mean_abs": 0.088128, "relative": 0.762068 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.9262050986289978" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7608693242073059" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.7772440314292908" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1061, "start_frame": 5305, "end_frame": 5324, "center_frame": 5314, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.183886, "mean_abs": 0.237668, "relative": 0.859054 }, { "name": "hand_right_joints", "l2": 10.491384, "mean_abs": 0.282108, "relative": 0.827002 }, { "name": "body_joints", "l2": 6.114621, "mean_abs": 0.093071, "relative": 0.825897 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.162626, "mean_abs": 0.022584, "relative": 0.383774 }, { "name": "camera_rotation_matrix", "l2": 3.022906, "mean_abs": 0.211136, "relative": 0.928285 }, { "name": "imu_accel_gyro", "l2": 17.049913, "mean_abs": 1.10559, "relative": 0.919902 }, { "name": "depth_confidence", "l2": 10.565925, "mean_abs": 0.19889, "relative": 0.000351 }, { "name": "video_fisheye_cam0", "l2": 5.212733, "mean_abs": 0.111211, "relative": 0.943914 }, { "name": "video_fisheye_cam1", "l2": 5.258226, "mean_abs": 0.117545, "relative": 0.868878 }, { "name": "video_fisheye_cam2", "l2": 5.515561, "mean_abs": 0.118915, "relative": 0.880477 }, { "name": "video_fisheye_cam3", "l2": 5.380619, "mean_abs": 0.113761, "relative": 0.90092 }, { "name": "video_stereo_left", "l2": 5.475925, "mean_abs": 0.1302, "relative": 0.889849 }, { "name": "video_stereo_right", "l2": 6.206464, "mean_abs": 0.146899, "relative": 0.874981 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.042951, "mean_abs": 0.081595, "relative": 0.731141 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8920544385910034" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.5887539982795715" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.4751960337162018" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1062, "start_frame": 5310, "end_frame": 5329, "center_frame": 5319, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.095857, "mean_abs": 0.232352, "relative": 0.851629 }, { "name": "hand_right_joints", "l2": 10.956627, "mean_abs": 0.294274, "relative": 0.863675 }, { "name": "body_joints", "l2": 6.16194, "mean_abs": 0.095543, "relative": 0.832288 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.153715, "mean_abs": 0.020468, "relative": 0.362747 }, { "name": "camera_rotation_matrix", "l2": 3.006174, "mean_abs": 0.200299, "relative": 0.923147 }, { "name": "imu_accel_gyro", "l2": 17.095976, "mean_abs": 1.09659, "relative": 0.922387 }, { "name": "depth_confidence", "l2": 10.555822, "mean_abs": 0.195936, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.184336, "mean_abs": 0.109101, "relative": 0.938772 }, { "name": "video_fisheye_cam1", "l2": 5.208056, "mean_abs": 0.11427, "relative": 0.860587 }, { "name": "video_fisheye_cam2", "l2": 5.453311, "mean_abs": 0.115176, "relative": 0.87054 }, { "name": "video_fisheye_cam3", "l2": 5.380761, "mean_abs": 0.112216, "relative": 0.900943 }, { "name": "video_stereo_left", "l2": 5.270944, "mean_abs": 0.122227, "relative": 0.856539 }, { "name": "video_stereo_right", "l2": 5.964655, "mean_abs": 0.138901, "relative": 0.840891 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.083838, "mean_abs": 0.085246, "relative": 0.745774 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6224517822265625" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.3918447196483612" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999979734420776" }, "next_action": { "true": "Pour coffee", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.3142651617527008" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "kettle|coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1063, "start_frame": 5315, "end_frame": 5334, "center_frame": 5324, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.057799, "mean_abs": 0.231863, "relative": 0.848418 }, { "name": "hand_right_joints", "l2": 11.449771, "mean_abs": 0.309316, "relative": 0.902548 }, { "name": "body_joints", "l2": 6.155567, "mean_abs": 0.097056, "relative": 0.831427 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.140553, "mean_abs": 0.019149, "relative": 0.331686 }, { "name": "camera_rotation_matrix", "l2": 3.004545, "mean_abs": 0.198472, "relative": 0.922647 }, { "name": "imu_accel_gyro", "l2": 17.12631, "mean_abs": 1.099885, "relative": 0.924023 }, { "name": "depth_confidence", "l2": 10.552123, "mean_abs": 0.195765, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.155351, "mean_abs": 0.108129, "relative": 0.933524 }, { "name": "video_fisheye_cam1", "l2": 5.137387, "mean_abs": 0.111177, "relative": 0.84891 }, { "name": "video_fisheye_cam2", "l2": 5.420561, "mean_abs": 0.112557, "relative": 0.865312 }, { "name": "video_fisheye_cam3", "l2": 5.397729, "mean_abs": 0.11286, "relative": 0.903785 }, { "name": "video_stereo_left", "l2": 5.194108, "mean_abs": 0.118376, "relative": 0.844053 }, { "name": "video_stereo_right", "l2": 5.922972, "mean_abs": 0.137485, "relative": 0.835015 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.115307, "mean_abs": 0.088121, "relative": 0.757037 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp gooseneck kettle", "correct": 0, "confidence": "0.45177412033081055" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.8908111453056335" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999998807907104" }, "next_action": { "true": "Pour coffee", "predicted": "Close bottle cap", "correct": 0, "confidence": "0.28911450505256653" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle", "correct": 0, "confidence": "" } } }, { "window_index": 1064, "start_frame": 5320, "end_frame": 5339, "center_frame": 5329, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.040987, "mean_abs": 0.231767, "relative": 0.847 }, { "name": "hand_right_joints", "l2": 11.746106, "mean_abs": 0.320431, "relative": 0.925907 }, { "name": "body_joints", "l2": 6.150333, "mean_abs": 0.097, "relative": 0.830721 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.143048, "mean_abs": 0.019709, "relative": 0.337573 }, { "name": "camera_rotation_matrix", "l2": 3.013784, "mean_abs": 0.208426, "relative": 0.925484 }, { "name": "imu_accel_gyro", "l2": 17.189968, "mean_abs": 1.150826, "relative": 0.927458 }, { "name": "depth_confidence", "l2": 10.564426, "mean_abs": 0.199792, "relative": 0.000351 }, { "name": "video_fisheye_cam0", "l2": 5.173344, "mean_abs": 0.109598, "relative": 0.936782 }, { "name": "video_fisheye_cam1", "l2": 5.13795, "mean_abs": 0.113368, "relative": 0.849003 }, { "name": "video_fisheye_cam2", "l2": 5.4884, "mean_abs": 0.115548, "relative": 0.876141 }, { "name": "video_fisheye_cam3", "l2": 5.425508, "mean_abs": 0.113544, "relative": 0.908436 }, { "name": "video_stereo_left", "l2": 5.273909, "mean_abs": 0.121465, "relative": 0.857021 }, { "name": "video_stereo_right", "l2": 6.064891, "mean_abs": 0.143256, "relative": 0.855022 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.107482, "mean_abs": 0.086336, "relative": 0.754236 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.5319464802742004" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9380927085876465" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.5687623620033264" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1065, "start_frame": 5325, "end_frame": 5344, "center_frame": 5334, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.036447, "mean_abs": 0.230159, "relative": 0.846617 }, { "name": "hand_right_joints", "l2": 11.963735, "mean_abs": 0.326064, "relative": 0.943062 }, { "name": "body_joints", "l2": 6.175089, "mean_abs": 0.09529, "relative": 0.834064 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.134704, "mean_abs": 0.018354, "relative": 0.317883 }, { "name": "camera_rotation_matrix", "l2": 3.020875, "mean_abs": 0.214634, "relative": 0.927662 }, { "name": "imu_accel_gyro", "l2": 17.31432, "mean_abs": 1.206371, "relative": 0.934167 }, { "name": "depth_confidence", "l2": 10.546222, "mean_abs": 0.200416, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.194818, "mean_abs": 0.110467, "relative": 0.940671 }, { "name": "video_fisheye_cam1", "l2": 5.137774, "mean_abs": 0.114502, "relative": 0.848974 }, { "name": "video_fisheye_cam2", "l2": 5.554345, "mean_abs": 0.117529, "relative": 0.886669 }, { "name": "video_fisheye_cam3", "l2": 5.464317, "mean_abs": 0.114319, "relative": 0.914934 }, { "name": "video_stereo_left", "l2": 5.359033, "mean_abs": 0.124228, "relative": 0.870853 }, { "name": "video_stereo_right", "l2": 6.143673, "mean_abs": 0.144799, "relative": 0.866129 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.163288, "mean_abs": 0.090322, "relative": 0.774208 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.554484486579895" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9746870398521423" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9646797180175781" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1066, "start_frame": 5330, "end_frame": 5349, "center_frame": 5339, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.0639, "mean_abs": 0.226272, "relative": 0.848933 }, { "name": "hand_right_joints", "l2": 12.249787, "mean_abs": 0.324169, "relative": 0.965611 }, { "name": "body_joints", "l2": 6.09249, "mean_abs": 0.092714, "relative": 0.822908 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.087261, "mean_abs": 0.011753, "relative": 0.205923 }, { "name": "camera_rotation_matrix", "l2": 3.014245, "mean_abs": 0.213213, "relative": 0.925626 }, { "name": "imu_accel_gyro", "l2": 17.226425, "mean_abs": 1.18465, "relative": 0.929425 }, { "name": "depth_confidence", "l2": 10.475014, "mean_abs": 0.197992, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.200393, "mean_abs": 0.110357, "relative": 0.94168 }, { "name": "video_fisheye_cam1", "l2": 5.108261, "mean_abs": 0.112866, "relative": 0.844097 }, { "name": "video_fisheye_cam2", "l2": 5.589038, "mean_abs": 0.117452, "relative": 0.892207 }, { "name": "video_fisheye_cam3", "l2": 5.502807, "mean_abs": 0.115113, "relative": 0.921379 }, { "name": "video_stereo_left", "l2": 5.369048, "mean_abs": 0.122944, "relative": 0.872481 }, { "name": "video_stereo_right", "l2": 6.153829, "mean_abs": 0.143596, "relative": 0.867561 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.142782, "mean_abs": 0.087901, "relative": 0.766869 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.33902937173843384" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.9770500659942627" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.990291178226471" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1067, "start_frame": 5335, "end_frame": 5354, "center_frame": 5344, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.113012, "mean_abs": 0.220516, "relative": 0.853076 }, { "name": "hand_right_joints", "l2": 12.410623, "mean_abs": 0.319719, "relative": 0.978289 }, { "name": "body_joints", "l2": 5.99226, "mean_abs": 0.088348, "relative": 0.80937 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081673, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.042085, "mean_abs": 0.006393, "relative": 0.099314 }, { "name": "camera_rotation_matrix", "l2": 3.005375, "mean_abs": 0.208837, "relative": 0.922902 }, { "name": "imu_accel_gyro", "l2": 17.220566, "mean_abs": 1.165329, "relative": 0.929109 }, { "name": "depth_confidence", "l2": 10.393909, "mean_abs": 0.194642, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.205649, "mean_abs": 0.10928, "relative": 0.942632 }, { "name": "video_fisheye_cam1", "l2": 5.111087, "mean_abs": 0.11147, "relative": 0.844564 }, { "name": "video_fisheye_cam2", "l2": 5.649, "mean_abs": 0.116781, "relative": 0.901779 }, { "name": "video_fisheye_cam3", "l2": 5.512185, "mean_abs": 0.11297, "relative": 0.922949 }, { "name": "video_stereo_left", "l2": 5.397878, "mean_abs": 0.121195, "relative": 0.877166 }, { "name": "video_stereo_right", "l2": 6.222954, "mean_abs": 0.142515, "relative": 0.877306 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.135106, "mean_abs": 0.086851, "relative": 0.764122 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.23454555869102478" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.8279412388801575" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.999998927116394" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.8666480779647827" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1068, "start_frame": 5340, "end_frame": 5359, "center_frame": 5349, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.120414, "mean_abs": 0.219315, "relative": 0.8537 }, { "name": "hand_right_joints", "l2": 12.265518, "mean_abs": 0.320508, "relative": 0.966851 }, { "name": "body_joints", "l2": 5.920483, "mean_abs": 0.088169, "relative": 0.799675 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.070383, "mean_abs": 0.009156, "relative": 0.166094 }, { "name": "camera_rotation_matrix", "l2": 3.000742, "mean_abs": 0.201987, "relative": 0.921479 }, { "name": "imu_accel_gyro", "l2": 17.218014, "mean_abs": 1.154825, "relative": 0.928971 }, { "name": "depth_confidence", "l2": 10.357319, "mean_abs": 0.190962, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.210099, "mean_abs": 0.10821, "relative": 0.943438 }, { "name": "video_fisheye_cam1", "l2": 5.158467, "mean_abs": 0.110879, "relative": 0.852393 }, { "name": "video_fisheye_cam2", "l2": 5.688041, "mean_abs": 0.115446, "relative": 0.908011 }, { "name": "video_fisheye_cam3", "l2": 5.568174, "mean_abs": 0.1138, "relative": 0.932324 }, { "name": "video_stereo_left", "l2": 5.410225, "mean_abs": 0.118499, "relative": 0.879172 }, { "name": "video_stereo_right", "l2": 6.341005, "mean_abs": 0.142233, "relative": 0.893949 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.13199, "mean_abs": 0.086887, "relative": 0.763007 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.01057, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.5923038721084595" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.7823457717895508" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999725818634033" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.6617807149887085" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle", "correct": 0, "confidence": "" } } }, { "window_index": 1069, "start_frame": 5345, "end_frame": 5364, "center_frame": 5354, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.090309, "mean_abs": 0.22185, "relative": 0.851161 }, { "name": "hand_right_joints", "l2": 12.074867, "mean_abs": 0.318417, "relative": 0.951823 }, { "name": "body_joints", "l2": 5.765368, "mean_abs": 0.090785, "relative": 0.778723 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.101141, "mean_abs": 0.013301, "relative": 0.238679 }, { "name": "camera_rotation_matrix", "l2": 2.999748, "mean_abs": 0.200382, "relative": 0.921174 }, { "name": "imu_accel_gyro", "l2": 17.120115, "mean_abs": 1.120863, "relative": 0.923689 }, { "name": "depth_confidence", "l2": 10.293898, "mean_abs": 0.190195, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.21414, "mean_abs": 0.108373, "relative": 0.944169 }, { "name": "video_fisheye_cam1", "l2": 5.218948, "mean_abs": 0.111717, "relative": 0.862387 }, { "name": "video_fisheye_cam2", "l2": 5.701018, "mean_abs": 0.11536, "relative": 0.910083 }, { "name": "video_fisheye_cam3", "l2": 5.632541, "mean_abs": 0.114556, "relative": 0.943101 }, { "name": "video_stereo_left", "l2": 5.448926, "mean_abs": 0.119579, "relative": 0.885461 }, { "name": "video_stereo_right", "l2": 6.402893, "mean_abs": 0.143364, "relative": 0.902674 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.03981, "mean_abs": 0.079225, "relative": 0.730017 }, { "name": "caption_objects_interaction_text", "l2": 1.868393, "mean_abs": 0.013979, "relative": 0.857921 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.856697142124176" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Secure coffee container", "correct": 0, "confidence": "0.8155156970024109" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9959025979042053" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.5454310178756714" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle", "correct": 0, "confidence": "" } } }, { "window_index": 1070, "start_frame": 5350, "end_frame": 5369, "center_frame": 5359, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.07163, "mean_abs": 0.219232, "relative": 0.849585 }, { "name": "hand_right_joints", "l2": 12.003426, "mean_abs": 0.317345, "relative": 0.946191 }, { "name": "body_joints", "l2": 5.761337, "mean_abs": 0.094493, "relative": 0.778179 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.114865, "mean_abs": 0.015534, "relative": 0.271064 }, { "name": "camera_rotation_matrix", "l2": 3.000458, "mean_abs": 0.201694, "relative": 0.921392 }, { "name": "imu_accel_gyro", "l2": 17.069443, "mean_abs": 1.114361, "relative": 0.920955 }, { "name": "depth_confidence", "l2": 10.290768, "mean_abs": 0.191356, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.218185, "mean_abs": 0.109001, "relative": 0.944902 }, { "name": "video_fisheye_cam1", "l2": 5.27737, "mean_abs": 0.11304, "relative": 0.872041 }, { "name": "video_fisheye_cam2", "l2": 5.727149, "mean_abs": 0.116483, "relative": 0.914254 }, { "name": "video_fisheye_cam3", "l2": 5.653548, "mean_abs": 0.114692, "relative": 0.946618 }, { "name": "video_stereo_left", "l2": 5.516487, "mean_abs": 0.122181, "relative": 0.89644 }, { "name": "video_stereo_right", "l2": 6.489319, "mean_abs": 0.146744, "relative": 0.914858 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.029502, "mean_abs": 0.077764, "relative": 0.726328 }, { "name": "caption_objects_interaction_text", "l2": 1.868393, "mean_abs": 0.01397, "relative": 0.857921 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9675118327140808" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Secure coffee container", "correct": 0, "confidence": "0.8898538947105408" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999918937683105" }, "next_action": { "true": "Pour coffee", "predicted": "Pick up kettle", "correct": 0, "confidence": "0.3888595402240753" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle", "correct": 0, "confidence": "" } } }, { "window_index": 1071, "start_frame": 5355, "end_frame": 5374, "center_frame": 5364, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.050819, "mean_abs": 0.214211, "relative": 0.84783 }, { "name": "hand_right_joints", "l2": 11.943987, "mean_abs": 0.315867, "relative": 0.941506 }, { "name": "body_joints", "l2": 5.764069, "mean_abs": 0.095212, "relative": 0.778548 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.117761, "mean_abs": 0.015654, "relative": 0.277899 }, { "name": "camera_rotation_matrix", "l2": 3.001928, "mean_abs": 0.205883, "relative": 0.921843 }, { "name": "imu_accel_gyro", "l2": 17.063288, "mean_abs": 1.13573, "relative": 0.920623 }, { "name": "depth_confidence", "l2": 10.351767, "mean_abs": 0.192751, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.216286, "mean_abs": 0.10926, "relative": 0.944558 }, { "name": "video_fisheye_cam1", "l2": 5.321694, "mean_abs": 0.114917, "relative": 0.879365 }, { "name": "video_fisheye_cam2", "l2": 5.762923, "mean_abs": 0.118284, "relative": 0.919965 }, { "name": "video_fisheye_cam3", "l2": 5.67351, "mean_abs": 0.11478, "relative": 0.949961 }, { "name": "video_stereo_left", "l2": 5.58993, "mean_abs": 0.125547, "relative": 0.908375 }, { "name": "video_stereo_right", "l2": 6.571385, "mean_abs": 0.150456, "relative": 0.926427 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.006733, "mean_abs": 0.077509, "relative": 0.71818 }, { "name": "caption_objects_interaction_text", "l2": 1.868393, "mean_abs": 0.013625, "relative": 0.857921 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9908761382102966" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Secure coffee container", "correct": 0, "confidence": "0.7258061766624451" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.5905745029449463" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1072, "start_frame": 5360, "end_frame": 5379, "center_frame": 5369, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.042562, "mean_abs": 0.212455, "relative": 0.847133 }, { "name": "hand_right_joints", "l2": 11.772886, "mean_abs": 0.296074, "relative": 0.928018 }, { "name": "body_joints", "l2": 5.505795, "mean_abs": 0.090409, "relative": 0.743663 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.098753, "mean_abs": 0.013235, "relative": 0.233042 }, { "name": "camera_rotation_matrix", "l2": 3.003093, "mean_abs": 0.207839, "relative": 0.922201 }, { "name": "imu_accel_gyro", "l2": 17.054499, "mean_abs": 1.139989, "relative": 0.920149 }, { "name": "depth_confidence", "l2": 10.371037, "mean_abs": 0.192458, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.20399, "mean_abs": 0.109055, "relative": 0.942331 }, { "name": "video_fisheye_cam1", "l2": 5.35919, "mean_abs": 0.115371, "relative": 0.885561 }, { "name": "video_fisheye_cam2", "l2": 5.777028, "mean_abs": 0.1189, "relative": 0.922217 }, { "name": "video_fisheye_cam3", "l2": 5.67141, "mean_abs": 0.114675, "relative": 0.949609 }, { "name": "video_stereo_left", "l2": 5.617477, "mean_abs": 0.126209, "relative": 0.912851 }, { "name": "video_stereo_right", "l2": 6.618209, "mean_abs": 0.151523, "relative": 0.933029 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.977705, "mean_abs": 0.078069, "relative": 0.707791 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008449, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9923466444015503" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.6154042482376099" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.8106060028076172" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1073, "start_frame": 5365, "end_frame": 5384, "center_frame": 5374, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.07521, "mean_abs": 0.21606, "relative": 0.849887 }, { "name": "hand_right_joints", "l2": 11.819631, "mean_abs": 0.27656, "relative": 0.931703 }, { "name": "body_joints", "l2": 5.25976, "mean_abs": 0.082079, "relative": 0.710432 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.07052, "mean_abs": 0.009811, "relative": 0.166417 }, { "name": "camera_rotation_matrix", "l2": 3.003405, "mean_abs": 0.207519, "relative": 0.922297 }, { "name": "imu_accel_gyro", "l2": 17.080698, "mean_abs": 1.145516, "relative": 0.921562 }, { "name": "depth_confidence", "l2": 10.428662, "mean_abs": 0.190839, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.202602, "mean_abs": 0.108841, "relative": 0.94208 }, { "name": "video_fisheye_cam1", "l2": 5.38064, "mean_abs": 0.114761, "relative": 0.889105 }, { "name": "video_fisheye_cam2", "l2": 5.782509, "mean_abs": 0.118437, "relative": 0.923092 }, { "name": "video_fisheye_cam3", "l2": 5.661283, "mean_abs": 0.114593, "relative": 0.947913 }, { "name": "video_stereo_left", "l2": 5.628609, "mean_abs": 0.124073, "relative": 0.91466 }, { "name": "video_stereo_right", "l2": 6.64536, "mean_abs": 0.14938, "relative": 0.936856 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.146646, "mean_abs": 0.08616, "relative": 0.768252 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008449, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.9066081643104553" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.8066345453262329" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999375343322754" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.5633077025413513" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|table|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1074, "start_frame": 5370, "end_frame": 5389, "center_frame": 5379, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.11017, "mean_abs": 0.221818, "relative": 0.852836 }, { "name": "hand_right_joints", "l2": 11.923772, "mean_abs": 0.284153, "relative": 0.939912 }, { "name": "body_joints", "l2": 5.36047, "mean_abs": 0.08713, "relative": 0.724034 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.048818, "mean_abs": 0.007254, "relative": 0.115203 }, { "name": "camera_rotation_matrix", "l2": 3.003284, "mean_abs": 0.208686, "relative": 0.92226 }, { "name": "imu_accel_gyro", "l2": 17.156384, "mean_abs": 1.157334, "relative": 0.925646 }, { "name": "depth_confidence", "l2": 10.495952, "mean_abs": 0.190422, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.199201, "mean_abs": 0.108462, "relative": 0.941464 }, { "name": "video_fisheye_cam1", "l2": 5.373858, "mean_abs": 0.114614, "relative": 0.887985 }, { "name": "video_fisheye_cam2", "l2": 5.766712, "mean_abs": 0.118283, "relative": 0.92057 }, { "name": "video_fisheye_cam3", "l2": 5.64607, "mean_abs": 0.113917, "relative": 0.945366 }, { "name": "video_stereo_left", "l2": 5.59694, "mean_abs": 0.122107, "relative": 0.909514 }, { "name": "video_stereo_right", "l2": 6.625856, "mean_abs": 0.146815, "relative": 0.934107 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.129531, "mean_abs": 0.082277, "relative": 0.762127 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008449, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.8385462760925293" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.44964125752449036" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9718577861785889" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.763824999332428" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|table|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1075, "start_frame": 5375, "end_frame": 5394, "center_frame": 5384, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.140276, "mean_abs": 0.226722, "relative": 0.855376 }, { "name": "hand_right_joints", "l2": 11.731905, "mean_abs": 0.295587, "relative": 0.924788 }, { "name": "body_joints", "l2": 5.536708, "mean_abs": 0.092974, "relative": 0.747839 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.031125, "mean_abs": 0.004556, "relative": 0.07345 }, { "name": "camera_rotation_matrix", "l2": 3.001563, "mean_abs": 0.209733, "relative": 0.921731 }, { "name": "imu_accel_gyro", "l2": 17.135288, "mean_abs": 1.154486, "relative": 0.924508 }, { "name": "depth_confidence", "l2": 10.509678, "mean_abs": 0.193575, "relative": 0.000349 }, { "name": "video_fisheye_cam0", "l2": 5.197432, "mean_abs": 0.108546, "relative": 0.941144 }, { "name": "video_fisheye_cam1", "l2": 5.36704, "mean_abs": 0.114756, "relative": 0.886858 }, { "name": "video_fisheye_cam2", "l2": 5.73688, "mean_abs": 0.117687, "relative": 0.915808 }, { "name": "video_fisheye_cam3", "l2": 5.655378, "mean_abs": 0.114544, "relative": 0.946925 }, { "name": "video_stereo_left", "l2": 5.570132, "mean_abs": 0.123686, "relative": 0.905158 }, { "name": "video_stereo_right", "l2": 6.60769, "mean_abs": 0.149389, "relative": 0.931546 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.14963, "mean_abs": 0.084963, "relative": 0.76932 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008449, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.7352426052093506" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.8149455785751343" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.8340427279472351" }, "next_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.7957364916801453" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|table|dripper|coffee filter|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1076, "start_frame": 5380, "end_frame": 5399, "center_frame": 5389, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.149143, "mean_abs": 0.231337, "relative": 0.856124 }, { "name": "hand_right_joints", "l2": 11.458209, "mean_abs": 0.303911, "relative": 0.903213 }, { "name": "body_joints", "l2": 5.787365, "mean_abs": 0.099621, "relative": 0.781695 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.043853, "mean_abs": 0.006109, "relative": 0.103487 }, { "name": "camera_rotation_matrix", "l2": 3.009706, "mean_abs": 0.213944, "relative": 0.924232 }, { "name": "imu_accel_gyro", "l2": 17.205328, "mean_abs": 1.168281, "relative": 0.928287 }, { "name": "depth_confidence", "l2": 10.570117, "mean_abs": 0.197689, "relative": 0.000351 }, { "name": "video_fisheye_cam0", "l2": 5.199047, "mean_abs": 0.109142, "relative": 0.941436 }, { "name": "video_fisheye_cam1", "l2": 5.385966, "mean_abs": 0.116692, "relative": 0.889985 }, { "name": "video_fisheye_cam2", "l2": 5.694821, "mean_abs": 0.117632, "relative": 0.909093 }, { "name": "video_fisheye_cam3", "l2": 5.692691, "mean_abs": 0.116338, "relative": 0.953172 }, { "name": "video_stereo_left", "l2": 5.562221, "mean_abs": 0.126119, "relative": 0.903872 }, { "name": "video_stereo_right", "l2": 6.609765, "mean_abs": 0.152613, "relative": 0.931838 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.229189, "mean_abs": 0.088156, "relative": 0.797793 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.008449, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Place kettle on table", "correct": 0, "confidence": "0.5416185259819031" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.936048150062561" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.5104843378067017" }, "next_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.5064749121665955" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|coffee mug|carafe", "predicted": "coffee dripper|scale|gooseneck kettle|table|dripper|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1077, "start_frame": 5385, "end_frame": 5404, "center_frame": 5394, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale", "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.113688, "mean_abs": 0.237115, "relative": 0.853133 }, { "name": "hand_right_joints", "l2": 11.146124, "mean_abs": 0.302061, "relative": 0.878613 }, { "name": "body_joints", "l2": 5.943096, "mean_abs": 0.101969, "relative": 0.802729 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.06276, "mean_abs": 0.007738, "relative": 0.148106 }, { "name": "camera_rotation_matrix", "l2": 3.016243, "mean_abs": 0.218369, "relative": 0.926239 }, { "name": "imu_accel_gyro", "l2": 17.18664, "mean_abs": 1.154138, "relative": 0.927278 }, { "name": "depth_confidence", "l2": 10.560711, "mean_abs": 0.200238, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.192657, "mean_abs": 0.109706, "relative": 0.940279 }, { "name": "video_fisheye_cam1", "l2": 5.41332, "mean_abs": 0.118683, "relative": 0.894505 }, { "name": "video_fisheye_cam2", "l2": 5.65133, "mean_abs": 0.118386, "relative": 0.902151 }, { "name": "video_fisheye_cam3", "l2": 5.701791, "mean_abs": 0.117819, "relative": 0.954696 }, { "name": "video_stereo_left", "l2": 5.551968, "mean_abs": 0.127677, "relative": 0.902206 }, { "name": "video_stereo_right", "l2": 6.649163, "mean_abs": 0.155487, "relative": 0.937392 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.092868, "mean_abs": 0.08163, "relative": 0.749006 }, { "name": "caption_objects_interaction_text", "l2": 1.917299, "mean_abs": 0.014544, "relative": 0.880377 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.720714807510376" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Set down kettle and retrieve white bottle", "correct": 0, "confidence": "0.8526344895362854" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.9996504783630371" }, "next_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.83284592628479" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|bottle|digital scale|coffee mug|carafe|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|gooseneck kettle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1078, "start_frame": 5390, "end_frame": 5409, "center_frame": 5399, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale", "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.097687, "mean_abs": 0.239698, "relative": 0.851783 }, { "name": "hand_right_joints", "l2": 10.861827, "mean_abs": 0.295521, "relative": 0.856203 }, { "name": "body_joints", "l2": 6.038055, "mean_abs": 0.101831, "relative": 0.815555 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.068803, "mean_abs": 0.008611, "relative": 0.162365 }, { "name": "camera_rotation_matrix", "l2": 3.020848, "mean_abs": 0.220001, "relative": 0.927653 }, { "name": "imu_accel_gyro", "l2": 17.335476, "mean_abs": 1.17889, "relative": 0.935309 }, { "name": "depth_confidence", "l2": 10.535793, "mean_abs": 0.201951, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.185352, "mean_abs": 0.110148, "relative": 0.938956 }, { "name": "video_fisheye_cam1", "l2": 5.435091, "mean_abs": 0.119939, "relative": 0.898103 }, { "name": "video_fisheye_cam2", "l2": 5.597661, "mean_abs": 0.118065, "relative": 0.893583 }, { "name": "video_fisheye_cam3", "l2": 5.722353, "mean_abs": 0.118893, "relative": 0.958139 }, { "name": "video_stereo_left", "l2": 5.559033, "mean_abs": 0.129078, "relative": 0.903354 }, { "name": "video_stereo_right", "l2": 6.71008, "mean_abs": 0.158492, "relative": 0.94598 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.063474, "mean_abs": 0.08074, "relative": 0.738486 }, { "name": "caption_objects_interaction_text", "l2": 1.917299, "mean_abs": 0.014918, "relative": 0.880377 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.8634116649627686" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.47600027918815613" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.9998980760574341" }, "next_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.9456624984741211" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|bottle|digital scale|coffee mug|carafe|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|gooseneck kettle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1079, "start_frame": 5395, "end_frame": 5414, "center_frame": 5404, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "carafe", "coffee mug", "scale", "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.06761, "mean_abs": 0.238559, "relative": 0.849246 }, { "name": "hand_right_joints", "l2": 10.49042, "mean_abs": 0.285435, "relative": 0.826926 }, { "name": "body_joints", "l2": 6.084439, "mean_abs": 0.097221, "relative": 0.82182 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.065561, "mean_abs": 0.008553, "relative": 0.154716 }, { "name": "camera_rotation_matrix", "l2": 3.014649, "mean_abs": 0.216953, "relative": 0.92575 }, { "name": "imu_accel_gyro", "l2": 17.201351, "mean_abs": 1.138438, "relative": 0.928072 }, { "name": "depth_confidence", "l2": 10.571307, "mean_abs": 0.200782, "relative": 0.000351 }, { "name": "video_fisheye_cam0", "l2": 5.18292, "mean_abs": 0.109772, "relative": 0.938516 }, { "name": "video_fisheye_cam1", "l2": 5.463452, "mean_abs": 0.119571, "relative": 0.902789 }, { "name": "video_fisheye_cam2", "l2": 5.559893, "mean_abs": 0.117138, "relative": 0.887554 }, { "name": "video_fisheye_cam3", "l2": 5.735033, "mean_abs": 0.119029, "relative": 0.960262 }, { "name": "video_stereo_left", "l2": 5.520286, "mean_abs": 0.12665, "relative": 0.897057 }, { "name": "video_stereo_right", "l2": 6.712696, "mean_abs": 0.157661, "relative": 0.946349 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.113033, "mean_abs": 0.083136, "relative": 0.756223 }, { "name": "caption_objects_interaction_text", "l2": 1.917299, "mean_abs": 0.014879, "relative": 0.880377 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.8354018330574036" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.7629497647285461" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.8380669951438904" }, "next_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.6592293381690979" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "scale|bottle|digital scale|coffee mug|carafe|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|gooseneck kettle|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1080, "start_frame": 5400, "end_frame": 5419, "center_frame": 5409, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.011016, "mean_abs": 0.2351, "relative": 0.844472 }, { "name": "hand_right_joints", "l2": 10.107449, "mean_abs": 0.272645, "relative": 0.796737 }, { "name": "body_joints", "l2": 6.159191, "mean_abs": 0.094191, "relative": 0.831917 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.043512, "mean_abs": 0.005939, "relative": 0.102683 }, { "name": "camera_rotation_matrix", "l2": 3.00609, "mean_abs": 0.211783, "relative": 0.923121 }, { "name": "imu_accel_gyro", "l2": 17.05015, "mean_abs": 1.077194, "relative": 0.919914 }, { "name": "depth_confidence", "l2": 10.574834, "mean_abs": 0.198356, "relative": 0.000351 }, { "name": "video_fisheye_cam0", "l2": 5.172476, "mean_abs": 0.108745, "relative": 0.936625 }, { "name": "video_fisheye_cam1", "l2": 5.489948, "mean_abs": 0.118677, "relative": 0.907168 }, { "name": "video_fisheye_cam2", "l2": 5.524887, "mean_abs": 0.115317, "relative": 0.881966 }, { "name": "video_fisheye_cam3", "l2": 5.720694, "mean_abs": 0.117875, "relative": 0.957861 }, { "name": "video_stereo_left", "l2": 5.491143, "mean_abs": 0.123679, "relative": 0.892322 }, { "name": "video_stereo_right", "l2": 6.731061, "mean_abs": 0.15485, "relative": 0.948938 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.127348, "mean_abs": 0.083058, "relative": 0.761346 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pick up white bottle", "correct": 0, "confidence": "0.3466179370880127" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.4458489716053009" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9906365275382996" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.43649834394454956" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1081, "start_frame": 5405, "end_frame": 5424, "center_frame": 5414, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.023505, "mean_abs": 0.234854, "relative": 0.845525 }, { "name": "hand_right_joints", "l2": 9.958531, "mean_abs": 0.267347, "relative": 0.784999 }, { "name": "body_joints", "l2": 6.209379, "mean_abs": 0.09336, "relative": 0.838696 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.023636, "mean_abs": 0.003598, "relative": 0.055778 }, { "name": "camera_rotation_matrix", "l2": 3.000866, "mean_abs": 0.206019, "relative": 0.921517 }, { "name": "imu_accel_gyro", "l2": 17.062679, "mean_abs": 1.048852, "relative": 0.92059 }, { "name": "depth_confidence", "l2": 10.544324, "mean_abs": 0.195341, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.166246, "mean_abs": 0.107873, "relative": 0.935497 }, { "name": "video_fisheye_cam1", "l2": 5.533321, "mean_abs": 0.118073, "relative": 0.914335 }, { "name": "video_fisheye_cam2", "l2": 5.484881, "mean_abs": 0.113339, "relative": 0.87558 }, { "name": "video_fisheye_cam3", "l2": 5.73026, "mean_abs": 0.116905, "relative": 0.959463 }, { "name": "video_stereo_left", "l2": 5.462441, "mean_abs": 0.121113, "relative": 0.887658 }, { "name": "video_stereo_right", "l2": 6.784956, "mean_abs": 0.153189, "relative": 0.956537 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.137351, "mean_abs": 0.083318, "relative": 0.764926 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.7035894393920898" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6513504981994629" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999895095825195" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.5174224376678467" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1082, "start_frame": 5410, "end_frame": 5429, "center_frame": 5419, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.028044, "mean_abs": 0.233916, "relative": 0.845908 }, { "name": "hand_right_joints", "l2": 9.916775, "mean_abs": 0.26454, "relative": 0.781707 }, { "name": "body_joints", "l2": 6.200377, "mean_abs": 0.092597, "relative": 0.83748 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.035251, "mean_abs": 0.005111, "relative": 0.083188 }, { "name": "camera_rotation_matrix", "l2": 3.00112, "mean_abs": 0.206259, "relative": 0.921595 }, { "name": "imu_accel_gyro", "l2": 17.093498, "mean_abs": 1.053645, "relative": 0.922253 }, { "name": "depth_confidence", "l2": 10.558625, "mean_abs": 0.194027, "relative": 0.00035 }, { "name": "video_fisheye_cam0", "l2": 5.169225, "mean_abs": 0.10797, "relative": 0.936036 }, { "name": "video_fisheye_cam1", "l2": 5.536923, "mean_abs": 0.118341, "relative": 0.91493 }, { "name": "video_fisheye_cam2", "l2": 5.460317, "mean_abs": 0.113117, "relative": 0.871658 }, { "name": "video_fisheye_cam3", "l2": 5.762827, "mean_abs": 0.117627, "relative": 0.964916 }, { "name": "video_stereo_left", "l2": 5.47327, "mean_abs": 0.120958, "relative": 0.889417 }, { "name": "video_stereo_right", "l2": 6.808354, "mean_abs": 0.152116, "relative": 0.959835 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.066862, "mean_abs": 0.081473, "relative": 0.739699 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.6737644672393799" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7110114097595215" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999860525131226" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.48010414838790894" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1083, "start_frame": 5415, "end_frame": 5434, "center_frame": 5424, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.065595, "mean_abs": 0.235901, "relative": 0.849076 }, { "name": "hand_right_joints", "l2": 9.882653, "mean_abs": 0.263283, "relative": 0.779017 }, { "name": "body_joints", "l2": 6.174057, "mean_abs": 0.092016, "relative": 0.833925 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.038582, "mean_abs": 0.005673, "relative": 0.091048 }, { "name": "camera_rotation_matrix", "l2": 3.001123, "mean_abs": 0.206245, "relative": 0.921596 }, { "name": "imu_accel_gyro", "l2": 17.056625, "mean_abs": 1.06327, "relative": 0.920264 }, { "name": "depth_confidence", "l2": 10.512783, "mean_abs": 0.192959, "relative": 0.000349 }, { "name": "video_fisheye_cam0", "l2": 5.169568, "mean_abs": 0.107817, "relative": 0.936098 }, { "name": "video_fisheye_cam1", "l2": 5.545298, "mean_abs": 0.118409, "relative": 0.916314 }, { "name": "video_fisheye_cam2", "l2": 5.463845, "mean_abs": 0.113435, "relative": 0.872222 }, { "name": "video_fisheye_cam3", "l2": 5.750221, "mean_abs": 0.118052, "relative": 0.962805 }, { "name": "video_stereo_left", "l2": 5.500876, "mean_abs": 0.121435, "relative": 0.893903 }, { "name": "video_stereo_right", "l2": 6.870559, "mean_abs": 0.153303, "relative": 0.968605 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.022258, "mean_abs": 0.080515, "relative": 0.723736 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.5195813179016113" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8168526887893677" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999940395355225" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.5785593390464783" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1084, "start_frame": 5420, "end_frame": 5439, "center_frame": 5429, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.08634, "mean_abs": 0.236026, "relative": 0.850826 }, { "name": "hand_right_joints", "l2": 9.855098, "mean_abs": 0.261819, "relative": 0.776845 }, { "name": "body_joints", "l2": 6.143055, "mean_abs": 0.091309, "relative": 0.829737 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.0336, "mean_abs": 0.004937, "relative": 0.079291 }, { "name": "camera_rotation_matrix", "l2": 3.000706, "mean_abs": 0.204259, "relative": 0.921468 }, { "name": "imu_accel_gyro", "l2": 17.01675, "mean_abs": 1.056429, "relative": 0.918112 }, { "name": "depth_confidence", "l2": 10.457817, "mean_abs": 0.191178, "relative": 0.000347 }, { "name": "video_fisheye_cam0", "l2": 5.174084, "mean_abs": 0.10765, "relative": 0.936916 }, { "name": "video_fisheye_cam1", "l2": 5.543175, "mean_abs": 0.117858, "relative": 0.915963 }, { "name": "video_fisheye_cam2", "l2": 5.461857, "mean_abs": 0.112886, "relative": 0.871904 }, { "name": "video_fisheye_cam3", "l2": 5.741807, "mean_abs": 0.117427, "relative": 0.961396 }, { "name": "video_stereo_left", "l2": 5.512327, "mean_abs": 0.120517, "relative": 0.895764 }, { "name": "video_stereo_right", "l2": 6.904069, "mean_abs": 0.152775, "relative": 0.973329 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.998772, "mean_abs": 0.079867, "relative": 0.71533 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.44726771116256714" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9165775775909424" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999957084655762" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.5041675567626953" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1085, "start_frame": 5425, "end_frame": 5444, "center_frame": 5434, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.103076, "mean_abs": 0.235522, "relative": 0.852238 }, { "name": "hand_right_joints", "l2": 9.808657, "mean_abs": 0.25932, "relative": 0.773185 }, { "name": "body_joints", "l2": 6.094603, "mean_abs": 0.089952, "relative": 0.823193 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081674, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.023531, "mean_abs": 0.003392, "relative": 0.05553 }, { "name": "camera_rotation_matrix", "l2": 3.000937, "mean_abs": 0.20224, "relative": 0.921539 }, { "name": "imu_accel_gyro", "l2": 17.035929, "mean_abs": 1.061313, "relative": 0.919147 }, { "name": "depth_confidence", "l2": 10.331931, "mean_abs": 0.188722, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.174675, "mean_abs": 0.107494, "relative": 0.937023 }, { "name": "video_fisheye_cam1", "l2": 5.514084, "mean_abs": 0.116922, "relative": 0.911156 }, { "name": "video_fisheye_cam2", "l2": 5.454975, "mean_abs": 0.112481, "relative": 0.870806 }, { "name": "video_fisheye_cam3", "l2": 5.690909, "mean_abs": 0.11669, "relative": 0.952874 }, { "name": "video_stereo_left", "l2": 5.513633, "mean_abs": 0.120043, "relative": 0.895976 }, { "name": "video_stereo_right", "l2": 6.908164, "mean_abs": 0.152159, "relative": 0.973906 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.961638, "mean_abs": 0.076984, "relative": 0.702041 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5167304277420044" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9569470286369324" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999984502792358" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.5877898335456848" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1086, "start_frame": 5430, "end_frame": 5449, "center_frame": 5439, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.125139, "mean_abs": 0.236382, "relative": 0.854099 }, { "name": "hand_right_joints", "l2": 9.778682, "mean_abs": 0.257877, "relative": 0.770822 }, { "name": "body_joints", "l2": 6.060678, "mean_abs": 0.089029, "relative": 0.818611 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012391, "mean_abs": 0.001787, "relative": 0.029241 }, { "name": "camera_rotation_matrix", "l2": 3.000387, "mean_abs": 0.201078, "relative": 0.92137 }, { "name": "imu_accel_gyro", "l2": 17.080744, "mean_abs": 1.049477, "relative": 0.921565 }, { "name": "depth_confidence", "l2": 10.254706, "mean_abs": 0.186803, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.173963, "mean_abs": 0.10743, "relative": 0.936894 }, { "name": "video_fisheye_cam1", "l2": 5.504434, "mean_abs": 0.116679, "relative": 0.909561 }, { "name": "video_fisheye_cam2", "l2": 5.449727, "mean_abs": 0.112068, "relative": 0.869968 }, { "name": "video_fisheye_cam3", "l2": 5.666616, "mean_abs": 0.116322, "relative": 0.948807 }, { "name": "video_stereo_left", "l2": 5.515061, "mean_abs": 0.119104, "relative": 0.896208 }, { "name": "video_stereo_right", "l2": 6.927413, "mean_abs": 0.151266, "relative": 0.97662 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.986651, "mean_abs": 0.078588, "relative": 0.710993 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6913955807685852" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9645357728004456" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999994039535522" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.5813976526260376" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1087, "start_frame": 5435, "end_frame": 5454, "center_frame": 5444, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.120358, "mean_abs": 0.235014, "relative": 0.853695 }, { "name": "hand_right_joints", "l2": 9.749713, "mean_abs": 0.256022, "relative": 0.768538 }, { "name": "body_joints", "l2": 6.032498, "mean_abs": 0.088098, "relative": 0.814805 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.010097, "mean_abs": 0.001686, "relative": 0.023828 }, { "name": "camera_rotation_matrix", "l2": 3.00019, "mean_abs": 0.200371, "relative": 0.92131 }, { "name": "imu_accel_gyro", "l2": 17.062262, "mean_abs": 1.050233, "relative": 0.920568 }, { "name": "depth_confidence", "l2": 10.204195, "mean_abs": 0.185379, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.178875, "mean_abs": 0.107487, "relative": 0.937784 }, { "name": "video_fisheye_cam1", "l2": 5.515833, "mean_abs": 0.11684, "relative": 0.911445 }, { "name": "video_fisheye_cam2", "l2": 5.447985, "mean_abs": 0.11207, "relative": 0.86969 }, { "name": "video_fisheye_cam3", "l2": 5.644601, "mean_abs": 0.115476, "relative": 0.94512 }, { "name": "video_stereo_left", "l2": 5.515886, "mean_abs": 0.119124, "relative": 0.896342 }, { "name": "video_stereo_right", "l2": 6.963086, "mean_abs": 0.151656, "relative": 0.981649 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.045738, "mean_abs": 0.081367, "relative": 0.732139 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.4861065745353699" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.932016909122467" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999997615814209" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.768185555934906" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1088, "start_frame": 5440, "end_frame": 5459, "center_frame": 5449, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.129007, "mean_abs": 0.236494, "relative": 0.854425 }, { "name": "hand_right_joints", "l2": 9.71816, "mean_abs": 0.253972, "relative": 0.766051 }, { "name": "body_joints", "l2": 6.020183, "mean_abs": 0.087767, "relative": 0.813141 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018036, "mean_abs": 0.002738, "relative": 0.042563 }, { "name": "camera_rotation_matrix", "l2": 3.000335, "mean_abs": 0.19909, "relative": 0.921354 }, { "name": "imu_accel_gyro", "l2": 17.049471, "mean_abs": 1.051336, "relative": 0.919878 }, { "name": "depth_confidence", "l2": 10.134283, "mean_abs": 0.183645, "relative": 0.000336 }, { "name": "video_fisheye_cam0", "l2": 5.180589, "mean_abs": 0.107598, "relative": 0.938094 }, { "name": "video_fisheye_cam1", "l2": 5.495933, "mean_abs": 0.116042, "relative": 0.908156 }, { "name": "video_fisheye_cam2", "l2": 5.4419, "mean_abs": 0.111797, "relative": 0.868718 }, { "name": "video_fisheye_cam3", "l2": 5.636512, "mean_abs": 0.116105, "relative": 0.943766 }, { "name": "video_stereo_left", "l2": 5.511073, "mean_abs": 0.119101, "relative": 0.89556 }, { "name": "video_stereo_right", "l2": 6.949024, "mean_abs": 0.150753, "relative": 0.979667 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.071088, "mean_abs": 0.084183, "relative": 0.741211 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.450313925743103" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8207755088806152" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999974966049194" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.7288050651550293" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1089, "start_frame": 5445, "end_frame": 5464, "center_frame": 5454, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.131048, "mean_abs": 0.237141, "relative": 0.854597 }, { "name": "hand_right_joints", "l2": 9.687666, "mean_abs": 0.252344, "relative": 0.763647 }, { "name": "body_joints", "l2": 6.009419, "mean_abs": 0.087405, "relative": 0.811687 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.021407, "mean_abs": 0.003077, "relative": 0.050518 }, { "name": "camera_rotation_matrix", "l2": 3.000118, "mean_abs": 0.199244, "relative": 0.921288 }, { "name": "imu_accel_gyro", "l2": 17.053921, "mean_abs": 1.053508, "relative": 0.920118 }, { "name": "depth_confidence", "l2": 10.091958, "mean_abs": 0.182681, "relative": 0.000335 }, { "name": "video_fisheye_cam0", "l2": 5.182887, "mean_abs": 0.107597, "relative": 0.93851 }, { "name": "video_fisheye_cam1", "l2": 5.500167, "mean_abs": 0.115926, "relative": 0.908856 }, { "name": "video_fisheye_cam2", "l2": 5.438797, "mean_abs": 0.111754, "relative": 0.868223 }, { "name": "video_fisheye_cam3", "l2": 5.621587, "mean_abs": 0.11516, "relative": 0.941267 }, { "name": "video_stereo_left", "l2": 5.50653, "mean_abs": 0.119175, "relative": 0.894822 }, { "name": "video_stereo_right", "l2": 6.966345, "mean_abs": 0.150491, "relative": 0.982109 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.1279, "mean_abs": 0.087681, "relative": 0.761543 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.41516485810279846" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6617971062660217" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999905824661255" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.7479664087295532" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1090, "start_frame": 5450, "end_frame": 5469, "center_frame": 5459, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.136112, "mean_abs": 0.238049, "relative": 0.855024 }, { "name": "hand_right_joints", "l2": 9.673428, "mean_abs": 0.251531, "relative": 0.762525 }, { "name": "body_joints", "l2": 6.013058, "mean_abs": 0.087297, "relative": 0.812179 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.024696, "mean_abs": 0.003438, "relative": 0.058278 }, { "name": "camera_rotation_matrix", "l2": 3.000315, "mean_abs": 0.200557, "relative": 0.921348 }, { "name": "imu_accel_gyro", "l2": 17.056517, "mean_abs": 1.05839, "relative": 0.920258 }, { "name": "depth_confidence", "l2": 10.067096, "mean_abs": 0.182151, "relative": 0.000334 }, { "name": "video_fisheye_cam0", "l2": 5.186041, "mean_abs": 0.107689, "relative": 0.939081 }, { "name": "video_fisheye_cam1", "l2": 5.510945, "mean_abs": 0.116246, "relative": 0.910637 }, { "name": "video_fisheye_cam2", "l2": 5.440764, "mean_abs": 0.111998, "relative": 0.868537 }, { "name": "video_fisheye_cam3", "l2": 5.616777, "mean_abs": 0.115193, "relative": 0.940462 }, { "name": "video_stereo_left", "l2": 5.515853, "mean_abs": 0.119578, "relative": 0.896337 }, { "name": "video_stereo_right", "l2": 6.990288, "mean_abs": 0.150931, "relative": 0.985484 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.093651, "mean_abs": 0.085367, "relative": 0.749286 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.46660831570625305" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.569770097732544" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999302625656128" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.7985426783561707" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1091, "start_frame": 5455, "end_frame": 5474, "center_frame": 5464, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.146531, "mean_abs": 0.238584, "relative": 0.855903 }, { "name": "hand_right_joints", "l2": 9.649812, "mean_abs": 0.251029, "relative": 0.760663 }, { "name": "body_joints", "l2": 6.018884, "mean_abs": 0.087032, "relative": 0.812966 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.02276, "mean_abs": 0.003269, "relative": 0.053711 }, { "name": "camera_rotation_matrix", "l2": 3.000341, "mean_abs": 0.200958, "relative": 0.921356 }, { "name": "imu_accel_gyro", "l2": 17.060019, "mean_abs": 1.057184, "relative": 0.920447 }, { "name": "depth_confidence", "l2": 10.069364, "mean_abs": 0.18199, "relative": 0.000334 }, { "name": "video_fisheye_cam0", "l2": 5.185818, "mean_abs": 0.107672, "relative": 0.939041 }, { "name": "video_fisheye_cam1", "l2": 5.513774, "mean_abs": 0.116279, "relative": 0.911105 }, { "name": "video_fisheye_cam2", "l2": 5.440204, "mean_abs": 0.112059, "relative": 0.868448 }, { "name": "video_fisheye_cam3", "l2": 5.619481, "mean_abs": 0.115158, "relative": 0.940914 }, { "name": "video_stereo_left", "l2": 5.519121, "mean_abs": 0.119826, "relative": 0.896868 }, { "name": "video_stereo_right", "l2": 7.004198, "mean_abs": 0.151197, "relative": 0.987445 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.041912, "mean_abs": 0.082255, "relative": 0.73077 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.39988890290260315" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6647174954414368" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9998735189437866" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.5693606734275818" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1092, "start_frame": 5460, "end_frame": 5479, "center_frame": 5469, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.140116, "mean_abs": 0.238531, "relative": 0.855362 }, { "name": "hand_right_joints", "l2": 9.62729, "mean_abs": 0.250708, "relative": 0.758888 }, { "name": "body_joints", "l2": 6.033328, "mean_abs": 0.087158, "relative": 0.814917 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017925, "mean_abs": 0.002638, "relative": 0.042302 }, { "name": "camera_rotation_matrix", "l2": 3.000226, "mean_abs": 0.200989, "relative": 0.921321 }, { "name": "imu_accel_gyro", "l2": 17.064133, "mean_abs": 1.065221, "relative": 0.920669 }, { "name": "depth_confidence", "l2": 10.043468, "mean_abs": 0.181365, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.185744, "mean_abs": 0.107702, "relative": 0.939027 }, { "name": "video_fisheye_cam1", "l2": 5.50666, "mean_abs": 0.116175, "relative": 0.909929 }, { "name": "video_fisheye_cam2", "l2": 5.438797, "mean_abs": 0.111965, "relative": 0.868223 }, { "name": "video_fisheye_cam3", "l2": 5.60464, "mean_abs": 0.114907, "relative": 0.938429 }, { "name": "video_stereo_left", "l2": 5.521704, "mean_abs": 0.120067, "relative": 0.897288 }, { "name": "video_stereo_right", "l2": 7.002533, "mean_abs": 0.151525, "relative": 0.98721 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.04319, "mean_abs": 0.082906, "relative": 0.731227 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.3505536913871765" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6807705163955688" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9990271329879761" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.6997290253639221" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1093, "start_frame": 5465, "end_frame": 5484, "center_frame": 5474, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.143643, "mean_abs": 0.239159, "relative": 0.85566 }, { "name": "hand_right_joints", "l2": 9.612627, "mean_abs": 0.250381, "relative": 0.757732 }, { "name": "body_joints", "l2": 6.049651, "mean_abs": 0.087151, "relative": 0.817121 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013066, "mean_abs": 0.001968, "relative": 0.030833 }, { "name": "camera_rotation_matrix", "l2": 3.000195, "mean_abs": 0.200918, "relative": 0.921311 }, { "name": "imu_accel_gyro", "l2": 17.016687, "mean_abs": 1.073818, "relative": 0.918109 }, { "name": "depth_confidence", "l2": 10.041012, "mean_abs": 0.180928, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.186399, "mean_abs": 0.107554, "relative": 0.939146 }, { "name": "video_fisheye_cam1", "l2": 5.498811, "mean_abs": 0.115689, "relative": 0.908632 }, { "name": "video_fisheye_cam2", "l2": 5.439767, "mean_abs": 0.111902, "relative": 0.868378 }, { "name": "video_fisheye_cam3", "l2": 5.609448, "mean_abs": 0.114486, "relative": 0.939234 }, { "name": "video_stereo_left", "l2": 5.526384, "mean_abs": 0.119977, "relative": 0.898048 }, { "name": "video_stereo_right", "l2": 6.995972, "mean_abs": 0.150689, "relative": 0.986285 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.086505, "mean_abs": 0.08548, "relative": 0.746729 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.38868919014930725" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.817051887512207" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9918013215065002" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.621292233467102" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1094, "start_frame": 5470, "end_frame": 5489, "center_frame": 5479, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.12974, "mean_abs": 0.239256, "relative": 0.854487 }, { "name": "hand_right_joints", "l2": 9.606566, "mean_abs": 0.250094, "relative": 0.757254 }, { "name": "body_joints", "l2": 6.059177, "mean_abs": 0.087313, "relative": 0.818408 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008097, "mean_abs": 0.001234, "relative": 0.019107 }, { "name": "camera_rotation_matrix", "l2": 3.000129, "mean_abs": 0.200367, "relative": 0.921291 }, { "name": "imu_accel_gyro", "l2": 17.01535, "mean_abs": 1.062998, "relative": 0.918037 }, { "name": "depth_confidence", "l2": 10.047951, "mean_abs": 0.18092, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.204319, "mean_abs": 0.108214, "relative": 0.942391 }, { "name": "video_fisheye_cam1", "l2": 5.500263, "mean_abs": 0.115418, "relative": 0.908872 }, { "name": "video_fisheye_cam2", "l2": 5.442204, "mean_abs": 0.111622, "relative": 0.868767 }, { "name": "video_fisheye_cam3", "l2": 5.615272, "mean_abs": 0.114404, "relative": 0.94021 }, { "name": "video_stereo_left", "l2": 5.533761, "mean_abs": 0.119514, "relative": 0.899247 }, { "name": "video_stereo_right", "l2": 7.006799, "mean_abs": 0.150466, "relative": 0.987812 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.169721, "mean_abs": 0.089687, "relative": 0.776511 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5390603542327881" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8234524130821228" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9880963563919067" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.606971800327301" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1095, "start_frame": 5475, "end_frame": 5494, "center_frame": 5484, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.129168, "mean_abs": 0.238823, "relative": 0.854438 }, { "name": "hand_right_joints", "l2": 9.598346, "mean_abs": 0.249969, "relative": 0.756606 }, { "name": "body_joints", "l2": 6.06494, "mean_abs": 0.087294, "relative": 0.819187 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005134, "mean_abs": 0.000819, "relative": 0.012115 }, { "name": "camera_rotation_matrix", "l2": 2.999874, "mean_abs": 0.200222, "relative": 0.921213 }, { "name": "imu_accel_gyro", "l2": 17.023827, "mean_abs": 1.065908, "relative": 0.918494 }, { "name": "depth_confidence", "l2": 10.025384, "mean_abs": 0.180377, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.210998, "mean_abs": 0.108302, "relative": 0.9436 }, { "name": "video_fisheye_cam1", "l2": 5.493988, "mean_abs": 0.115281, "relative": 0.907835 }, { "name": "video_fisheye_cam2", "l2": 5.445017, "mean_abs": 0.111553, "relative": 0.869216 }, { "name": "video_fisheye_cam3", "l2": 5.611854, "mean_abs": 0.114495, "relative": 0.939637 }, { "name": "video_stereo_left", "l2": 5.540612, "mean_abs": 0.119463, "relative": 0.90036 }, { "name": "video_stereo_right", "l2": 7.001244, "mean_abs": 0.150111, "relative": 0.987029 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.137401, "mean_abs": 0.086198, "relative": 0.764944 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5887999534606934" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6819961667060852" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9986903071403503" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.6300854682922363" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1096, "start_frame": 5480, "end_frame": 5499, "center_frame": 5489, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.122605, "mean_abs": 0.23896, "relative": 0.853885 }, { "name": "hand_right_joints", "l2": 9.591391, "mean_abs": 0.249618, "relative": 0.756058 }, { "name": "body_joints", "l2": 6.075928, "mean_abs": 0.087335, "relative": 0.820671 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003538, "mean_abs": 0.000564, "relative": 0.00835 }, { "name": "camera_rotation_matrix", "l2": 2.999707, "mean_abs": 0.19995, "relative": 0.921161 }, { "name": "imu_accel_gyro", "l2": 17.026436, "mean_abs": 1.063592, "relative": 0.918635 }, { "name": "depth_confidence", "l2": 10.024909, "mean_abs": 0.180671, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.224195, "mean_abs": 0.108435, "relative": 0.94599 }, { "name": "video_fisheye_cam1", "l2": 5.483755, "mean_abs": 0.115196, "relative": 0.906144 }, { "name": "video_fisheye_cam2", "l2": 5.451316, "mean_abs": 0.111554, "relative": 0.870221 }, { "name": "video_fisheye_cam3", "l2": 5.596085, "mean_abs": 0.115068, "relative": 0.936997 }, { "name": "video_stereo_left", "l2": 5.552402, "mean_abs": 0.119455, "relative": 0.902276 }, { "name": "video_stereo_right", "l2": 6.993433, "mean_abs": 0.150236, "relative": 0.985927 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.15521, "mean_abs": 0.08678, "relative": 0.771317 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6301208734512329" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.718791127204895" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9978051781654358" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.4781215488910675" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1097, "start_frame": 5485, "end_frame": 5504, "center_frame": 5494, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.118291, "mean_abs": 0.23844, "relative": 0.853521 }, { "name": "hand_right_joints", "l2": 9.590428, "mean_abs": 0.250029, "relative": 0.755982 }, { "name": "body_joints", "l2": 6.077159, "mean_abs": 0.087355, "relative": 0.820837 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003422, "mean_abs": 0.000581, "relative": 0.008077 }, { "name": "camera_rotation_matrix", "l2": 2.99987, "mean_abs": 0.199472, "relative": 0.921211 }, { "name": "imu_accel_gyro", "l2": 17.048326, "mean_abs": 1.064882, "relative": 0.919816 }, { "name": "depth_confidence", "l2": 10.023652, "mean_abs": 0.180541, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.23072, "mean_abs": 0.108553, "relative": 0.947172 }, { "name": "video_fisheye_cam1", "l2": 5.471146, "mean_abs": 0.11427, "relative": 0.904061 }, { "name": "video_fisheye_cam2", "l2": 5.45806, "mean_abs": 0.11162, "relative": 0.871298 }, { "name": "video_fisheye_cam3", "l2": 5.57158, "mean_abs": 0.113588, "relative": 0.932894 }, { "name": "video_stereo_left", "l2": 5.564104, "mean_abs": 0.119502, "relative": 0.904178 }, { "name": "video_stereo_right", "l2": 6.992963, "mean_abs": 0.149477, "relative": 0.985861 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.039804, "mean_abs": 0.081415, "relative": 0.730015 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5938012003898621" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7612749338150024" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9983514547348022" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.553778350353241" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1098, "start_frame": 5490, "end_frame": 5509, "center_frame": 5499, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.111295, "mean_abs": 0.239169, "relative": 0.852931 }, { "name": "hand_right_joints", "l2": 9.585089, "mean_abs": 0.249455, "relative": 0.755561 }, { "name": "body_joints", "l2": 6.078843, "mean_abs": 0.087566, "relative": 0.821064 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.002701, "mean_abs": 0.000461, "relative": 0.006375 }, { "name": "camera_rotation_matrix", "l2": 3.000228, "mean_abs": 0.19924, "relative": 0.921321 }, { "name": "imu_accel_gyro", "l2": 17.020037, "mean_abs": 1.056882, "relative": 0.91829 }, { "name": "depth_confidence", "l2": 10.017679, "mean_abs": 0.179808, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.242856, "mean_abs": 0.108531, "relative": 0.949369 }, { "name": "video_fisheye_cam1", "l2": 5.474015, "mean_abs": 0.114411, "relative": 0.904535 }, { "name": "video_fisheye_cam2", "l2": 5.461108, "mean_abs": 0.111748, "relative": 0.871785 }, { "name": "video_fisheye_cam3", "l2": 5.569013, "mean_abs": 0.113559, "relative": 0.932464 }, { "name": "video_stereo_left", "l2": 5.568752, "mean_abs": 0.119603, "relative": 0.904933 }, { "name": "video_stereo_right", "l2": 7.002156, "mean_abs": 0.14976, "relative": 0.987157 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.061609, "mean_abs": 0.084048, "relative": 0.737819 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.4004683494567871" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7344683408737183" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9983839988708496" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.5578931570053101" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1099, "start_frame": 5495, "end_frame": 5514, "center_frame": 5504, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.111578, "mean_abs": 0.239455, "relative": 0.852955 }, { "name": "hand_right_joints", "l2": 9.583762, "mean_abs": 0.250455, "relative": 0.755457 }, { "name": "body_joints", "l2": 6.089525, "mean_abs": 0.088008, "relative": 0.822507 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003575, "mean_abs": 0.000548, "relative": 0.008436 }, { "name": "camera_rotation_matrix", "l2": 3.000019, "mean_abs": 0.19887, "relative": 0.921257 }, { "name": "imu_accel_gyro", "l2": 17.021803, "mean_abs": 1.056155, "relative": 0.918385 }, { "name": "depth_confidence", "l2": 10.020696, "mean_abs": 0.179818, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.246243, "mean_abs": 0.108588, "relative": 0.949982 }, { "name": "video_fisheye_cam1", "l2": 5.477751, "mean_abs": 0.114403, "relative": 0.905152 }, { "name": "video_fisheye_cam2", "l2": 5.465174, "mean_abs": 0.111722, "relative": 0.872434 }, { "name": "video_fisheye_cam3", "l2": 5.567742, "mean_abs": 0.113612, "relative": 0.932251 }, { "name": "video_stereo_left", "l2": 5.57606, "mean_abs": 0.119497, "relative": 0.906121 }, { "name": "video_stereo_right", "l2": 7.009566, "mean_abs": 0.149641, "relative": 0.988202 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.070672, "mean_abs": 0.084008, "relative": 0.741063 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.39144906401634216" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.785860538482666" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9992814660072327" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.47992461919784546" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1100, "start_frame": 5500, "end_frame": 5519, "center_frame": 5509, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.112583, "mean_abs": 0.241367, "relative": 0.85304 }, { "name": "hand_right_joints", "l2": 9.589215, "mean_abs": 0.252044, "relative": 0.755887 }, { "name": "body_joints", "l2": 6.111058, "mean_abs": 0.089139, "relative": 0.825416 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004309, "mean_abs": 0.000592, "relative": 0.010169 }, { "name": "camera_rotation_matrix", "l2": 3.000038, "mean_abs": 0.199029, "relative": 0.921263 }, { "name": "imu_accel_gyro", "l2": 17.018915, "mean_abs": 1.059741, "relative": 0.918229 }, { "name": "depth_confidence", "l2": 10.01861, "mean_abs": 0.179684, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.249496, "mean_abs": 0.108491, "relative": 0.950572 }, { "name": "video_fisheye_cam1", "l2": 5.486127, "mean_abs": 0.114496, "relative": 0.906536 }, { "name": "video_fisheye_cam2", "l2": 5.467702, "mean_abs": 0.111769, "relative": 0.872837 }, { "name": "video_fisheye_cam3", "l2": 5.578444, "mean_abs": 0.113276, "relative": 0.934043 }, { "name": "video_stereo_left", "l2": 5.581684, "mean_abs": 0.119565, "relative": 0.907035 }, { "name": "video_stereo_right", "l2": 7.022694, "mean_abs": 0.149499, "relative": 0.990053 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.026117, "mean_abs": 0.083914, "relative": 0.725117 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010458, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.41047894954681396" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8864577412605286" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9978103041648865" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.39191678166389465" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1101, "start_frame": 5505, "end_frame": 5524, "center_frame": 5514, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.113896, "mean_abs": 0.242479, "relative": 0.85315 }, { "name": "hand_right_joints", "l2": 9.59023, "mean_abs": 0.253559, "relative": 0.755967 }, { "name": "body_joints", "l2": 6.137932, "mean_abs": 0.08979, "relative": 0.829045 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003854, "mean_abs": 0.000623, "relative": 0.009096 }, { "name": "camera_rotation_matrix", "l2": 2.999963, "mean_abs": 0.198885, "relative": 0.92124 }, { "name": "imu_accel_gyro", "l2": 17.023579, "mean_abs": 1.059332, "relative": 0.918481 }, { "name": "depth_confidence", "l2": 10.013029, "mean_abs": 0.179086, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.261058, "mean_abs": 0.108565, "relative": 0.952665 }, { "name": "video_fisheye_cam1", "l2": 5.507977, "mean_abs": 0.114642, "relative": 0.910147 }, { "name": "video_fisheye_cam2", "l2": 5.47388, "mean_abs": 0.111804, "relative": 0.873824 }, { "name": "video_fisheye_cam3", "l2": 5.618037, "mean_abs": 0.113682, "relative": 0.940673 }, { "name": "video_stereo_left", "l2": 5.592829, "mean_abs": 0.119706, "relative": 0.908846 }, { "name": "video_stereo_right", "l2": 7.057199, "mean_abs": 0.149875, "relative": 0.994917 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.076367, "mean_abs": 0.086674, "relative": 0.743101 }, { "name": "caption_objects_interaction_text", "l2": 1.842907, "mean_abs": 0.013885, "relative": 0.846218 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6049293279647827" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9398118257522583" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.9999887943267822" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.5049875974655151" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1102, "start_frame": 5510, "end_frame": 5529, "center_frame": 5519, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.10859, "mean_abs": 0.243746, "relative": 0.852703 }, { "name": "hand_right_joints", "l2": 9.593459, "mean_abs": 0.254683, "relative": 0.756221 }, { "name": "body_joints", "l2": 6.165212, "mean_abs": 0.090884, "relative": 0.83273 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006811, "mean_abs": 0.000981, "relative": 0.016072 }, { "name": "camera_rotation_matrix", "l2": 2.999962, "mean_abs": 0.198705, "relative": 0.92124 }, { "name": "imu_accel_gyro", "l2": 17.036003, "mean_abs": 1.058634, "relative": 0.919151 }, { "name": "depth_confidence", "l2": 10.016273, "mean_abs": 0.180129, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.252061, "mean_abs": 0.108498, "relative": 0.951036 }, { "name": "video_fisheye_cam1", "l2": 5.498538, "mean_abs": 0.114526, "relative": 0.908587 }, { "name": "video_fisheye_cam2", "l2": 5.471735, "mean_abs": 0.111784, "relative": 0.873481 }, { "name": "video_fisheye_cam3", "l2": 5.587826, "mean_abs": 0.11345, "relative": 0.935614 }, { "name": "video_stereo_left", "l2": 5.589712, "mean_abs": 0.119699, "relative": 0.908339 }, { "name": "video_stereo_right", "l2": 7.052666, "mean_abs": 0.149952, "relative": 0.994278 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.055539, "mean_abs": 0.084713, "relative": 0.735647 }, { "name": "caption_objects_interaction_text", "l2": 1.842907, "mean_abs": 0.013928, "relative": 0.846218 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6593788266181946" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9590763449668884" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.9999179840087891" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.4263857305049896" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1103, "start_frame": 5515, "end_frame": 5534, "center_frame": 5524, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.111799, "mean_abs": 0.24567, "relative": 0.852973 }, { "name": "hand_right_joints", "l2": 9.600753, "mean_abs": 0.255636, "relative": 0.756796 }, { "name": "body_joints", "l2": 6.200362, "mean_abs": 0.091531, "relative": 0.837478 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008884, "mean_abs": 0.00129, "relative": 0.020965 }, { "name": "camera_rotation_matrix", "l2": 2.999963, "mean_abs": 0.19882, "relative": 0.92124 }, { "name": "imu_accel_gyro", "l2": 17.039665, "mean_abs": 1.056973, "relative": 0.919349 }, { "name": "depth_confidence", "l2": 10.016855, "mean_abs": 0.179733, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.246352, "mean_abs": 0.108554, "relative": 0.950002 }, { "name": "video_fisheye_cam1", "l2": 5.494454, "mean_abs": 0.11457, "relative": 0.907912 }, { "name": "video_fisheye_cam2", "l2": 5.470335, "mean_abs": 0.111825, "relative": 0.873258 }, { "name": "video_fisheye_cam3", "l2": 5.571735, "mean_abs": 0.113579, "relative": 0.93292 }, { "name": "video_stereo_left", "l2": 5.588647, "mean_abs": 0.119839, "relative": 0.908166 }, { "name": "video_stereo_right", "l2": 7.052155, "mean_abs": 0.150195, "relative": 0.994206 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.067264, "mean_abs": 0.085267, "relative": 0.739843 }, { "name": "caption_objects_interaction_text", "l2": 1.842907, "mean_abs": 0.013858, "relative": 0.846218 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7078307271003723" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9605588912963867" }, "transition_detection": { "true": "steady", "predicted": "transition", "correct": 0, "confidence": "0.9997813105583191" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle away", "correct": 0, "confidence": "0.40649980306625366" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1104, "start_frame": 5520, "end_frame": 5539, "center_frame": 5529, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.108273, "mean_abs": 0.24638, "relative": 0.852676 }, { "name": "hand_right_joints", "l2": 9.592548, "mean_abs": 0.255985, "relative": 0.756149 }, { "name": "body_joints", "l2": 6.224066, "mean_abs": 0.091697, "relative": 0.84068 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.011166, "mean_abs": 0.001505, "relative": 0.02635 }, { "name": "camera_rotation_matrix", "l2": 2.999877, "mean_abs": 0.198805, "relative": 0.921214 }, { "name": "imu_accel_gyro", "l2": 17.050165, "mean_abs": 1.061692, "relative": 0.919915 }, { "name": "depth_confidence", "l2": 10.01254, "mean_abs": 0.179799, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.243348, "mean_abs": 0.108505, "relative": 0.949458 }, { "name": "video_fisheye_cam1", "l2": 5.496747, "mean_abs": 0.114642, "relative": 0.908291 }, { "name": "video_fisheye_cam2", "l2": 5.47068, "mean_abs": 0.111872, "relative": 0.873313 }, { "name": "video_fisheye_cam3", "l2": 5.56577, "mean_abs": 0.113556, "relative": 0.931921 }, { "name": "video_stereo_left", "l2": 5.590473, "mean_abs": 0.119964, "relative": 0.908463 }, { "name": "video_stereo_right", "l2": 7.058217, "mean_abs": 0.150361, "relative": 0.99506 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.022776, "mean_abs": 0.082696, "relative": 0.723921 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.4912412166595459" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9479674100875854" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.986160397529602" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle away", "correct": 0, "confidence": "0.48188483715057373" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1105, "start_frame": 5525, "end_frame": 5544, "center_frame": 5534, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.102787, "mean_abs": 0.245909, "relative": 0.852213 }, { "name": "hand_right_joints", "l2": 9.604616, "mean_abs": 0.258505, "relative": 0.757101 }, { "name": "body_joints", "l2": 6.265959, "mean_abs": 0.092361, "relative": 0.846338 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013839, "mean_abs": 0.001837, "relative": 0.032658 }, { "name": "camera_rotation_matrix", "l2": 2.999809, "mean_abs": 0.199062, "relative": 0.921193 }, { "name": "imu_accel_gyro", "l2": 17.045303, "mean_abs": 1.06391, "relative": 0.919653 }, { "name": "depth_confidence", "l2": 10.013896, "mean_abs": 0.18016, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.239851, "mean_abs": 0.108386, "relative": 0.948825 }, { "name": "video_fisheye_cam1", "l2": 5.496793, "mean_abs": 0.114828, "relative": 0.908299 }, { "name": "video_fisheye_cam2", "l2": 5.468751, "mean_abs": 0.111919, "relative": 0.873005 }, { "name": "video_fisheye_cam3", "l2": 5.558561, "mean_abs": 0.113392, "relative": 0.930714 }, { "name": "video_stereo_left", "l2": 5.591524, "mean_abs": 0.120114, "relative": 0.908634 }, { "name": "video_stereo_right", "l2": 7.053867, "mean_abs": 0.150747, "relative": 0.994447 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.001994, "mean_abs": 0.079278, "relative": 0.716484 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.4996519088745117" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9378976225852966" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9965194463729858" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle away", "correct": 0, "confidence": "0.4925818145275116" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1106, "start_frame": 5530, "end_frame": 5549, "center_frame": 5539, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.107602, "mean_abs": 0.246469, "relative": 0.852619 }, { "name": "hand_right_joints", "l2": 9.643986, "mean_abs": 0.263294, "relative": 0.760204 }, { "name": "body_joints", "l2": 6.333585, "mean_abs": 0.094011, "relative": 0.855472 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012747, "mean_abs": 0.001778, "relative": 0.030081 }, { "name": "camera_rotation_matrix", "l2": 3.000216, "mean_abs": 0.200067, "relative": 0.921318 }, { "name": "imu_accel_gyro", "l2": 17.039913, "mean_abs": 1.065842, "relative": 0.919362 }, { "name": "depth_confidence", "l2": 10.010085, "mean_abs": 0.181051, "relative": 0.000332 }, { "name": "video_fisheye_cam0", "l2": 5.238814, "mean_abs": 0.108524, "relative": 0.948637 }, { "name": "video_fisheye_cam1", "l2": 5.488897, "mean_abs": 0.115896, "relative": 0.906994 }, { "name": "video_fisheye_cam2", "l2": 5.463728, "mean_abs": 0.112422, "relative": 0.872203 }, { "name": "video_fisheye_cam3", "l2": 5.541646, "mean_abs": 0.113257, "relative": 0.927882 }, { "name": "video_stereo_left", "l2": 5.591866, "mean_abs": 0.121445, "relative": 0.908689 }, { "name": "video_stereo_right", "l2": 7.014042, "mean_abs": 0.152727, "relative": 0.988833 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.02309, "mean_abs": 0.080747, "relative": 0.724034 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.6068305373191833" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.892570972442627" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9687994718551636" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle away", "correct": 0, "confidence": "0.448794960975647" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1107, "start_frame": 5535, "end_frame": 5554, "center_frame": 5544, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.085181, "mean_abs": 0.245945, "relative": 0.850728 }, { "name": "hand_right_joints", "l2": 9.650794, "mean_abs": 0.265662, "relative": 0.760741 }, { "name": "body_joints", "l2": 6.369666, "mean_abs": 0.095488, "relative": 0.860346 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.029958, "mean_abs": 0.004161, "relative": 0.070697 }, { "name": "camera_rotation_matrix", "l2": 3.00146, "mean_abs": 0.202623, "relative": 0.9217 }, { "name": "imu_accel_gyro", "l2": 17.083143, "mean_abs": 1.075873, "relative": 0.921694 }, { "name": "depth_confidence", "l2": 10.035302, "mean_abs": 0.18339, "relative": 0.000333 }, { "name": "video_fisheye_cam0", "l2": 5.24628, "mean_abs": 0.109265, "relative": 0.949989 }, { "name": "video_fisheye_cam1", "l2": 5.489853, "mean_abs": 0.117303, "relative": 0.907152 }, { "name": "video_fisheye_cam2", "l2": 5.467663, "mean_abs": 0.113742, "relative": 0.872831 }, { "name": "video_fisheye_cam3", "l2": 5.520529, "mean_abs": 0.113884, "relative": 0.924346 }, { "name": "video_stereo_left", "l2": 5.621458, "mean_abs": 0.124039, "relative": 0.913498 }, { "name": "video_stereo_right", "l2": 6.994122, "mean_abs": 0.155948, "relative": 0.986024 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.219479, "mean_abs": 0.0917, "relative": 0.794318 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.5957943797111511" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8206225633621216" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9970190525054932" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.4504070580005646" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1108, "start_frame": 5540, "end_frame": 5559, "center_frame": 5549, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.090313, "mean_abs": 0.245437, "relative": 0.851161 }, { "name": "hand_right_joints", "l2": 9.64048, "mean_abs": 0.266109, "relative": 0.759928 }, { "name": "body_joints", "l2": 6.395395, "mean_abs": 0.096253, "relative": 0.863821 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.051793, "mean_abs": 0.007579, "relative": 0.122225 }, { "name": "camera_rotation_matrix", "l2": 3.003853, "mean_abs": 0.205537, "relative": 0.922434 }, { "name": "imu_accel_gyro", "l2": 17.072723, "mean_abs": 1.087038, "relative": 0.921132 }, { "name": "depth_confidence", "l2": 10.084154, "mean_abs": 0.18563, "relative": 0.000335 }, { "name": "video_fisheye_cam0", "l2": 5.255097, "mean_abs": 0.110063, "relative": 0.951586 }, { "name": "video_fisheye_cam1", "l2": 5.495264, "mean_abs": 0.118421, "relative": 0.908046 }, { "name": "video_fisheye_cam2", "l2": 5.48276, "mean_abs": 0.115224, "relative": 0.875241 }, { "name": "video_fisheye_cam3", "l2": 5.50367, "mean_abs": 0.114433, "relative": 0.921523 }, { "name": "video_stereo_left", "l2": 5.637194, "mean_abs": 0.126078, "relative": 0.916055 }, { "name": "video_stereo_right", "l2": 6.96931, "mean_abs": 0.157336, "relative": 0.982526 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.149414, "mean_abs": 0.088654, "relative": 0.769243 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6390894055366516" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.85584956407547" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999922513961792" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.5173118710517883" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1109, "start_frame": 5545, "end_frame": 5564, "center_frame": 5554, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.041581, "mean_abs": 0.2439, "relative": 0.84705 }, { "name": "hand_right_joints", "l2": 9.653179, "mean_abs": 0.265875, "relative": 0.760929 }, { "name": "body_joints", "l2": 6.418985, "mean_abs": 0.096169, "relative": 0.867007 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.064254, "mean_abs": 0.009636, "relative": 0.151631 }, { "name": "camera_rotation_matrix", "l2": 3.005622, "mean_abs": 0.206977, "relative": 0.922978 }, { "name": "imu_accel_gyro", "l2": 17.063828, "mean_abs": 1.086685, "relative": 0.920652 }, { "name": "depth_confidence", "l2": 10.140713, "mean_abs": 0.188381, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.25801, "mean_abs": 0.11052, "relative": 0.952113 }, { "name": "video_fisheye_cam1", "l2": 5.477252, "mean_abs": 0.118297, "relative": 0.90507 }, { "name": "video_fisheye_cam2", "l2": 5.495077, "mean_abs": 0.115992, "relative": 0.877207 }, { "name": "video_fisheye_cam3", "l2": 5.473711, "mean_abs": 0.114088, "relative": 0.916507 }, { "name": "video_stereo_left", "l2": 5.636329, "mean_abs": 0.127179, "relative": 0.915915 }, { "name": "video_stereo_right", "l2": 6.896499, "mean_abs": 0.15648, "relative": 0.972262 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.150254, "mean_abs": 0.089779, "relative": 0.769544 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6411758661270142" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6890159845352173" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999998807907104" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.7085299491882324" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1110, "start_frame": 5550, "end_frame": 5569, "center_frame": 5559, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.036097, "mean_abs": 0.241849, "relative": 0.846588 }, { "name": "hand_right_joints", "l2": 9.667607, "mean_abs": 0.264715, "relative": 0.762066 }, { "name": "body_joints", "l2": 6.421783, "mean_abs": 0.094812, "relative": 0.867385 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.056997, "mean_abs": 0.008624, "relative": 0.134505 }, { "name": "camera_rotation_matrix", "l2": 3.004336, "mean_abs": 0.205335, "relative": 0.922583 }, { "name": "imu_accel_gyro", "l2": 17.07918, "mean_abs": 1.083125, "relative": 0.921481 }, { "name": "depth_confidence", "l2": 10.186917, "mean_abs": 0.188936, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.259779, "mean_abs": 0.110335, "relative": 0.952433 }, { "name": "video_fisheye_cam1", "l2": 5.445335, "mean_abs": 0.11638, "relative": 0.899796 }, { "name": "video_fisheye_cam2", "l2": 5.497616, "mean_abs": 0.115623, "relative": 0.877613 }, { "name": "video_fisheye_cam3", "l2": 5.431407, "mean_abs": 0.112277, "relative": 0.909424 }, { "name": "video_stereo_left", "l2": 5.617797, "mean_abs": 0.12583, "relative": 0.912903 }, { "name": "video_stereo_right", "l2": 6.765387, "mean_abs": 0.152326, "relative": 0.953778 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.131666, "mean_abs": 0.089434, "relative": 0.762891 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8746318221092224" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7970624566078186" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "1.0" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.41855207085609436" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1111, "start_frame": 5555, "end_frame": 5574, "center_frame": 5564, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.036263, "mean_abs": 0.24141, "relative": 0.846602 }, { "name": "hand_right_joints", "l2": 9.667553, "mean_abs": 0.263245, "relative": 0.762062 }, { "name": "body_joints", "l2": 6.405913, "mean_abs": 0.093203, "relative": 0.865241 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081672, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.035703, "mean_abs": 0.005341, "relative": 0.084255 }, { "name": "camera_rotation_matrix", "l2": 3.00263, "mean_abs": 0.201456, "relative": 0.922059 }, { "name": "imu_accel_gyro", "l2": 17.065811, "mean_abs": 1.048729, "relative": 0.920759 }, { "name": "depth_confidence", "l2": 10.268428, "mean_abs": 0.189772, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.260271, "mean_abs": 0.109729, "relative": 0.952523 }, { "name": "video_fisheye_cam1", "l2": 5.430198, "mean_abs": 0.114603, "relative": 0.897294 }, { "name": "video_fisheye_cam2", "l2": 5.509249, "mean_abs": 0.114584, "relative": 0.87947 }, { "name": "video_fisheye_cam3", "l2": 5.43752, "mean_abs": 0.111663, "relative": 0.910447 }, { "name": "video_stereo_left", "l2": 5.60382, "mean_abs": 0.124207, "relative": 0.910632 }, { "name": "video_stereo_right", "l2": 6.747622, "mean_abs": 0.150481, "relative": 0.951273 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.076748, "mean_abs": 0.086621, "relative": 0.743237 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6987560987472534" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.5991231203079224" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999995231628418" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.3792625069618225" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1112, "start_frame": 5560, "end_frame": 5579, "center_frame": 5569, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.041191, "mean_abs": 0.242518, "relative": 0.847017 }, { "name": "hand_right_joints", "l2": 9.678164, "mean_abs": 0.265094, "relative": 0.762898 }, { "name": "body_joints", "l2": 6.406469, "mean_abs": 0.093737, "relative": 0.865317 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.024944, "mean_abs": 0.003891, "relative": 0.058864 }, { "name": "camera_rotation_matrix", "l2": 3.001336, "mean_abs": 0.198788, "relative": 0.921662 }, { "name": "imu_accel_gyro", "l2": 17.131136, "mean_abs": 1.027014, "relative": 0.924284 }, { "name": "depth_confidence", "l2": 10.33075, "mean_abs": 0.189977, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.258105, "mean_abs": 0.10919, "relative": 0.95213 }, { "name": "video_fisheye_cam1", "l2": 5.42812, "mean_abs": 0.113891, "relative": 0.896951 }, { "name": "video_fisheye_cam2", "l2": 5.520674, "mean_abs": 0.113857, "relative": 0.881294 }, { "name": "video_fisheye_cam3", "l2": 5.435189, "mean_abs": 0.111268, "relative": 0.910057 }, { "name": "video_stereo_left", "l2": 5.582658, "mean_abs": 0.122722, "relative": 0.907193 }, { "name": "video_stereo_right", "l2": 6.743847, "mean_abs": 0.149477, "relative": 0.950741 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.054134, "mean_abs": 0.085732, "relative": 0.735144 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.46799468994140625" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.5253011584281921" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999964237213135" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.4549344480037689" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1113, "start_frame": 5565, "end_frame": 5584, "center_frame": 5574, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.082886, "mean_abs": 0.246482, "relative": 0.850534 }, { "name": "hand_right_joints", "l2": 9.678915, "mean_abs": 0.265568, "relative": 0.762957 }, { "name": "body_joints", "l2": 6.417953, "mean_abs": 0.095048, "relative": 0.866868 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.027823, "mean_abs": 0.003279, "relative": 0.065658 }, { "name": "camera_rotation_matrix", "l2": 3.000566, "mean_abs": 0.198962, "relative": 0.921425 }, { "name": "imu_accel_gyro", "l2": 17.096422, "mean_abs": 1.015995, "relative": 0.922411 }, { "name": "depth_confidence", "l2": 10.348877, "mean_abs": 0.190412, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.25922, "mean_abs": 0.109003, "relative": 0.952332 }, { "name": "video_fisheye_cam1", "l2": 5.437927, "mean_abs": 0.114169, "relative": 0.898572 }, { "name": "video_fisheye_cam2", "l2": 5.534887, "mean_abs": 0.114006, "relative": 0.883562 }, { "name": "video_fisheye_cam3", "l2": 5.462915, "mean_abs": 0.111878, "relative": 0.914699 }, { "name": "video_stereo_left", "l2": 5.571985, "mean_abs": 0.121821, "relative": 0.905459 }, { "name": "video_stereo_right", "l2": 6.78148, "mean_abs": 0.149988, "relative": 0.956046 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.081894, "mean_abs": 0.084509, "relative": 0.745079 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.6102291345596313" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.5423208475112915" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999897480010986" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.44246113300323486" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1114, "start_frame": 5570, "end_frame": 5589, "center_frame": 5579, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.083976, "mean_abs": 0.246887, "relative": 0.850626 }, { "name": "hand_right_joints", "l2": 9.679103, "mean_abs": 0.266836, "relative": 0.762972 }, { "name": "body_joints", "l2": 6.415413, "mean_abs": 0.095694, "relative": 0.866525 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.034909, "mean_abs": 0.00441, "relative": 0.08238 }, { "name": "camera_rotation_matrix", "l2": 3.000816, "mean_abs": 0.199935, "relative": 0.921502 }, { "name": "imu_accel_gyro", "l2": 17.095026, "mean_abs": 1.010934, "relative": 0.922336 }, { "name": "depth_confidence", "l2": 10.359422, "mean_abs": 0.189528, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.263835, "mean_abs": 0.109282, "relative": 0.953168 }, { "name": "video_fisheye_cam1", "l2": 5.466561, "mean_abs": 0.115603, "relative": 0.903303 }, { "name": "video_fisheye_cam2", "l2": 5.552249, "mean_abs": 0.114968, "relative": 0.886334 }, { "name": "video_fisheye_cam3", "l2": 5.532433, "mean_abs": 0.113559, "relative": 0.926339 }, { "name": "video_stereo_left", "l2": 5.613861, "mean_abs": 0.123279, "relative": 0.912264 }, { "name": "video_stereo_right", "l2": 6.892237, "mean_abs": 0.153084, "relative": 0.971661 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.176756, "mean_abs": 0.087922, "relative": 0.779028 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.6595046520233154" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.6106578707695007" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999958276748657" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.43493154644966125" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1115, "start_frame": 5575, "end_frame": 5594, "center_frame": 5584, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.092298, "mean_abs": 0.247667, "relative": 0.851328 }, { "name": "hand_right_joints", "l2": 9.666866, "mean_abs": 0.266948, "relative": 0.762008 }, { "name": "body_joints", "l2": 6.410131, "mean_abs": 0.095819, "relative": 0.865811 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.043761, "mean_abs": 0.006107, "relative": 0.103269 }, { "name": "camera_rotation_matrix", "l2": 3.00258, "mean_abs": 0.2044, "relative": 0.922044 }, { "name": "imu_accel_gyro", "l2": 17.137878, "mean_abs": 1.045668, "relative": 0.924648 }, { "name": "depth_confidence", "l2": 10.369846, "mean_abs": 0.190929, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.261155, "mean_abs": 0.109984, "relative": 0.952683 }, { "name": "video_fisheye_cam1", "l2": 5.475274, "mean_abs": 0.116858, "relative": 0.904743 }, { "name": "video_fisheye_cam2", "l2": 5.561148, "mean_abs": 0.116176, "relative": 0.887754 }, { "name": "video_fisheye_cam3", "l2": 5.599716, "mean_abs": 0.115451, "relative": 0.937605 }, { "name": "video_stereo_left", "l2": 5.659313, "mean_abs": 0.125705, "relative": 0.91965 }, { "name": "video_stereo_right", "l2": 6.947928, "mean_abs": 0.155325, "relative": 0.979512 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.153969, "mean_abs": 0.085481, "relative": 0.770873 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.6951003670692444" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.6381993889808655" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9994183778762817" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.42541012167930603" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1116, "start_frame": 5580, "end_frame": 5599, "center_frame": 5589, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.072794, "mean_abs": 0.246658, "relative": 0.849683 }, { "name": "hand_right_joints", "l2": 9.684363, "mean_abs": 0.266768, "relative": 0.763387 }, { "name": "body_joints", "l2": 6.398221, "mean_abs": 0.094776, "relative": 0.864202 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.043224, "mean_abs": 0.00638, "relative": 0.102004 }, { "name": "camera_rotation_matrix", "l2": 3.002799, "mean_abs": 0.205392, "relative": 0.922111 }, { "name": "imu_accel_gyro", "l2": 17.094763, "mean_abs": 1.044645, "relative": 0.922321 }, { "name": "depth_confidence", "l2": 10.365761, "mean_abs": 0.190362, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.25843, "mean_abs": 0.110089, "relative": 0.952189 }, { "name": "video_fisheye_cam1", "l2": 5.483582, "mean_abs": 0.117135, "relative": 0.906116 }, { "name": "video_fisheye_cam2", "l2": 5.561074, "mean_abs": 0.116383, "relative": 0.887743 }, { "name": "video_fisheye_cam3", "l2": 5.627817, "mean_abs": 0.115419, "relative": 0.94231 }, { "name": "video_stereo_left", "l2": 5.677073, "mean_abs": 0.126235, "relative": 0.922536 }, { "name": "video_stereo_right", "l2": 6.989275, "mean_abs": 0.155719, "relative": 0.985341 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.14822, "mean_abs": 0.085836, "relative": 0.768816 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.7345848679542542" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.7322412133216858" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9948686361312866" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5009293556213379" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1117, "start_frame": 5585, "end_frame": 5604, "center_frame": 5594, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.089364, "mean_abs": 0.246389, "relative": 0.851081 }, { "name": "hand_right_joints", "l2": 9.678212, "mean_abs": 0.264245, "relative": 0.762902 }, { "name": "body_joints", "l2": 6.379162, "mean_abs": 0.093543, "relative": 0.861628 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028734, "mean_abs": 0.00435, "relative": 0.067808 }, { "name": "camera_rotation_matrix", "l2": 3.001095, "mean_abs": 0.203121, "relative": 0.921588 }, { "name": "imu_accel_gyro", "l2": 17.090862, "mean_abs": 1.044879, "relative": 0.922111 }, { "name": "depth_confidence", "l2": 10.350608, "mean_abs": 0.188737, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.259205, "mean_abs": 0.109582, "relative": 0.95233 }, { "name": "video_fisheye_cam1", "l2": 5.499857, "mean_abs": 0.11633, "relative": 0.908805 }, { "name": "video_fisheye_cam2", "l2": 5.553028, "mean_abs": 0.115387, "relative": 0.886458 }, { "name": "video_fisheye_cam3", "l2": 5.674598, "mean_abs": 0.115181, "relative": 0.950143 }, { "name": "video_stereo_left", "l2": 5.680532, "mean_abs": 0.124543, "relative": 0.923098 }, { "name": "video_stereo_right", "l2": 7.042661, "mean_abs": 0.15393, "relative": 0.992867 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.082132, "mean_abs": 0.083747, "relative": 0.745164 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.6186107993125916" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.8192020654678345" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9991812109947205" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.3942295014858246" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1118, "start_frame": 5590, "end_frame": 5609, "center_frame": 5599, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.088116, "mean_abs": 0.246783, "relative": 0.850976 }, { "name": "hand_right_joints", "l2": 9.688703, "mean_abs": 0.262457, "relative": 0.763729 }, { "name": "body_joints", "l2": 6.362207, "mean_abs": 0.092898, "relative": 0.859338 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014367, "mean_abs": 0.002279, "relative": 0.033904 }, { "name": "camera_rotation_matrix", "l2": 2.999735, "mean_abs": 0.200291, "relative": 0.92117 }, { "name": "imu_accel_gyro", "l2": 17.041182, "mean_abs": 1.038658, "relative": 0.91943 }, { "name": "depth_confidence", "l2": 10.345289, "mean_abs": 0.187252, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.257629, "mean_abs": 0.108975, "relative": 0.952044 }, { "name": "video_fisheye_cam1", "l2": 5.506531, "mean_abs": 0.11544, "relative": 0.909908 }, { "name": "video_fisheye_cam2", "l2": 5.540373, "mean_abs": 0.11405, "relative": 0.884438 }, { "name": "video_fisheye_cam3", "l2": 5.690974, "mean_abs": 0.11476, "relative": 0.952885 }, { "name": "video_stereo_left", "l2": 5.679788, "mean_abs": 0.123022, "relative": 0.922977 }, { "name": "video_stereo_right", "l2": 7.056367, "mean_abs": 0.152596, "relative": 0.9948 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.096077, "mean_abs": 0.084155, "relative": 0.750154 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6330573558807373" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pick up and position kettle", "correct": 0, "confidence": "0.5031293034553528" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9800348281860352" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.49381154775619507" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1119, "start_frame": 5595, "end_frame": 5614, "center_frame": 5604, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.08709, "mean_abs": 0.246791, "relative": 0.850889 }, { "name": "hand_right_joints", "l2": 9.687319, "mean_abs": 0.262886, "relative": 0.76362 }, { "name": "body_joints", "l2": 6.357581, "mean_abs": 0.092829, "relative": 0.858713 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.023083, "mean_abs": 0.003106, "relative": 0.054472 }, { "name": "camera_rotation_matrix", "l2": 2.999983, "mean_abs": 0.200756, "relative": 0.921246 }, { "name": "imu_accel_gyro", "l2": 17.044775, "mean_abs": 1.028385, "relative": 0.919624 }, { "name": "depth_confidence", "l2": 10.349348, "mean_abs": 0.187403, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.256239, "mean_abs": 0.108893, "relative": 0.951792 }, { "name": "video_fisheye_cam1", "l2": 5.504726, "mean_abs": 0.115666, "relative": 0.90961 }, { "name": "video_fisheye_cam2", "l2": 5.534459, "mean_abs": 0.114019, "relative": 0.883494 }, { "name": "video_fisheye_cam3", "l2": 5.697642, "mean_abs": 0.11473, "relative": 0.954001 }, { "name": "video_stereo_left", "l2": 5.675355, "mean_abs": 0.122805, "relative": 0.922256 }, { "name": "video_stereo_right", "l2": 7.041795, "mean_abs": 0.151893, "relative": 0.992745 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.122098, "mean_abs": 0.085755, "relative": 0.759467 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7370864748954773" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.6979039311408997" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9985359907150269" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.45497262477874756" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1120, "start_frame": 5600, "end_frame": 5619, "center_frame": 5609, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.098175, "mean_abs": 0.246327, "relative": 0.851824 }, { "name": "hand_right_joints", "l2": 9.669856, "mean_abs": 0.262642, "relative": 0.762243 }, { "name": "body_joints", "l2": 6.344663, "mean_abs": 0.092964, "relative": 0.856968 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.018398, "mean_abs": 0.002719, "relative": 0.043418 }, { "name": "camera_rotation_matrix", "l2": 2.999931, "mean_abs": 0.200042, "relative": 0.92123 }, { "name": "imu_accel_gyro", "l2": 17.037006, "mean_abs": 1.026826, "relative": 0.919205 }, { "name": "depth_confidence", "l2": 10.318822, "mean_abs": 0.186762, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.256478, "mean_abs": 0.108844, "relative": 0.951836 }, { "name": "video_fisheye_cam1", "l2": 5.504366, "mean_abs": 0.115183, "relative": 0.90955 }, { "name": "video_fisheye_cam2", "l2": 5.527776, "mean_abs": 0.113693, "relative": 0.882427 }, { "name": "video_fisheye_cam3", "l2": 5.705717, "mean_abs": 0.1148, "relative": 0.955353 }, { "name": "video_stereo_left", "l2": 5.666404, "mean_abs": 0.122497, "relative": 0.920802 }, { "name": "video_stereo_right", "l2": 7.038042, "mean_abs": 0.151618, "relative": 0.992216 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.162854, "mean_abs": 0.088741, "relative": 0.774053 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7160112857818604" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7233752012252808" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9016844034194946" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.45549720525741577" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1121, "start_frame": 5605, "end_frame": 5624, "center_frame": 5614, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.080184, "mean_abs": 0.244029, "relative": 0.850307 }, { "name": "hand_right_joints", "l2": 9.659556, "mean_abs": 0.262021, "relative": 0.761431 }, { "name": "body_joints", "l2": 6.333073, "mean_abs": 0.093084, "relative": 0.855403 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017285, "mean_abs": 0.002644, "relative": 0.040791 }, { "name": "camera_rotation_matrix", "l2": 3.000487, "mean_abs": 0.201377, "relative": 0.921401 }, { "name": "imu_accel_gyro", "l2": 17.034962, "mean_abs": 1.03785, "relative": 0.919095 }, { "name": "depth_confidence", "l2": 10.313492, "mean_abs": 0.187407, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.255621, "mean_abs": 0.109005, "relative": 0.951681 }, { "name": "video_fisheye_cam1", "l2": 5.501058, "mean_abs": 0.115642, "relative": 0.909003 }, { "name": "video_fisheye_cam2", "l2": 5.519539, "mean_abs": 0.113944, "relative": 0.881112 }, { "name": "video_fisheye_cam3", "l2": 5.691091, "mean_abs": 0.115279, "relative": 0.952904 }, { "name": "video_stereo_left", "l2": 5.654908, "mean_abs": 0.122456, "relative": 0.918934 }, { "name": "video_stereo_right", "l2": 7.030536, "mean_abs": 0.151497, "relative": 0.991158 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.20273, "mean_abs": 0.089851, "relative": 0.788324 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6557028889656067" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7613067030906677" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.5957896709442139" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.47684919834136963" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1122, "start_frame": 5610, "end_frame": 5629, "center_frame": 5619, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.074183, "mean_abs": 0.243577, "relative": 0.8498 }, { "name": "hand_right_joints", "l2": 9.652191, "mean_abs": 0.261355, "relative": 0.760851 }, { "name": "body_joints", "l2": 6.315122, "mean_abs": 0.092933, "relative": 0.852978 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.019397, "mean_abs": 0.002533, "relative": 0.045775 }, { "name": "camera_rotation_matrix", "l2": 3.000576, "mean_abs": 0.20185, "relative": 0.921428 }, { "name": "imu_accel_gyro", "l2": 17.053913, "mean_abs": 1.034418, "relative": 0.920117 }, { "name": "depth_confidence", "l2": 10.298356, "mean_abs": 0.186675, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.255842, "mean_abs": 0.109116, "relative": 0.951721 }, { "name": "video_fisheye_cam1", "l2": 5.500782, "mean_abs": 0.116171, "relative": 0.908958 }, { "name": "video_fisheye_cam2", "l2": 5.521195, "mean_abs": 0.114121, "relative": 0.881377 }, { "name": "video_fisheye_cam3", "l2": 5.693786, "mean_abs": 0.115463, "relative": 0.953356 }, { "name": "video_stereo_left", "l2": 5.665559, "mean_abs": 0.122778, "relative": 0.920665 }, { "name": "video_stereo_right", "l2": 7.033163, "mean_abs": 0.15183, "relative": 0.991528 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.162098, "mean_abs": 0.087481, "relative": 0.773782 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7430909276008606" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7628434300422668" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.567429780960083" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6841712594032288" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1123, "start_frame": 5615, "end_frame": 5634, "center_frame": 5624, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.074167, "mean_abs": 0.243976, "relative": 0.849799 }, { "name": "hand_right_joints", "l2": 9.647745, "mean_abs": 0.26002, "relative": 0.7605 }, { "name": "body_joints", "l2": 6.296272, "mean_abs": 0.092951, "relative": 0.850432 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.024296, "mean_abs": 0.003101, "relative": 0.057335 }, { "name": "camera_rotation_matrix", "l2": 3.000955, "mean_abs": 0.202619, "relative": 0.921545 }, { "name": "imu_accel_gyro", "l2": 17.054588, "mean_abs": 1.030804, "relative": 0.920154 }, { "name": "depth_confidence", "l2": 10.24849, "mean_abs": 0.186132, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.255137, "mean_abs": 0.109125, "relative": 0.951593 }, { "name": "video_fisheye_cam1", "l2": 5.503868, "mean_abs": 0.116719, "relative": 0.909468 }, { "name": "video_fisheye_cam2", "l2": 5.525248, "mean_abs": 0.114416, "relative": 0.882024 }, { "name": "video_fisheye_cam3", "l2": 5.700381, "mean_abs": 0.115994, "relative": 0.95446 }, { "name": "video_stereo_left", "l2": 5.676533, "mean_abs": 0.123143, "relative": 0.922448 }, { "name": "video_stereo_right", "l2": 7.045177, "mean_abs": 0.152731, "relative": 0.993222 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.132058, "mean_abs": 0.086837, "relative": 0.763032 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6620863080024719" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7532759308815002" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.8276044130325317" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6873948574066162" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1124, "start_frame": 5620, "end_frame": 5639, "center_frame": 5629, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.073007, "mean_abs": 0.244624, "relative": 0.849701 }, { "name": "hand_right_joints", "l2": 9.635659, "mean_abs": 0.258786, "relative": 0.759548 }, { "name": "body_joints", "l2": 6.285637, "mean_abs": 0.09305, "relative": 0.848996 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028066, "mean_abs": 0.003604, "relative": 0.066231 }, { "name": "camera_rotation_matrix", "l2": 3.000961, "mean_abs": 0.20292, "relative": 0.921546 }, { "name": "imu_accel_gyro", "l2": 17.042467, "mean_abs": 1.043174, "relative": 0.9195 }, { "name": "depth_confidence", "l2": 10.235185, "mean_abs": 0.185821, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.253918, "mean_abs": 0.109207, "relative": 0.951372 }, { "name": "video_fisheye_cam1", "l2": 5.505261, "mean_abs": 0.116888, "relative": 0.909698 }, { "name": "video_fisheye_cam2", "l2": 5.527734, "mean_abs": 0.114499, "relative": 0.88242 }, { "name": "video_fisheye_cam3", "l2": 5.677322, "mean_abs": 0.115961, "relative": 0.950599 }, { "name": "video_stereo_left", "l2": 5.690005, "mean_abs": 0.12366, "relative": 0.924637 }, { "name": "video_stereo_right", "l2": 7.05971, "mean_abs": 0.153559, "relative": 0.995271 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.078506, "mean_abs": 0.084176, "relative": 0.743866 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6634331345558167" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.812817394733429" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9568713903427124" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5800459980964661" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1125, "start_frame": 5625, "end_frame": 5644, "center_frame": 5634, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.076389, "mean_abs": 0.244281, "relative": 0.849986 }, { "name": "hand_right_joints", "l2": 9.620611, "mean_abs": 0.258194, "relative": 0.758362 }, { "name": "body_joints", "l2": 6.275856, "mean_abs": 0.092952, "relative": 0.847675 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028543, "mean_abs": 0.003829, "relative": 0.067357 }, { "name": "camera_rotation_matrix", "l2": 3.000618, "mean_abs": 0.202871, "relative": 0.921441 }, { "name": "imu_accel_gyro", "l2": 17.057402, "mean_abs": 1.049082, "relative": 0.920306 }, { "name": "depth_confidence", "l2": 10.223891, "mean_abs": 0.185471, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.25413, "mean_abs": 0.109277, "relative": 0.951411 }, { "name": "video_fisheye_cam1", "l2": 5.506237, "mean_abs": 0.116768, "relative": 0.909859 }, { "name": "video_fisheye_cam2", "l2": 5.533779, "mean_abs": 0.114581, "relative": 0.883385 }, { "name": "video_fisheye_cam3", "l2": 5.684092, "mean_abs": 0.116088, "relative": 0.951733 }, { "name": "video_stereo_left", "l2": 5.70201, "mean_abs": 0.123785, "relative": 0.926588 }, { "name": "video_stereo_right", "l2": 7.069167, "mean_abs": 0.153608, "relative": 0.996604 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.046573, "mean_abs": 0.081783, "relative": 0.732438 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5002846717834473" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.786747932434082" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9945935606956482" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.41599586606025696" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1126, "start_frame": 5630, "end_frame": 5649, "center_frame": 5639, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.069017, "mean_abs": 0.244482, "relative": 0.849365 }, { "name": "hand_right_joints", "l2": 9.609586, "mean_abs": 0.257286, "relative": 0.757492 }, { "name": "body_joints", "l2": 6.271724, "mean_abs": 0.092798, "relative": 0.847117 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.0227, "mean_abs": 0.003151, "relative": 0.053568 }, { "name": "camera_rotation_matrix", "l2": 3.000274, "mean_abs": 0.202117, "relative": 0.921336 }, { "name": "imu_accel_gyro", "l2": 17.034954, "mean_abs": 1.049908, "relative": 0.919094 }, { "name": "depth_confidence", "l2": 10.210525, "mean_abs": 0.185006, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.257982, "mean_abs": 0.109209, "relative": 0.952108 }, { "name": "video_fisheye_cam1", "l2": 5.515373, "mean_abs": 0.116373, "relative": 0.911369 }, { "name": "video_fisheye_cam2", "l2": 5.540795, "mean_abs": 0.114337, "relative": 0.884505 }, { "name": "video_fisheye_cam3", "l2": 5.683741, "mean_abs": 0.116014, "relative": 0.951674 }, { "name": "video_stereo_left", "l2": 5.712617, "mean_abs": 0.123332, "relative": 0.928312 }, { "name": "video_stereo_right", "l2": 7.092172, "mean_abs": 0.153112, "relative": 0.999847 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.034874, "mean_abs": 0.080628, "relative": 0.728251 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6709438562393188" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9070062637329102" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.979550302028656" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.3954063355922699" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1127, "start_frame": 5635, "end_frame": 5654, "center_frame": 5644, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.069978, "mean_abs": 0.244339, "relative": 0.849446 }, { "name": "hand_right_joints", "l2": 9.572803, "mean_abs": 0.256492, "relative": 0.754593 }, { "name": "body_joints", "l2": 6.268434, "mean_abs": 0.092551, "relative": 0.846672 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.013616, "mean_abs": 0.001823, "relative": 0.032132 }, { "name": "camera_rotation_matrix", "l2": 2.999972, "mean_abs": 0.200787, "relative": 0.921243 }, { "name": "imu_accel_gyro", "l2": 17.040863, "mean_abs": 1.049719, "relative": 0.919413 }, { "name": "depth_confidence", "l2": 10.208815, "mean_abs": 0.184643, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.262264, "mean_abs": 0.108988, "relative": 0.952884 }, { "name": "video_fisheye_cam1", "l2": 5.513257, "mean_abs": 0.115629, "relative": 0.911019 }, { "name": "video_fisheye_cam2", "l2": 5.545141, "mean_abs": 0.113762, "relative": 0.885199 }, { "name": "video_fisheye_cam3", "l2": 5.67145, "mean_abs": 0.11558, "relative": 0.949616 }, { "name": "video_stereo_left", "l2": 5.718, "mean_abs": 0.122372, "relative": 0.929186 }, { "name": "video_stereo_right", "l2": 7.093254, "mean_abs": 0.152203, "relative": 1.0 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.045872, "mean_abs": 0.081293, "relative": 0.732187 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7470940351486206" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9534345269203186" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9972978234291077" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.38968807458877563" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1128, "start_frame": 5640, "end_frame": 5659, "center_frame": 5649, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.065212, "mean_abs": 0.244032, "relative": 0.849044 }, { "name": "hand_right_joints", "l2": 9.559566, "mean_abs": 0.255605, "relative": 0.75355 }, { "name": "body_joints", "l2": 6.269441, "mean_abs": 0.092437, "relative": 0.846808 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005086, "mean_abs": 0.00081, "relative": 0.012002 }, { "name": "camera_rotation_matrix", "l2": 2.999996, "mean_abs": 0.199982, "relative": 0.92125 }, { "name": "imu_accel_gyro", "l2": 17.038719, "mean_abs": 1.044209, "relative": 0.919298 }, { "name": "depth_confidence", "l2": 10.236306, "mean_abs": 0.18405, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.266851, "mean_abs": 0.108922, "relative": 0.953714 }, { "name": "video_fisheye_cam1", "l2": 5.502502, "mean_abs": 0.115254, "relative": 0.909242 }, { "name": "video_fisheye_cam2", "l2": 5.548537, "mean_abs": 0.11354, "relative": 0.885741 }, { "name": "video_fisheye_cam3", "l2": 5.655172, "mean_abs": 0.114985, "relative": 0.94689 }, { "name": "video_stereo_left", "l2": 5.722906, "mean_abs": 0.122002, "relative": 0.929984 }, { "name": "video_stereo_right", "l2": 7.088493, "mean_abs": 0.15167, "relative": 0.999329 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.050895, "mean_abs": 0.0816, "relative": 0.733985 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7996389269828796" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9672910571098328" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9971423745155334" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5211641788482666" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1129, "start_frame": 5645, "end_frame": 5664, "center_frame": 5654, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.063338, "mean_abs": 0.243716, "relative": 0.848886 }, { "name": "hand_right_joints", "l2": 9.54018, "mean_abs": 0.255836, "relative": 0.752021 }, { "name": "body_joints", "l2": 6.26743, "mean_abs": 0.092667, "relative": 0.846537 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004179, "mean_abs": 0.000565, "relative": 0.009862 }, { "name": "camera_rotation_matrix", "l2": 2.999943, "mean_abs": 0.19997, "relative": 0.921234 }, { "name": "imu_accel_gyro", "l2": 17.035929, "mean_abs": 1.041735, "relative": 0.919147 }, { "name": "depth_confidence", "l2": 10.243747, "mean_abs": 0.184699, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.269203, "mean_abs": 0.108952, "relative": 0.95414 }, { "name": "video_fisheye_cam1", "l2": 5.497125, "mean_abs": 0.115341, "relative": 0.908353 }, { "name": "video_fisheye_cam2", "l2": 5.551106, "mean_abs": 0.11356, "relative": 0.886152 }, { "name": "video_fisheye_cam3", "l2": 5.648496, "mean_abs": 0.114893, "relative": 0.945773 }, { "name": "video_stereo_left", "l2": 5.72657, "mean_abs": 0.122049, "relative": 0.930579 }, { "name": "video_stereo_right", "l2": 7.086268, "mean_abs": 0.151859, "relative": 0.999015 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.070024, "mean_abs": 0.079207, "relative": 0.740831 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8268386721611023" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9618256688117981" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9984737038612366" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.533637285232544" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1130, "start_frame": 5650, "end_frame": 5669, "center_frame": 5659, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.078746, "mean_abs": 0.243527, "relative": 0.850185 }, { "name": "hand_right_joints", "l2": 9.519706, "mean_abs": 0.254963, "relative": 0.750407 }, { "name": "body_joints", "l2": 6.261005, "mean_abs": 0.092795, "relative": 0.845669 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.007166, "mean_abs": 0.000902, "relative": 0.016911 }, { "name": "camera_rotation_matrix", "l2": 2.999943, "mean_abs": 0.199862, "relative": 0.921234 }, { "name": "imu_accel_gyro", "l2": 17.049614, "mean_abs": 1.048585, "relative": 0.919885 }, { "name": "depth_confidence", "l2": 10.231695, "mean_abs": 0.183564, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.264944, "mean_abs": 0.109133, "relative": 0.953369 }, { "name": "video_fisheye_cam1", "l2": 5.485186, "mean_abs": 0.11511, "relative": 0.906381 }, { "name": "video_fisheye_cam2", "l2": 5.552149, "mean_abs": 0.113654, "relative": 0.886318 }, { "name": "video_fisheye_cam3", "l2": 5.620965, "mean_abs": 0.114643, "relative": 0.941163 }, { "name": "video_stereo_left", "l2": 5.727011, "mean_abs": 0.12217, "relative": 0.930651 }, { "name": "video_stereo_right", "l2": 7.072985, "mean_abs": 0.15169, "relative": 0.997142 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.137127, "mean_abs": 0.084476, "relative": 0.764846 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8260108828544617" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9721234440803528" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9579705595970154" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.49553442001342773" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1131, "start_frame": 5655, "end_frame": 5674, "center_frame": 5664, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.074367, "mean_abs": 0.243269, "relative": 0.849816 }, { "name": "hand_right_joints", "l2": 9.511518, "mean_abs": 0.254014, "relative": 0.749762 }, { "name": "body_joints", "l2": 6.258208, "mean_abs": 0.092802, "relative": 0.845291 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009001, "mean_abs": 0.001173, "relative": 0.021242 }, { "name": "camera_rotation_matrix", "l2": 2.999979, "mean_abs": 0.200143, "relative": 0.921245 }, { "name": "imu_accel_gyro", "l2": 17.051472, "mean_abs": 1.046808, "relative": 0.919986 }, { "name": "depth_confidence", "l2": 10.236891, "mean_abs": 0.184253, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.261805, "mean_abs": 0.1091, "relative": 0.9528 }, { "name": "video_fisheye_cam1", "l2": 5.467733, "mean_abs": 0.114464, "relative": 0.903497 }, { "name": "video_fisheye_cam2", "l2": 5.552132, "mean_abs": 0.113777, "relative": 0.886315 }, { "name": "video_fisheye_cam3", "l2": 5.618605, "mean_abs": 0.114686, "relative": 0.940768 }, { "name": "video_stereo_left", "l2": 5.727709, "mean_abs": 0.122446, "relative": 0.930764 }, { "name": "video_stereo_right", "l2": 7.05787, "mean_abs": 0.151015, "relative": 0.995012 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.04406, "mean_abs": 0.079341, "relative": 0.731539 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8124262094497681" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9722992181777954" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.981903076171875" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5363159775733948" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1132, "start_frame": 5660, "end_frame": 5679, "center_frame": 5669, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.073606, "mean_abs": 0.243253, "relative": 0.849752 }, { "name": "hand_right_joints", "l2": 9.495184, "mean_abs": 0.253026, "relative": 0.748475 }, { "name": "body_joints", "l2": 6.249032, "mean_abs": 0.092677, "relative": 0.844052 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.00776, "mean_abs": 0.000977, "relative": 0.018313 }, { "name": "camera_rotation_matrix", "l2": 2.999992, "mean_abs": 0.199901, "relative": 0.921249 }, { "name": "imu_accel_gyro", "l2": 17.056587, "mean_abs": 1.046693, "relative": 0.920262 }, { "name": "depth_confidence", "l2": 10.216034, "mean_abs": 0.183671, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.260322, "mean_abs": 0.109085, "relative": 0.952532 }, { "name": "video_fisheye_cam1", "l2": 5.470292, "mean_abs": 0.114534, "relative": 0.90392 }, { "name": "video_fisheye_cam2", "l2": 5.55092, "mean_abs": 0.113689, "relative": 0.886122 }, { "name": "video_fisheye_cam3", "l2": 5.59493, "mean_abs": 0.114812, "relative": 0.936804 }, { "name": "video_stereo_left", "l2": 5.728023, "mean_abs": 0.122415, "relative": 0.930815 }, { "name": "video_stereo_right", "l2": 7.067912, "mean_abs": 0.151285, "relative": 0.996427 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.039, "mean_abs": 0.079478, "relative": 0.729728 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8579425811767578" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9508742690086365" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9988930821418762" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5478416085243225" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1133, "start_frame": 5665, "end_frame": 5684, "center_frame": 5674, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.067592, "mean_abs": 0.243409, "relative": 0.849244 }, { "name": "hand_right_joints", "l2": 9.479585, "mean_abs": 0.252995, "relative": 0.747245 }, { "name": "body_joints", "l2": 6.237403, "mean_abs": 0.092572, "relative": 0.842481 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004892, "mean_abs": 0.000697, "relative": 0.011544 }, { "name": "camera_rotation_matrix", "l2": 3.000045, "mean_abs": 0.200029, "relative": 0.921265 }, { "name": "imu_accel_gyro", "l2": 17.039528, "mean_abs": 1.047451, "relative": 0.919341 }, { "name": "depth_confidence", "l2": 10.22156, "mean_abs": 0.183867, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.257761, "mean_abs": 0.108843, "relative": 0.952068 }, { "name": "video_fisheye_cam1", "l2": 5.465128, "mean_abs": 0.114311, "relative": 0.903066 }, { "name": "video_fisheye_cam2", "l2": 5.550216, "mean_abs": 0.113633, "relative": 0.886009 }, { "name": "video_fisheye_cam3", "l2": 5.574593, "mean_abs": 0.114233, "relative": 0.933398 }, { "name": "video_stereo_left", "l2": 5.725475, "mean_abs": 0.122293, "relative": 0.930401 }, { "name": "video_stereo_right", "l2": 7.061422, "mean_abs": 0.151039, "relative": 0.995512 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.023367, "mean_abs": 0.077578, "relative": 0.724133 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8754261136054993" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9666813015937805" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.999097466468811" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5729432106018066" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1134, "start_frame": 5670, "end_frame": 5689, "center_frame": 5679, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.063907, "mean_abs": 0.243277, "relative": 0.848933 }, { "name": "hand_right_joints", "l2": 9.460475, "mean_abs": 0.25188, "relative": 0.745739 }, { "name": "body_joints", "l2": 6.23009, "mean_abs": 0.092473, "relative": 0.841493 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008978, "mean_abs": 0.001057, "relative": 0.021187 }, { "name": "camera_rotation_matrix", "l2": 3.000039, "mean_abs": 0.200358, "relative": 0.921263 }, { "name": "imu_accel_gyro", "l2": 17.027206, "mean_abs": 1.043661, "relative": 0.918676 }, { "name": "depth_confidence", "l2": 10.223166, "mean_abs": 0.183821, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.258534, "mean_abs": 0.109071, "relative": 0.952208 }, { "name": "video_fisheye_cam1", "l2": 5.463791, "mean_abs": 0.114774, "relative": 0.902845 }, { "name": "video_fisheye_cam2", "l2": 5.548092, "mean_abs": 0.113911, "relative": 0.88567 }, { "name": "video_fisheye_cam3", "l2": 5.565274, "mean_abs": 0.114737, "relative": 0.931838 }, { "name": "video_stereo_left", "l2": 5.718848, "mean_abs": 0.122702, "relative": 0.929324 }, { "name": "video_stereo_right", "l2": 7.059354, "mean_abs": 0.151589, "relative": 0.995221 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.042519, "mean_abs": 0.079381, "relative": 0.730987 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.836940348148346" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9463140964508057" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9987229704856873" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle away", "correct": 0, "confidence": "0.3463343679904938" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1135, "start_frame": 5675, "end_frame": 5694, "center_frame": 5684, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.062366, "mean_abs": 0.242802, "relative": 0.848803 }, { "name": "hand_right_joints", "l2": 9.434732, "mean_abs": 0.250792, "relative": 0.743709 }, { "name": "body_joints", "l2": 6.224891, "mean_abs": 0.092186, "relative": 0.840791 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008007, "mean_abs": 0.001087, "relative": 0.018895 }, { "name": "camera_rotation_matrix", "l2": 2.999925, "mean_abs": 0.199677, "relative": 0.921228 }, { "name": "imu_accel_gyro", "l2": 17.015226, "mean_abs": 1.040601, "relative": 0.91803 }, { "name": "depth_confidence", "l2": 10.214436, "mean_abs": 0.183724, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.26233, "mean_abs": 0.109137, "relative": 0.952895 }, { "name": "video_fisheye_cam1", "l2": 5.470016, "mean_abs": 0.114629, "relative": 0.903874 }, { "name": "video_fisheye_cam2", "l2": 5.550038, "mean_abs": 0.113708, "relative": 0.885981 }, { "name": "video_fisheye_cam3", "l2": 5.552131, "mean_abs": 0.114465, "relative": 0.929637 }, { "name": "video_stereo_left", "l2": 5.719078, "mean_abs": 0.122334, "relative": 0.929362 }, { "name": "video_stereo_right", "l2": 7.073784, "mean_abs": 0.15168, "relative": 0.997255 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.013669, "mean_abs": 0.075602, "relative": 0.720662 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.8294323682785034" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9479753971099854" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9976825714111328" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.3489214777946472" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1136, "start_frame": 5680, "end_frame": 5699, "center_frame": 5689, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.064062, "mean_abs": 0.242507, "relative": 0.848947 }, { "name": "hand_right_joints", "l2": 9.423446, "mean_abs": 0.249451, "relative": 0.74282 }, { "name": "body_joints", "l2": 6.223437, "mean_abs": 0.092392, "relative": 0.840595 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006911, "mean_abs": 0.001132, "relative": 0.016309 }, { "name": "camera_rotation_matrix", "l2": 3.000054, "mean_abs": 0.200232, "relative": 0.921268 }, { "name": "imu_accel_gyro", "l2": 17.031872, "mean_abs": 1.043695, "relative": 0.918928 }, { "name": "depth_confidence", "l2": 10.220632, "mean_abs": 0.184205, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.258864, "mean_abs": 0.108958, "relative": 0.952268 }, { "name": "video_fisheye_cam1", "l2": 5.464369, "mean_abs": 0.114306, "relative": 0.902941 }, { "name": "video_fisheye_cam2", "l2": 5.548045, "mean_abs": 0.113695, "relative": 0.885663 }, { "name": "video_fisheye_cam3", "l2": 5.524914, "mean_abs": 0.11251, "relative": 0.92508 }, { "name": "video_stereo_left", "l2": 5.713972, "mean_abs": 0.122418, "relative": 0.928532 }, { "name": "video_stereo_right", "l2": 7.062739, "mean_abs": 0.15117, "relative": 0.995698 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.00479, "mean_abs": 0.076235, "relative": 0.717484 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7252582907676697" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9546493887901306" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9727597236633301" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle away", "correct": 0, "confidence": "0.3580490052700043" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1137, "start_frame": 5685, "end_frame": 5704, "center_frame": 5694, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.086362, "mean_abs": 0.244537, "relative": 0.850828 }, { "name": "hand_right_joints", "l2": 9.429072, "mean_abs": 0.250662, "relative": 0.743263 }, { "name": "body_joints", "l2": 6.233791, "mean_abs": 0.093042, "relative": 0.841993 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.01421, "mean_abs": 0.002018, "relative": 0.033533 }, { "name": "camera_rotation_matrix", "l2": 3.000134, "mean_abs": 0.201512, "relative": 0.921293 }, { "name": "imu_accel_gyro", "l2": 17.05825, "mean_abs": 1.064984, "relative": 0.920351 }, { "name": "depth_confidence", "l2": 10.170514, "mean_abs": 0.185312, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.25758, "mean_abs": 0.109237, "relative": 0.952035 }, { "name": "video_fisheye_cam1", "l2": 5.465993, "mean_abs": 0.114964, "relative": 0.903209 }, { "name": "video_fisheye_cam2", "l2": 5.545391, "mean_abs": 0.114056, "relative": 0.885239 }, { "name": "video_fisheye_cam3", "l2": 5.531321, "mean_abs": 0.113492, "relative": 0.926153 }, { "name": "video_stereo_left", "l2": 5.710792, "mean_abs": 0.122953, "relative": 0.928015 }, { "name": "video_stereo_right", "l2": 7.064595, "mean_abs": 0.152125, "relative": 0.99596 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.213629, "mean_abs": 0.085409, "relative": 0.792225 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6135447025299072" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9196842908859253" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9781398773193359" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.33424824476242065" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1138, "start_frame": 5690, "end_frame": 5709, "center_frame": 5699, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.095641, "mean_abs": 0.245151, "relative": 0.85161 }, { "name": "hand_right_joints", "l2": 9.443933, "mean_abs": 0.25255, "relative": 0.744435 }, { "name": "body_joints", "l2": 6.250021, "mean_abs": 0.09343, "relative": 0.844185 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.012806, "mean_abs": 0.001777, "relative": 0.030219 }, { "name": "camera_rotation_matrix", "l2": 3.000101, "mean_abs": 0.201309, "relative": 0.921282 }, { "name": "imu_accel_gyro", "l2": 17.040339, "mean_abs": 1.052267, "relative": 0.919385 }, { "name": "depth_confidence", "l2": 10.174012, "mean_abs": 0.185347, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.257417, "mean_abs": 0.109238, "relative": 0.952006 }, { "name": "video_fisheye_cam1", "l2": 5.468487, "mean_abs": 0.114937, "relative": 0.903621 }, { "name": "video_fisheye_cam2", "l2": 5.545998, "mean_abs": 0.113942, "relative": 0.885336 }, { "name": "video_fisheye_cam3", "l2": 5.532106, "mean_abs": 0.113463, "relative": 0.926284 }, { "name": "video_stereo_left", "l2": 5.711677, "mean_abs": 0.122905, "relative": 0.928159 }, { "name": "video_stereo_right", "l2": 7.066646, "mean_abs": 0.152107, "relative": 0.996249 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.312998, "mean_abs": 0.088828, "relative": 0.827787 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5833576917648315" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9270859360694885" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9505512118339539" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.3451521694660187" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1139, "start_frame": 5695, "end_frame": 5714, "center_frame": 5704, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.092175, "mean_abs": 0.245365, "relative": 0.851318 }, { "name": "hand_right_joints", "l2": 9.470158, "mean_abs": 0.252795, "relative": 0.746502 }, { "name": "body_joints", "l2": 6.267929, "mean_abs": 0.093736, "relative": 0.846604 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.008242, "mean_abs": 0.001146, "relative": 0.01945 }, { "name": "camera_rotation_matrix", "l2": 2.999643, "mean_abs": 0.200977, "relative": 0.921142 }, { "name": "imu_accel_gyro", "l2": 17.045952, "mean_abs": 1.053968, "relative": 0.919688 }, { "name": "depth_confidence", "l2": 10.159229, "mean_abs": 0.184653, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.254735, "mean_abs": 0.108921, "relative": 0.95152 }, { "name": "video_fisheye_cam1", "l2": 5.463255, "mean_abs": 0.114482, "relative": 0.902757 }, { "name": "video_fisheye_cam2", "l2": 5.541255, "mean_abs": 0.113628, "relative": 0.884579 }, { "name": "video_fisheye_cam3", "l2": 5.522076, "mean_abs": 0.112831, "relative": 0.924605 }, { "name": "video_stereo_left", "l2": 5.701389, "mean_abs": 0.122379, "relative": 0.926487 }, { "name": "video_stereo_right", "l2": 7.050337, "mean_abs": 0.151203, "relative": 0.99395 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.191717, "mean_abs": 0.085168, "relative": 0.784383 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5483040809631348" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9412733316421509" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9946086406707764" }, "next_action": { "true": "Pour coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.2823852300643921" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1140, "start_frame": 5700, "end_frame": 5719, "center_frame": 5709, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.097123, "mean_abs": 0.246536, "relative": 0.851735 }, { "name": "hand_right_joints", "l2": 9.484316, "mean_abs": 0.253926, "relative": 0.747618 }, { "name": "body_joints", "l2": 6.288382, "mean_abs": 0.094346, "relative": 0.849367 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006476, "mean_abs": 0.001089, "relative": 0.015283 }, { "name": "camera_rotation_matrix", "l2": 3.000037, "mean_abs": 0.200697, "relative": 0.921263 }, { "name": "imu_accel_gyro", "l2": 17.030117, "mean_abs": 1.057282, "relative": 0.918833 }, { "name": "depth_confidence", "l2": 10.146946, "mean_abs": 0.18457, "relative": 0.000337 }, { "name": "video_fisheye_cam0", "l2": 5.254378, "mean_abs": 0.108771, "relative": 0.951455 }, { "name": "video_fisheye_cam1", "l2": 5.465064, "mean_abs": 0.114352, "relative": 0.903056 }, { "name": "video_fisheye_cam2", "l2": 5.538476, "mean_abs": 0.113496, "relative": 0.884135 }, { "name": "video_fisheye_cam3", "l2": 5.526787, "mean_abs": 0.112763, "relative": 0.925394 }, { "name": "video_stereo_left", "l2": 5.694042, "mean_abs": 0.121941, "relative": 0.925293 }, { "name": "video_stereo_right", "l2": 7.046618, "mean_abs": 0.15073, "relative": 0.993425 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.156142, "mean_abs": 0.083304, "relative": 0.771651 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5899620652198792" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9449734687805176" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9960111379623413" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.33416852355003357" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1141, "start_frame": 5705, "end_frame": 5724, "center_frame": 5714, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.092442, "mean_abs": 0.247324, "relative": 0.851341 }, { "name": "hand_right_joints", "l2": 9.489449, "mean_abs": 0.254364, "relative": 0.748022 }, { "name": "body_joints", "l2": 6.310168, "mean_abs": 0.094861, "relative": 0.852309 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006699, "mean_abs": 0.001064, "relative": 0.015808 }, { "name": "camera_rotation_matrix", "l2": 3.000117, "mean_abs": 0.200878, "relative": 0.921287 }, { "name": "imu_accel_gyro", "l2": 17.046568, "mean_abs": 1.052246, "relative": 0.919721 }, { "name": "depth_confidence", "l2": 10.186992, "mean_abs": 0.183916, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.257379, "mean_abs": 0.108852, "relative": 0.951999 }, { "name": "video_fisheye_cam1", "l2": 5.472472, "mean_abs": 0.114542, "relative": 0.90428 }, { "name": "video_fisheye_cam2", "l2": 5.540291, "mean_abs": 0.113556, "relative": 0.884425 }, { "name": "video_fisheye_cam3", "l2": 5.519523, "mean_abs": 0.112575, "relative": 0.924178 }, { "name": "video_stereo_left", "l2": 5.704764, "mean_abs": 0.122204, "relative": 0.927035 }, { "name": "video_stereo_right", "l2": 7.066731, "mean_abs": 0.151273, "relative": 0.996261 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.156763, "mean_abs": 0.085291, "relative": 0.771873 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6369968056678772" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9347750544548035" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.993444561958313" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.3586760461330414" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1142, "start_frame": 5710, "end_frame": 5729, "center_frame": 5719, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.084433, "mean_abs": 0.247139, "relative": 0.850665 }, { "name": "hand_right_joints", "l2": 9.483509, "mean_abs": 0.254141, "relative": 0.747554 }, { "name": "body_joints", "l2": 6.314161, "mean_abs": 0.094944, "relative": 0.852849 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081681, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004415, "mean_abs": 0.000761, "relative": 0.010419 }, { "name": "camera_rotation_matrix", "l2": 2.999795, "mean_abs": 0.200631, "relative": 0.921188 }, { "name": "imu_accel_gyro", "l2": 17.058647, "mean_abs": 1.056452, "relative": 0.920373 }, { "name": "depth_confidence", "l2": 10.186899, "mean_abs": 0.183957, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.25877, "mean_abs": 0.108839, "relative": 0.952251 }, { "name": "video_fisheye_cam1", "l2": 5.473012, "mean_abs": 0.114461, "relative": 0.904369 }, { "name": "video_fisheye_cam2", "l2": 5.541061, "mean_abs": 0.11345, "relative": 0.884548 }, { "name": "video_fisheye_cam3", "l2": 5.51535, "mean_abs": 0.112259, "relative": 0.923479 }, { "name": "video_stereo_left", "l2": 5.707391, "mean_abs": 0.122239, "relative": 0.927462 }, { "name": "video_stereo_right", "l2": 7.071162, "mean_abs": 0.151524, "relative": 0.996885 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.06581, "mean_abs": 0.081325, "relative": 0.739322 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7437241673469543" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9536181688308716" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9970182180404663" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.39474138617515564" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1143, "start_frame": 5715, "end_frame": 5734, "center_frame": 5724, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.087306, "mean_abs": 0.246752, "relative": 0.850907 }, { "name": "hand_right_joints", "l2": 9.479769, "mean_abs": 0.253822, "relative": 0.747259 }, { "name": "body_joints", "l2": 6.310988, "mean_abs": 0.094942, "relative": 0.85242 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.005035, "mean_abs": 0.000706, "relative": 0.011883 }, { "name": "camera_rotation_matrix", "l2": 3.000136, "mean_abs": 0.200885, "relative": 0.921293 }, { "name": "imu_accel_gyro", "l2": 17.059399, "mean_abs": 1.052033, "relative": 0.920413 }, { "name": "depth_confidence", "l2": 10.197806, "mean_abs": 0.184531, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.25872, "mean_abs": 0.108865, "relative": 0.952242 }, { "name": "video_fisheye_cam1", "l2": 5.474783, "mean_abs": 0.114612, "relative": 0.904662 }, { "name": "video_fisheye_cam2", "l2": 5.544226, "mean_abs": 0.113572, "relative": 0.885053 }, { "name": "video_fisheye_cam3", "l2": 5.512935, "mean_abs": 0.112259, "relative": 0.923074 }, { "name": "video_stereo_left", "l2": 5.710423, "mean_abs": 0.12231, "relative": 0.927955 }, { "name": "video_stereo_right", "l2": 7.075365, "mean_abs": 0.151595, "relative": 0.997478 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.026793, "mean_abs": 0.077569, "relative": 0.725359 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7498315572738647" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9622494578361511" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9154862761497498" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.40659990906715393" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1144, "start_frame": 5720, "end_frame": 5739, "center_frame": 5729, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.090603, "mean_abs": 0.246872, "relative": 0.851185 }, { "name": "hand_right_joints", "l2": 9.475477, "mean_abs": 0.253584, "relative": 0.746921 }, { "name": "body_joints", "l2": 6.312623, "mean_abs": 0.094668, "relative": 0.852641 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.003614, "mean_abs": 0.000597, "relative": 0.008529 }, { "name": "camera_rotation_matrix", "l2": 3.000133, "mean_abs": 0.200812, "relative": 0.921292 }, { "name": "imu_accel_gyro", "l2": 17.059254, "mean_abs": 1.053892, "relative": 0.920406 }, { "name": "depth_confidence", "l2": 10.241829, "mean_abs": 0.184715, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.260107, "mean_abs": 0.108878, "relative": 0.952493 }, { "name": "video_fisheye_cam1", "l2": 5.475323, "mean_abs": 0.114487, "relative": 0.904751 }, { "name": "video_fisheye_cam2", "l2": 5.546144, "mean_abs": 0.113556, "relative": 0.885359 }, { "name": "video_fisheye_cam3", "l2": 5.512387, "mean_abs": 0.112307, "relative": 0.922983 }, { "name": "video_stereo_left", "l2": 5.711954, "mean_abs": 0.122109, "relative": 0.928204 }, { "name": "video_stereo_right", "l2": 7.077909, "mean_abs": 0.151188, "relative": 0.997837 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.04119, "mean_abs": 0.078851, "relative": 0.730511 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7253323197364807" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9690845608711243" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.7527762055397034" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.38282790780067444" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1145, "start_frame": 5725, "end_frame": 5744, "center_frame": 5734, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.093362, "mean_abs": 0.247032, "relative": 0.851418 }, { "name": "hand_right_joints", "l2": 9.482217, "mean_abs": 0.253969, "relative": 0.747452 }, { "name": "body_joints", "l2": 6.308859, "mean_abs": 0.094641, "relative": 0.852132 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.002846, "mean_abs": 0.000454, "relative": 0.006716 }, { "name": "camera_rotation_matrix", "l2": 2.999981, "mean_abs": 0.20037, "relative": 0.921246 }, { "name": "imu_accel_gyro", "l2": 17.045809, "mean_abs": 1.057233, "relative": 0.91968 }, { "name": "depth_confidence", "l2": 10.199845, "mean_abs": 0.184181, "relative": 0.000339 }, { "name": "video_fisheye_cam0", "l2": 5.24461, "mean_abs": 0.108312, "relative": 0.949687 }, { "name": "video_fisheye_cam1", "l2": 5.442099, "mean_abs": 0.113576, "relative": 0.899261 }, { "name": "video_fisheye_cam2", "l2": 5.539443, "mean_abs": 0.113226, "relative": 0.88429 }, { "name": "video_fisheye_cam3", "l2": 5.471778, "mean_abs": 0.111365, "relative": 0.916183 }, { "name": "video_stereo_left", "l2": 5.691505, "mean_abs": 0.121315, "relative": 0.924881 }, { "name": "video_stereo_right", "l2": 7.02743, "mean_abs": 0.14957, "relative": 0.99072 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.086701, "mean_abs": 0.081135, "relative": 0.746799 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5327427387237549" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9739237427711487" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.5975668430328369" }, "next_action": { "true": "Pour coffee", "predicted": "Move kettle away", "correct": 0, "confidence": "0.35638996958732605" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1146, "start_frame": 5730, "end_frame": 5749, "center_frame": 5739, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.099405, "mean_abs": 0.246544, "relative": 0.851928 }, { "name": "hand_right_joints", "l2": 9.482005, "mean_abs": 0.253486, "relative": 0.747436 }, { "name": "body_joints", "l2": 6.305134, "mean_abs": 0.094493, "relative": 0.851629 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006917, "mean_abs": 0.001013, "relative": 0.016324 }, { "name": "camera_rotation_matrix", "l2": 3.000047, "mean_abs": 0.200825, "relative": 0.921266 }, { "name": "imu_accel_gyro", "l2": 17.041649, "mean_abs": 1.054955, "relative": 0.919456 }, { "name": "depth_confidence", "l2": 10.197343, "mean_abs": 0.183915, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.246989, "mean_abs": 0.108498, "relative": 0.950117 }, { "name": "video_fisheye_cam1", "l2": 5.443689, "mean_abs": 0.113916, "relative": 0.899524 }, { "name": "video_fisheye_cam2", "l2": 5.541657, "mean_abs": 0.11346, "relative": 0.884643 }, { "name": "video_fisheye_cam3", "l2": 5.479082, "mean_abs": 0.111548, "relative": 0.917406 }, { "name": "video_stereo_left", "l2": 5.692785, "mean_abs": 0.1217, "relative": 0.925089 }, { "name": "video_stereo_right", "l2": 7.029201, "mean_abs": 0.149968, "relative": 0.99097 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.025537, "mean_abs": 0.077654, "relative": 0.724909 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.577114999294281" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9753735065460205" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9394167065620422" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.32109957933425903" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "scale|coffee jar|coffee scoop|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1147, "start_frame": 5735, "end_frame": 5754, "center_frame": 5744, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.095865, "mean_abs": 0.246377, "relative": 0.851629 }, { "name": "hand_right_joints", "l2": 9.452999, "mean_abs": 0.253983, "relative": 0.745149 }, { "name": "body_joints", "l2": 6.31294, "mean_abs": 0.094902, "relative": 0.852684 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081683, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.006185, "mean_abs": 0.000992, "relative": 0.014596 }, { "name": "camera_rotation_matrix", "l2": 2.999957, "mean_abs": 0.200864, "relative": 0.921238 }, { "name": "imu_accel_gyro", "l2": 17.042131, "mean_abs": 1.052184, "relative": 0.919482 }, { "name": "depth_confidence", "l2": 10.192371, "mean_abs": 0.184257, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.263728, "mean_abs": 0.109183, "relative": 0.953148 }, { "name": "video_fisheye_cam1", "l2": 5.474863, "mean_abs": 0.114911, "relative": 0.904675 }, { "name": "video_fisheye_cam2", "l2": 5.549911, "mean_abs": 0.113912, "relative": 0.885961 }, { "name": "video_fisheye_cam3", "l2": 5.541595, "mean_abs": 0.113464, "relative": 0.927873 }, { "name": "video_stereo_left", "l2": 5.708506, "mean_abs": 0.122576, "relative": 0.927643 }, { "name": "video_stereo_right", "l2": 7.078542, "mean_abs": 0.152197, "relative": 0.997926 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.042722, "mean_abs": 0.079629, "relative": 0.731059 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7014224529266357" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9567964673042297" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9979169964790344" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.589388906955719" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1148, "start_frame": 5740, "end_frame": 5759, "center_frame": 5749, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.096919, "mean_abs": 0.246313, "relative": 0.851718 }, { "name": "hand_right_joints", "l2": 9.460622, "mean_abs": 0.254474, "relative": 0.74575 }, { "name": "body_joints", "l2": 6.344164, "mean_abs": 0.095958, "relative": 0.856901 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081682, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.004664, "mean_abs": 0.000782, "relative": 0.011006 }, { "name": "camera_rotation_matrix", "l2": 3.000037, "mean_abs": 0.20056, "relative": 0.921263 }, { "name": "imu_accel_gyro", "l2": 17.04711, "mean_abs": 1.050149, "relative": 0.91975 }, { "name": "depth_confidence", "l2": 10.190425, "mean_abs": 0.18419, "relative": 0.000338 }, { "name": "video_fisheye_cam0", "l2": 5.263543, "mean_abs": 0.108943, "relative": 0.953115 }, { "name": "video_fisheye_cam1", "l2": 5.465238, "mean_abs": 0.114714, "relative": 0.903084 }, { "name": "video_fisheye_cam2", "l2": 5.549244, "mean_abs": 0.11382, "relative": 0.885854 }, { "name": "video_fisheye_cam3", "l2": 5.541514, "mean_abs": 0.113035, "relative": 0.92786 }, { "name": "video_stereo_left", "l2": 5.707737, "mean_abs": 0.1226, "relative": 0.927519 }, { "name": "video_stereo_right", "l2": 7.04599, "mean_abs": 0.151988, "relative": 0.993337 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.075892, "mean_abs": 0.080153, "relative": 0.742931 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.010333, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5886632800102234" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9465656876564026" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9975457787513733" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.42387688159942627" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1149, "start_frame": 5745, "end_frame": 5764, "center_frame": 5754, "action": "Pour coffee", "subtask": "Pour coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.084815, "mean_abs": 0.24629, "relative": 0.850697 }, { "name": "hand_right_joints", "l2": 9.499365, "mean_abs": 0.259598, "relative": 0.748804 }, { "name": "body_joints", "l2": 6.384457, "mean_abs": 0.097425, "relative": 0.862343 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.009078, "mean_abs": 0.001307, "relative": 0.021422 }, { "name": "camera_rotation_matrix", "l2": 3.00011, "mean_abs": 0.201558, "relative": 0.921285 }, { "name": "imu_accel_gyro", "l2": 17.055698, "mean_abs": 1.060232, "relative": 0.920214 }, { "name": "depth_confidence", "l2": 10.250218, "mean_abs": 0.186336, "relative": 0.00034 }, { "name": "video_fisheye_cam0", "l2": 5.264552, "mean_abs": 0.109054, "relative": 0.953298 }, { "name": "video_fisheye_cam1", "l2": 5.450603, "mean_abs": 0.115127, "relative": 0.900666 }, { "name": "video_fisheye_cam2", "l2": 5.551125, "mean_abs": 0.11421, "relative": 0.886154 }, { "name": "video_fisheye_cam3", "l2": 5.543911, "mean_abs": 0.113074, "relative": 0.928261 }, { "name": "video_stereo_left", "l2": 5.725635, "mean_abs": 0.123949, "relative": 0.930427 }, { "name": "video_stereo_right", "l2": 6.988119, "mean_abs": 0.153336, "relative": 0.985178 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.037431, "mean_abs": 0.078824, "relative": 0.729166 }, { "name": "caption_objects_interaction_text", "l2": 1.829797, "mean_abs": 0.015642, "relative": 0.840199 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6596003770828247" }, "timeline_subtask": { "true": "Pour coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7727599740028381" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9139751195907593" }, "next_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.4004504978656769" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1150, "start_frame": 5750, "end_frame": 5769, "center_frame": 5759, "action": "Pour coffee", "subtask": "", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.071373, "mean_abs": 0.245925, "relative": 0.849563 }, { "name": "hand_right_joints", "l2": 9.528248, "mean_abs": 0.263457, "relative": 0.751081 }, { "name": "body_joints", "l2": 6.411811, "mean_abs": 0.098518, "relative": 0.866038 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.024132, "mean_abs": 0.003473, "relative": 0.056948 }, { "name": "camera_rotation_matrix", "l2": 3.000512, "mean_abs": 0.203181, "relative": 0.921408 }, { "name": "imu_accel_gyro", "l2": 17.041826, "mean_abs": 1.069275, "relative": 0.919465 }, { "name": "depth_confidence", "l2": 10.27035, "mean_abs": 0.187555, "relative": 0.000341 }, { "name": "video_fisheye_cam0", "l2": 5.266533, "mean_abs": 0.109436, "relative": 0.953656 }, { "name": "video_fisheye_cam1", "l2": 5.436998, "mean_abs": 0.116116, "relative": 0.898418 }, { "name": "video_fisheye_cam2", "l2": 5.549828, "mean_abs": 0.114984, "relative": 0.885947 }, { "name": "video_fisheye_cam3", "l2": 5.528473, "mean_abs": 0.11339, "relative": 0.925676 }, { "name": "video_stereo_left", "l2": 5.725069, "mean_abs": 0.125204, "relative": 0.930335 }, { "name": "video_stereo_right", "l2": 6.949545, "mean_abs": 0.154801, "relative": 0.97974 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.085651, "mean_abs": 0.081122, "relative": 0.746423 }, { "name": "caption_objects_interaction_text", "l2": 1.829797, "mean_abs": 0.01603, "relative": 0.840199 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5507420897483826" }, "timeline_subtask": null, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9858696460723877" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.3756162226200104" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1151, "start_frame": 5755, "end_frame": 5774, "center_frame": 5764, "action": "Pour coffee", "subtask": "Pour milk into coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.067454, "mean_abs": 0.245517, "relative": 0.849233 }, { "name": "hand_right_joints", "l2": 9.563107, "mean_abs": 0.264319, "relative": 0.753829 }, { "name": "body_joints", "l2": 6.427335, "mean_abs": 0.098589, "relative": 0.868135 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.08168, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.04151, "mean_abs": 0.006107, "relative": 0.097957 }, { "name": "camera_rotation_matrix", "l2": 3.001894, "mean_abs": 0.205723, "relative": 0.921833 }, { "name": "imu_accel_gyro", "l2": 17.0669, "mean_abs": 1.076867, "relative": 0.920818 }, { "name": "depth_confidence", "l2": 10.304723, "mean_abs": 0.189957, "relative": 0.000342 }, { "name": "video_fisheye_cam0", "l2": 5.2583, "mean_abs": 0.109734, "relative": 0.952166 }, { "name": "video_fisheye_cam1", "l2": 5.412863, "mean_abs": 0.116782, "relative": 0.89443 }, { "name": "video_fisheye_cam2", "l2": 5.546545, "mean_abs": 0.115716, "relative": 0.885423 }, { "name": "video_fisheye_cam3", "l2": 5.480196, "mean_abs": 0.112875, "relative": 0.917593 }, { "name": "video_stereo_left", "l2": 5.707369, "mean_abs": 0.126642, "relative": 0.927459 }, { "name": "video_stereo_right", "l2": 6.858477, "mean_abs": 0.154607, "relative": 0.966901 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.115391, "mean_abs": 0.082878, "relative": 0.757067 }, { "name": "caption_objects_interaction_text", "l2": 1.829797, "mean_abs": 0.01609, "relative": 0.840199 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.5202066898345947" }, "timeline_subtask": { "true": "Pour milk into coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7727410793304443" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9975638389587402" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.6571423411369324" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "scale|coffee jar|coffee scoop|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1152, "start_frame": 5760, "end_frame": 5779, "center_frame": 5769, "action": "Pour coffee", "subtask": "Pour milk into coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.061744, "mean_abs": 0.244116, "relative": 0.848751 }, { "name": "hand_right_joints", "l2": 9.601555, "mean_abs": 0.263905, "relative": 0.756859 }, { "name": "body_joints", "l2": 6.442283, "mean_abs": 0.097106, "relative": 0.870154 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081679, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.04314, "mean_abs": 0.006585, "relative": 0.101805 }, { "name": "camera_rotation_matrix", "l2": 3.001729, "mean_abs": 0.205794, "relative": 0.921782 }, { "name": "imu_accel_gyro", "l2": 17.023399, "mean_abs": 1.067536, "relative": 0.918471 }, { "name": "depth_confidence", "l2": 10.346087, "mean_abs": 0.189624, "relative": 0.000343 }, { "name": "video_fisheye_cam0", "l2": 5.256841, "mean_abs": 0.109639, "relative": 0.951901 }, { "name": "video_fisheye_cam1", "l2": 5.38561, "mean_abs": 0.115988, "relative": 0.889927 }, { "name": "video_fisheye_cam2", "l2": 5.540584, "mean_abs": 0.115455, "relative": 0.884472 }, { "name": "video_fisheye_cam3", "l2": 5.464839, "mean_abs": 0.112481, "relative": 0.915021 }, { "name": "video_stereo_left", "l2": 5.687788, "mean_abs": 0.125757, "relative": 0.924277 }, { "name": "video_stereo_right", "l2": 6.735651, "mean_abs": 0.151448, "relative": 0.949585 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.047162, "mean_abs": 0.079772, "relative": 0.732649 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013017, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.6833415627479553" }, "timeline_subtask": { "true": "Pour milk into coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8758783936500549" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999032020568848" }, "next_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.5249497890472412" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "scale|coffee jar|coffee scoop|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1153, "start_frame": 5765, "end_frame": 5784, "center_frame": 5774, "action": "Pour coffee", "subtask": "Pour milk into coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.017312, "mean_abs": 0.24178, "relative": 0.845003 }, { "name": "hand_right_joints", "l2": 9.637421, "mean_abs": 0.261295, "relative": 0.759687 }, { "name": "body_joints", "l2": 6.430256, "mean_abs": 0.095771, "relative": 0.868529 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.037358, "mean_abs": 0.00554, "relative": 0.088161 }, { "name": "camera_rotation_matrix", "l2": 3.00138, "mean_abs": 0.204798, "relative": 0.921675 }, { "name": "imu_accel_gyro", "l2": 17.026627, "mean_abs": 1.059569, "relative": 0.918645 }, { "name": "depth_confidence", "l2": 10.377076, "mean_abs": 0.190157, "relative": 0.000344 }, { "name": "video_fisheye_cam0", "l2": 5.258606, "mean_abs": 0.109446, "relative": 0.952221 }, { "name": "video_fisheye_cam1", "l2": 5.356891, "mean_abs": 0.114583, "relative": 0.885181 }, { "name": "video_fisheye_cam2", "l2": 5.539943, "mean_abs": 0.114986, "relative": 0.884369 }, { "name": "video_fisheye_cam3", "l2": 5.463233, "mean_abs": 0.112305, "relative": 0.914753 }, { "name": "video_stereo_left", "l2": 5.674423, "mean_abs": 0.124409, "relative": 0.922105 }, { "name": "video_stereo_right", "l2": 6.609617, "mean_abs": 0.146211, "relative": 0.931817 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.001575, "mean_abs": 0.077819, "relative": 0.716334 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013017, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7926583290100098" }, "timeline_subtask": { "true": "Pour milk into coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9480108618736267" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9999595880508423" }, "next_action": { "true": "Pour milk into coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.4794361889362335" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "scale|coffee jar|coffee scoop|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1154, "start_frame": 5770, "end_frame": 5789, "center_frame": 5779, "action": "Pour coffee", "subtask": "Pour milk into coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 10.004093, "mean_abs": 0.24063, "relative": 0.843888 }, { "name": "hand_right_joints", "l2": 9.650276, "mean_abs": 0.259932, "relative": 0.7607 }, { "name": "body_joints", "l2": 6.411302, "mean_abs": 0.094597, "relative": 0.865969 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.020059, "mean_abs": 0.002991, "relative": 0.047337 }, { "name": "camera_rotation_matrix", "l2": 3.000632, "mean_abs": 0.202406, "relative": 0.921445 }, { "name": "imu_accel_gyro", "l2": 17.044659, "mean_abs": 1.044443, "relative": 0.919618 }, { "name": "depth_confidence", "l2": 10.425308, "mean_abs": 0.189318, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.262877, "mean_abs": 0.109156, "relative": 0.952994 }, { "name": "video_fisheye_cam1", "l2": 5.351563, "mean_abs": 0.112793, "relative": 0.884301 }, { "name": "video_fisheye_cam2", "l2": 5.535053, "mean_abs": 0.113945, "relative": 0.883589 }, { "name": "video_fisheye_cam3", "l2": 5.490536, "mean_abs": 0.11267, "relative": 0.919324 }, { "name": "video_stereo_left", "l2": 5.641026, "mean_abs": 0.122053, "relative": 0.916678 }, { "name": "video_stereo_right", "l2": 6.564358, "mean_abs": 0.142725, "relative": 0.925437 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.966352, "mean_abs": 0.07848, "relative": 0.703728 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013017, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.7460301518440247" }, "timeline_subtask": { "true": "Pour milk into coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.9136845469474792" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9988318085670471" }, "next_action": { "true": "Pour milk into coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.5746459364891052" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1155, "start_frame": 5775, "end_frame": 5794, "center_frame": 5784, "action": "Pour coffee", "subtask": "Pour milk into coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.995137, "mean_abs": 0.238414, "relative": 0.843132 }, { "name": "hand_right_joints", "l2": 9.652073, "mean_abs": 0.259458, "relative": 0.760842 }, { "name": "body_joints", "l2": 6.40255, "mean_abs": 0.093896, "relative": 0.864787 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.014874, "mean_abs": 0.002249, "relative": 0.035099 }, { "name": "camera_rotation_matrix", "l2": 2.999837, "mean_abs": 0.201791, "relative": 0.921201 }, { "name": "imu_accel_gyro", "l2": 17.034506, "mean_abs": 1.031186, "relative": 0.91907 }, { "name": "depth_confidence", "l2": 10.473739, "mean_abs": 0.19011, "relative": 0.000348 }, { "name": "video_fisheye_cam0", "l2": 5.259641, "mean_abs": 0.108744, "relative": 0.952408 }, { "name": "video_fisheye_cam1", "l2": 5.353755, "mean_abs": 0.112207, "relative": 0.884663 }, { "name": "video_fisheye_cam2", "l2": 5.528993, "mean_abs": 0.113394, "relative": 0.882622 }, { "name": "video_fisheye_cam3", "l2": 5.523591, "mean_abs": 0.11347, "relative": 0.924859 }, { "name": "video_stereo_left", "l2": 5.620975, "mean_abs": 0.121122, "relative": 0.91342 }, { "name": "video_stereo_right", "l2": 6.545689, "mean_abs": 0.141797, "relative": 0.922805 }, { "name": "audio_fisheye_cam0_aac", "l2": 1.933824, "mean_abs": 0.074595, "relative": 0.692086 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013017, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.557888925075531" }, "timeline_subtask": { "true": "Pour milk into coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8432262539863586" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9977885484695435" }, "next_action": { "true": "Pour milk into coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.5135923624038696" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1156, "start_frame": 5780, "end_frame": 5799, "center_frame": 5789, "action": "Pour coffee", "subtask": "Pour milk into coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.962274, "mean_abs": 0.236907, "relative": 0.84036 }, { "name": "hand_right_joints", "l2": 9.694099, "mean_abs": 0.260171, "relative": 0.764154 }, { "name": "body_joints", "l2": 6.408846, "mean_abs": 0.093843, "relative": 0.865638 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.017651, "mean_abs": 0.00237, "relative": 0.041654 }, { "name": "camera_rotation_matrix", "l2": 3.000073, "mean_abs": 0.20226, "relative": 0.921274 }, { "name": "imu_accel_gyro", "l2": 17.040993, "mean_abs": 1.03015, "relative": 0.91942 }, { "name": "depth_confidence", "l2": 10.503313, "mean_abs": 0.190115, "relative": 0.000349 }, { "name": "video_fisheye_cam0", "l2": 5.259331, "mean_abs": 0.108731, "relative": 0.952352 }, { "name": "video_fisheye_cam1", "l2": 5.359968, "mean_abs": 0.112429, "relative": 0.885689 }, { "name": "video_fisheye_cam2", "l2": 5.529462, "mean_abs": 0.113384, "relative": 0.882696 }, { "name": "video_fisheye_cam3", "l2": 5.57421, "mean_abs": 0.114506, "relative": 0.933334 }, { "name": "video_stereo_left", "l2": 5.617218, "mean_abs": 0.121047, "relative": 0.912809 }, { "name": "video_stereo_right", "l2": 6.546333, "mean_abs": 0.141755, "relative": 0.922896 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.045172, "mean_abs": 0.080443, "relative": 0.731936 }, { "name": "caption_objects_interaction_text", "l2": 1.732051, "mean_abs": 0.013017, "relative": 0.795316 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.48130643367767334" }, "timeline_subtask": { "true": "Pour milk into coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.730074405670166" }, "transition_detection": { "true": "steady", "predicted": "steady", "correct": 1, "confidence": "0.9985013008117676" }, "next_action": { "true": "Pour milk into coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.5427289605140686" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|milk pitcher|coffee cup", "predicted": "coffee dripper|scale|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1157, "start_frame": 5785, "end_frame": 5804, "center_frame": 5794, "action": "Pour coffee", "subtask": "Pour milk into coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle", "coffee mug", "stainless steel milk pitcher", "table", "milk bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.968721, "mean_abs": 0.239843, "relative": 0.840904 }, { "name": "hand_right_joints", "l2": 9.677197, "mean_abs": 0.262593, "relative": 0.762822 }, { "name": "body_joints", "l2": 6.421746, "mean_abs": 0.095277, "relative": 0.86738 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081675, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.028063, "mean_abs": 0.003104, "relative": 0.066224 }, { "name": "camera_rotation_matrix", "l2": 2.999536, "mean_abs": 0.203797, "relative": 0.921109 }, { "name": "imu_accel_gyro", "l2": 17.080948, "mean_abs": 1.04667, "relative": 0.921576 }, { "name": "depth_confidence", "l2": 10.420365, "mean_abs": 0.190581, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.258084, "mean_abs": 0.10893, "relative": 0.952127 }, { "name": "video_fisheye_cam1", "l2": 5.379653, "mean_abs": 0.113268, "relative": 0.888942 }, { "name": "video_fisheye_cam2", "l2": 5.529314, "mean_abs": 0.113725, "relative": 0.882673 }, { "name": "video_fisheye_cam3", "l2": 5.619283, "mean_abs": 0.115313, "relative": 0.940881 }, { "name": "video_stereo_left", "l2": 5.610614, "mean_abs": 0.121662, "relative": 0.911736 }, { "name": "video_stereo_right", "l2": 6.592054, "mean_abs": 0.144423, "relative": 0.929341 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.028754, "mean_abs": 0.078705, "relative": 0.726061 }, { "name": "caption_objects_interaction_text", "l2": 1.828376, "mean_abs": 0.017758, "relative": 0.839546 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour coffee", "predicted": "Pour liquid from white bottle", "correct": 0, "confidence": "0.38013845682144165" }, "timeline_subtask": { "true": "Pour milk into coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8854793310165405" }, "transition_detection": { "true": "transition", "predicted": "transition", "correct": 1, "confidence": "0.9999809265136719" }, "next_action": { "true": "Pour milk into coffee", "predicted": "Grasp coffee scoop", "correct": 0, "confidence": "0.3596675395965576" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|table|coffee mug|milk pitcher|coffee cup|stainless steel milk pitcher|milk bottle", "predicted": "coffee dripper|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1158, "start_frame": 5790, "end_frame": 5809, "center_frame": 5799, "action": "", "subtask": "Pour milk into coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle", "coffee mug", "stainless steel milk pitcher", "table", "milk bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.963935, "mean_abs": 0.243749, "relative": 0.8405 }, { "name": "hand_right_joints", "l2": 9.647164, "mean_abs": 0.263854, "relative": 0.760455 }, { "name": "body_joints", "l2": 6.441633, "mean_abs": 0.097127, "relative": 0.870066 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081676, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.045207, "mean_abs": 0.00552, "relative": 0.106683 }, { "name": "camera_rotation_matrix", "l2": 3.001215, "mean_abs": 0.207822, "relative": 0.921624 }, { "name": "imu_accel_gyro", "l2": 17.104153, "mean_abs": 1.06616, "relative": 0.922828 }, { "name": "depth_confidence", "l2": 10.405989, "mean_abs": 0.191959, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.250416, "mean_abs": 0.109412, "relative": 0.950738 }, { "name": "video_fisheye_cam1", "l2": 5.399474, "mean_abs": 0.115442, "relative": 0.892218 }, { "name": "video_fisheye_cam2", "l2": 5.538034, "mean_abs": 0.115238, "relative": 0.884065 }, { "name": "video_fisheye_cam3", "l2": 5.659094, "mean_abs": 0.115992, "relative": 0.947547 }, { "name": "video_stereo_left", "l2": 5.636265, "mean_abs": 0.124429, "relative": 0.915904 }, { "name": "video_stereo_right", "l2": 6.685865, "mean_abs": 0.149579, "relative": 0.942567 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.087415, "mean_abs": 0.081222, "relative": 0.747054 }, { "name": "caption_objects_interaction_text", "l2": 1.758681, "mean_abs": 0.017042, "relative": 0.807544 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": null, "timeline_subtask": { "true": "Pour milk into coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8767520189285278" }, "transition_detection": { "true": "transition", "predicted": "transition", "correct": 1, "confidence": "0.9999933242797852" }, "next_action": { "true": "Pour milk into coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.6127960681915283" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|table|coffee mug|milk pitcher|coffee cup|stainless steel milk pitcher|milk bottle", "predicted": "coffee dripper|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1159, "start_frame": 5795, "end_frame": 5814, "center_frame": 5804, "action": "Pour milk into coffee", "subtask": "Pour milk into coffee", "objects": [ "milk pitcher", "coffee cup", "digital scale", "bottle", "coffee mug", "stainless steel milk pitcher", "table", "milk bottle" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.971428, "mean_abs": 0.246807, "relative": 0.841132 }, { "name": "hand_right_joints", "l2": 9.622434, "mean_abs": 0.265105, "relative": 0.758505 }, { "name": "body_joints", "l2": 6.464365, "mean_abs": 0.098461, "relative": 0.873137 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081677, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.05063, "mean_abs": 0.006764, "relative": 0.119481 }, { "name": "camera_rotation_matrix", "l2": 3.002879, "mean_abs": 0.209539, "relative": 0.922135 }, { "name": "imu_accel_gyro", "l2": 17.105722, "mean_abs": 1.05631, "relative": 0.922913 }, { "name": "depth_confidence", "l2": 10.41189, "mean_abs": 0.192109, "relative": 0.000346 }, { "name": "video_fisheye_cam0", "l2": 5.250942, "mean_abs": 0.109649, "relative": 0.950833 }, { "name": "video_fisheye_cam1", "l2": 5.421518, "mean_abs": 0.116913, "relative": 0.89586 }, { "name": "video_fisheye_cam2", "l2": 5.541958, "mean_abs": 0.115863, "relative": 0.884691 }, { "name": "video_fisheye_cam3", "l2": 5.69235, "mean_abs": 0.116732, "relative": 0.953115 }, { "name": "video_stereo_left", "l2": 5.662914, "mean_abs": 0.126099, "relative": 0.920235 }, { "name": "video_stereo_right", "l2": 6.760176, "mean_abs": 0.152501, "relative": 0.953043 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.131392, "mean_abs": 0.082896, "relative": 0.762793 }, { "name": "caption_objects_interaction_text", "l2": 1.686108, "mean_abs": 0.015741, "relative": 0.77422 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour milk into coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.7695194482803345" }, "timeline_subtask": { "true": "Pour milk into coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.7595806121826172" }, "transition_detection": { "true": "transition", "predicted": "transition", "correct": 1, "confidence": "0.9979351758956909" }, "next_action": { "true": "Pour milk into coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.6660147309303284" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "bottle|digital scale|table|coffee mug|milk pitcher|coffee cup|stainless steel milk pitcher|milk bottle", "predicted": "coffee dripper|coffee jar|white bottle", "correct": 0, "confidence": "" } } }, { "window_index": 1160, "start_frame": 5800, "end_frame": 5819, "center_frame": 5809, "action": "Pour milk into coffee", "subtask": "Pour milk into coffee", "objects": [ "coffee mug", "stainless steel milk pitcher", "table", "milk bottle", "digital scale" ], "feature_stats": [ { "name": "hand_left_joints", "l2": 9.983866, "mean_abs": 0.248821, "relative": 0.842182 }, { "name": "hand_right_joints", "l2": 9.574771, "mean_abs": 0.265407, "relative": 0.754748 }, { "name": "body_joints", "l2": 6.497583, "mean_abs": 0.099801, "relative": 0.877623 }, { "name": "body_contacts", "l2": 3.464102, "mean_abs": 0.081678, "relative": 0.999625 }, { "name": "camera_translation", "l2": 0.046385, "mean_abs": 0.006704, "relative": 0.109462 }, { "name": "camera_rotation_matrix", "l2": 3.002163, "mean_abs": 0.209666, "relative": 0.921916 }, { "name": "imu_accel_gyro", "l2": 17.071604, "mean_abs": 1.065196, "relative": 0.921072 }, { "name": "depth_confidence", "l2": 10.408319, "mean_abs": 0.191223, "relative": 0.000345 }, { "name": "video_fisheye_cam0", "l2": 5.252712, "mean_abs": 0.109735, "relative": 0.951154 }, { "name": "video_fisheye_cam1", "l2": 5.441864, "mean_abs": 0.1172, "relative": 0.899222 }, { "name": "video_fisheye_cam2", "l2": 5.548408, "mean_abs": 0.115988, "relative": 0.885721 }, { "name": "video_fisheye_cam3", "l2": 5.712443, "mean_abs": 0.117142, "relative": 0.95648 }, { "name": "video_stereo_left", "l2": 5.682262, "mean_abs": 0.126113, "relative": 0.923379 }, { "name": "video_stereo_right", "l2": 6.826939, "mean_abs": 0.153528, "relative": 0.962455 }, { "name": "audio_fisheye_cam0_aac", "l2": 2.207377, "mean_abs": 0.086336, "relative": 0.789987 }, { "name": "caption_objects_interaction_text", "l2": 1.450045, "mean_abs": 0.010417, "relative": 0.665826 }, { "name": "slam_point_cloud", "l2": 12.724784, "mean_abs": 1.984679, "relative": 1.0 }, { "name": "calibration", "l2": 2087.725098, "mean_abs": 76.022057, "relative": 1.0 } ], "predictions": { "timeline_action": { "true": "Pour milk into coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.7730115056037903" }, "timeline_subtask": { "true": "Pour milk into coffee", "predicted": "Pour and close white bottle", "correct": 0, "confidence": "0.8281679153442383" }, "transition_detection": { "true": "transition", "predicted": "transition", "correct": 1, "confidence": "0.7152870297431946" }, "next_action": { "true": "Pour milk into coffee", "predicted": "Position kettle to pour", "correct": 0, "confidence": "0.6966760754585266" }, "contact_prediction": { "true": "contact", "predicted": "contact", "correct": 1, "confidence": "1.0" }, "object_relevance": { "true": "digital scale|table|coffee mug|stainless steel milk pitcher|milk bottle", "predicted": "coffee dripper|coffee jar|white bottle", "correct": 0, "confidence": "" } } } ], "ablation": { "best_by_task": { "caption_grounding": { "task": "caption_grounding", "task_display_name": "Language Grounding", "best": { "modality_group": "language", "modality_display": "Language", "score": 0.24527303874492645, "primary_metric": "mrr", "target_source_overlap": "true" }, "best_non_overlap": { "modality_group": "audio", "modality_display": "Audio", "score": 0.03902389109134674, "primary_metric": "mrr" } }, "contact_prediction": { "task": "contact_prediction", "task_display_name": "Contact State Prediction", "best": { "modality_group": "all_features", "modality_display": "All Features", "score": 1.0, "primary_metric": "macro_f1", "target_source_overlap": "false" }, "best_non_overlap": { "modality_group": "all_features", "modality_display": "All Features", "score": 1.0, "primary_metric": "macro_f1" } }, "cross_modal_retrieval": { "task": "cross_modal_retrieval", "task_display_name": "Cross-Modal Retrieval", "best": { "modality_group": "all_features", "modality_display": "All Features", "score": 0.972757875919342, "primary_metric": "mrr", "target_source_overlap": "true" }, "best_non_overlap": { "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "score": 0.42622581124305725, "primary_metric": "mrr" } }, "hand_trajectory_forecast": { "task": "hand_trajectory_forecast", "task_display_name": "Hand Trajectory Forecasting", "best": { "modality_group": "inertial", "modality_display": "Inertial", "score": 0.5679183061202404, "primary_metric": "mae", "target_source_overlap": "false" }, "best_non_overlap": { "modality_group": "inertial", "modality_display": "Inertial", "score": 0.5679183061202404, "primary_metric": "mae" } }, "misalignment_detection": { "task": "misalignment_detection", "task_display_name": "Multimodal Synchronization Detection", "best": { "modality_group": "audio", "modality_display": "Audio", "score": 0.5152624305034046, "primary_metric": "macro_f1", "target_source_overlap": "false" }, "best_non_overlap": { "modality_group": "audio", "modality_display": "Audio", "score": 0.5152624305034046, "primary_metric": "macro_f1" } }, "modality_reconstruction": { "task": "modality_reconstruction", "task_display_name": "Cross-Modal Reconstruction", "best": { "modality_group": "video", "modality_display": "Video", "score": 0.611318891594774, "primary_metric": "mae", "target_source_overlap": "true" }, "best_non_overlap": { "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "score": 0.5359235021455191, "primary_metric": "mae" } }, "next_action": { "task": "next_action", "task_display_name": "Next-Action Prediction", "best": { "modality_group": "language", "modality_display": "Language", "score": 0.04193971166448231, "primary_metric": "macro_f1", "target_source_overlap": "false" }, "best_non_overlap": { "modality_group": "language", "modality_display": "Language", "score": 0.04193971166448231, "primary_metric": "macro_f1" } }, "object_relevance": { "task": "object_relevance", "task_display_name": "Object Relevance Prediction", "best": { "modality_group": "language", "modality_display": "Language", "score": 0.23021032504780117, "primary_metric": "micro_f1", "target_source_overlap": "true" }, "best_non_overlap": { "modality_group": "depth", "modality_display": "Depth", "score": 0.20134228187919462, "primary_metric": "micro_f1" } }, "temporal_order": { "task": "temporal_order", "task_display_name": "Temporal Order Verification", "best": { "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "score": 0.5258620689655172, "primary_metric": "macro_f1", "target_source_overlap": "false" }, "best_non_overlap": { "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "score": 0.5258620689655172, "primary_metric": "macro_f1" } }, "timeline_action": { "task": "timeline_action", "task_display_name": "Action Recognition", "best": { "modality_group": "language", "modality_display": "Language", "score": 0.027777777777777776, "primary_metric": "macro_f1", "target_source_overlap": "false" }, "best_non_overlap": { "modality_group": "language", "modality_display": "Language", "score": 0.027777777777777776, "primary_metric": "macro_f1" } }, "timeline_subtask": { "task": "timeline_subtask", "task_display_name": "Procedure Step Recognition", "best": { "modality_group": "language", "modality_display": "Language", "score": 0.04828150572831424, "primary_metric": "macro_f1", "target_source_overlap": "false" }, "best_non_overlap": { "modality_group": "language", "modality_display": "Language", "score": 0.04828150572831424, "primary_metric": "macro_f1" } }, "transition_detection": { "task": "transition_detection", "task_display_name": "Action Boundary Detection", "best": { "modality_group": "language", "modality_display": "Language", "score": 0.7051957831325302, "primary_metric": "macro_f1", "target_source_overlap": "false" }, "best_non_overlap": { "modality_group": "language", "modality_display": "Language", "score": 0.7051957831325302, "primary_metric": "macro_f1" } } }, "rows": [ { "task": "timeline_action", "task_display": "Current Action Recognition", "modality_group": "all_features", "modality_display": "All Features", "status": "computed", "score": "0.00905456968081885", "primary_metric": "macro_f1", "primary_metric_value": "0.00905456968081885", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.017241379310344827", "macro_f1": "0.00905456968081885", "balanced_accuracy": "0.01720647773279352", "num_classes": "19", "num_train": "813", "num_test": "348", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Recognition" }, { "task": "timeline_action", "task_display": "Current Action Recognition", "modality_group": "video", "modality_display": "Video", "status": "computed", "score": "0.0066280033140016575", "primary_metric": "macro_f1", "primary_metric_value": "0.0066280033140016575", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.011494252873563218", "macro_f1": "0.0066280033140016575", "balanced_accuracy": "0.0036199095022624436", "num_classes": "19", "num_train": "813", "num_test": "348", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Recognition" }, { "task": "timeline_action", "task_display": "Current Action Recognition", "modality_group": "depth", "modality_display": "Depth", "status": "computed", "score": "0.0030075187969924814", "primary_metric": "macro_f1", "primary_metric_value": "0.0030075187969924814", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.005747126436781609", "macro_f1": "0.0030075187969924814", "balanced_accuracy": "0.001619433198380567", "num_classes": "19", "num_train": "813", "num_test": "348", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Recognition" }, { "task": "timeline_action", "task_display": "Current Action Recognition", "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "status": "computed", "score": "0.0", "primary_metric": "macro_f1", "primary_metric_value": "0.0", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.0", "macro_f1": "0.0", "balanced_accuracy": "0.0", "num_classes": "19", "num_train": "813", "num_test": "348", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Recognition" }, { "task": "timeline_action", "task_display": "Current Action Recognition", "modality_group": "motion_capture", "modality_display": "Motion Capture", "status": "computed", "score": "0.0055147058823529415", "primary_metric": "macro_f1", "primary_metric_value": "0.0055147058823529415", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.008620689655172414", "macro_f1": "0.0055147058823529415", "balanced_accuracy": "0.0028846153846153848", "num_classes": "19", "num_train": "813", "num_test": "348", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Recognition" }, { "task": "timeline_action", "task_display": "Current Action Recognition", "modality_group": "inertial", "modality_display": "Inertial", "status": "computed", "score": "0.003055767761650115", "primary_metric": "macro_f1", "primary_metric_value": "0.003055767761650115", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.005747126436781609", "macro_f1": "0.003055767761650115", "balanced_accuracy": "0.0018099547511312218", "num_classes": "19", "num_train": "813", "num_test": "348", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Recognition" }, { "task": "timeline_action", "task_display": "Current Action Recognition", "modality_group": "audio", "modality_display": "Audio", "status": "computed", "score": "0.006925207756232688", "primary_metric": "macro_f1", "primary_metric_value": "0.006925207756232688", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.014367816091954023", "macro_f1": "0.006925207756232688", "balanced_accuracy": "0.004048582995951417", "num_classes": "19", "num_train": "813", "num_test": "348", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Recognition" }, { "task": "timeline_action", "task_display": "Current Action Recognition", "modality_group": "language", "modality_display": "Language", "status": "computed", "score": "0.027777777777777776", "primary_metric": "macro_f1", "primary_metric_value": "0.027777777777777776", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.05747126436781609", "macro_f1": "0.027777777777777776", "balanced_accuracy": "0.03615384615384616", "num_classes": "19", "num_train": "813", "num_test": "348", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Recognition" }, { "task": "timeline_action", "task_display": "Current Action Recognition", "modality_group": "no_language", "modality_display": "All Except Language", "status": "computed", "score": "0.0043859649122807015", "primary_metric": "macro_f1", "primary_metric_value": "0.0043859649122807015", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.008620689655172414", "macro_f1": "0.0043859649122807015", "balanced_accuracy": "0.0024291497975708503", "num_classes": "19", "num_train": "813", "num_test": "348", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Recognition" }, { "task": "timeline_subtask", "task_display": "Current Subtask Recognition", "modality_group": "all_features", "modality_display": "All Features", "status": "computed", "score": "0.011256354393609296", "primary_metric": "macro_f1", "primary_metric_value": "0.011256354393609296", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.02586206896551724", "macro_f1": "0.011256354393609296", "balanced_accuracy": "0.02788220551378446", "num_classes": "15", "num_train": "813", "num_test": "348", "unseen_test_classes": "Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Procedure Step Recognition" }, { "task": "timeline_subtask", "task_display": "Current Subtask Recognition", "modality_group": "video", "modality_display": "Video", "status": "computed", "score": "0.011740041928721174", "primary_metric": "macro_f1", "primary_metric_value": "0.011740041928721174", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.040229885057471264", "macro_f1": "0.011740041928721174", "balanced_accuracy": "0.01637426900584795", "num_classes": "15", "num_train": "813", "num_test": "348", "unseen_test_classes": "Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Procedure Step Recognition" }, { "task": "timeline_subtask", "task_display": "Current Subtask Recognition", "modality_group": "depth", "modality_display": "Depth", "status": "computed", "score": "0.009467455621301775", "primary_metric": "macro_f1", "primary_metric_value": "0.009467455621301775", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.022988505747126436", "macro_f1": "0.009467455621301775", "balanced_accuracy": "0.010796221322537112", "num_classes": "15", "num_train": "813", "num_test": "348", "unseen_test_classes": "Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Procedure Step Recognition" }, { "task": "timeline_subtask", "task_display": "Current Subtask Recognition", "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "status": "computed", "score": "0.002331002331002331", "primary_metric": "macro_f1", "primary_metric_value": "0.002331002331002331", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.0028735632183908046", "macro_f1": "0.002331002331002331", "balanced_accuracy": "0.001349527665317139", "num_classes": "15", "num_train": "813", "num_test": "348", "unseen_test_classes": "Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Procedure Step Recognition" }, { "task": "timeline_subtask", "task_display": "Current Subtask Recognition", "modality_group": "motion_capture", "modality_display": "Motion Capture", "status": "computed", "score": "0.006756756756756756", "primary_metric": "macro_f1", "primary_metric_value": "0.006756756756756756", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.008620689655172414", "macro_f1": "0.006756756756756756", "balanced_accuracy": "0.0043859649122807015", "num_classes": "15", "num_train": "813", "num_test": "348", "unseen_test_classes": "Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Procedure Step Recognition" }, { "task": "timeline_subtask", "task_display": "Current Subtask Recognition", "modality_group": "inertial", "modality_display": "Inertial", "status": "computed", "score": "0.004662004662004662", "primary_metric": "macro_f1", "primary_metric_value": "0.004662004662004662", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.005747126436781609", "macro_f1": "0.004662004662004662", "balanced_accuracy": "0.002699055330634278", "num_classes": "15", "num_train": "813", "num_test": "348", "unseen_test_classes": "Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Procedure Step Recognition" }, { "task": "timeline_subtask", "task_display": "Current Subtask Recognition", "modality_group": "audio", "modality_display": "Audio", "status": "computed", "score": "0.016194331983805668", "primary_metric": "macro_f1", "primary_metric_value": "0.016194331983805668", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.022988505747126436", "macro_f1": "0.016194331983805668", "balanced_accuracy": "0.010796221322537112", "num_classes": "15", "num_train": "813", "num_test": "348", "unseen_test_classes": "Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Procedure Step Recognition" }, { "task": "timeline_subtask", "task_display": "Current Subtask Recognition", "modality_group": "language", "modality_display": "Language", "status": "computed", "score": "0.04828150572831424", "primary_metric": "macro_f1", "primary_metric_value": "0.04828150572831424", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.14655172413793102", "macro_f1": "0.04828150572831424", "balanced_accuracy": "0.0939327485380117", "num_classes": "15", "num_train": "813", "num_test": "348", "unseen_test_classes": "Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Procedure Step Recognition" }, { "task": "timeline_subtask", "task_display": "Current Subtask Recognition", "modality_group": "no_language", "modality_display": "All Except Language", "status": "computed", "score": "0.007427055702917773", "primary_metric": "macro_f1", "primary_metric_value": "0.007427055702917773", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.020114942528735632", "macro_f1": "0.007427055702917773", "balanced_accuracy": "0.009446693657219972", "num_classes": "15", "num_train": "813", "num_test": "348", "unseen_test_classes": "Move bottle to coffee equipment|Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Procedure Step Recognition" }, { "task": "transition_detection", "task_display": "Action Transition Detection", "modality_group": "all_features", "modality_display": "All Features", "status": "computed", "score": "0.46213292117465227", "primary_metric": "macro_f1", "primary_metric_value": "0.46213292117465227", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.8591954022988506", "macro_f1": "0.46213292117465227", "balanced_accuracy": "0.4503012048192771", "num_classes": "2", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Boundary Detection" }, { "task": "transition_detection", "task_display": "Action Transition Detection", "modality_group": "video", "modality_display": "Video", "status": "computed", "score": "0.46625766871165636", "primary_metric": "macro_f1", "primary_metric_value": "0.46625766871165636", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.8735632183908046", "macro_f1": "0.46625766871165636", "balanced_accuracy": "0.4578313253012048", "num_classes": "2", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Boundary Detection" }, { "task": "transition_detection", "task_display": "Action Transition Detection", "modality_group": "depth", "modality_display": "Depth", "status": "computed", "score": "0.4604651162790698", "primary_metric": "macro_f1", "primary_metric_value": "0.4604651162790698", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.853448275862069", "macro_f1": "0.4604651162790698", "balanced_accuracy": "0.44728915662650603", "num_classes": "2", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Boundary Detection" }, { "task": "transition_detection", "task_display": "Action Transition Detection", "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "status": "computed", "score": "0.48444444444444446", "primary_metric": "macro_f1", "primary_metric_value": "0.48444444444444446", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.9396551724137931", "macro_f1": "0.48444444444444446", "balanced_accuracy": "0.4924698795180723", "num_classes": "2", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Boundary Detection" }, { "task": "transition_detection", "task_display": "Action Transition Detection", "modality_group": "motion_capture", "modality_display": "Motion Capture", "status": "computed", "score": "0.5439056356487549", "primary_metric": "macro_f1", "primary_metric_value": "0.5439056356487549", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.896551724137931", "macro_f1": "0.5439056356487549", "balanced_accuracy": "0.5591114457831325", "num_classes": "2", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Boundary Detection" }, { "task": "transition_detection", "task_display": "Action Transition Detection", "modality_group": "inertial", "modality_display": "Inertial", "status": "computed", "score": "0.48520710059171596", "primary_metric": "macro_f1", "primary_metric_value": "0.48520710059171596", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.9425287356321839", "macro_f1": "0.48520710059171596", "balanced_accuracy": "0.4939759036144578", "num_classes": "2", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Boundary Detection" }, { "task": "transition_detection", "task_display": "Action Transition Detection", "modality_group": "audio", "modality_display": "Audio", "status": "computed", "score": "0.48444444444444446", "primary_metric": "macro_f1", "primary_metric_value": "0.48444444444444446", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.9396551724137931", "macro_f1": "0.48444444444444446", "balanced_accuracy": "0.4924698795180723", "num_classes": "2", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Boundary Detection" }, { "task": "transition_detection", "task_display": "Action Transition Detection", "modality_group": "language", "modality_display": "Language", "status": "computed", "score": "0.7051957831325302", "primary_metric": "macro_f1", "primary_metric_value": "0.7051957831325302", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.9482758620689655", "macro_f1": "0.7051957831325302", "balanced_accuracy": "0.7051957831325302", "num_classes": "2", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Boundary Detection" }, { "task": "transition_detection", "task_display": "Action Transition Detection", "modality_group": "no_language", "modality_display": "All Except Language", "status": "computed", "score": "0.45454545454545453", "primary_metric": "macro_f1", "primary_metric_value": "0.45454545454545453", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.8333333333333334", "macro_f1": "0.45454545454545453", "balanced_accuracy": "0.4367469879518072", "num_classes": "2", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Action Boundary Detection" }, { "task": "next_action", "task_display": "Next-Action Prediction", "modality_group": "all_features", "modality_display": "All Features", "status": "computed", "score": "0.004971002485501243", "primary_metric": "macro_f1", "primary_metric_value": "0.004971002485501243", "target_variant": "future action label from windows.csv", "target_source_overlap": "false", "reason": "", "accuracy": "0.008645533141210375", "macro_f1": "0.004971002485501243", "balanced_accuracy": "0.0027573529411764708", "num_classes": "19", "num_train": "810", "num_test": "347", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Next-Action Prediction" }, { "task": "next_action", "task_display": "Next-Action Prediction", "modality_group": "video", "modality_display": "Video", "status": "computed", "score": "0.006349206349206349", "primary_metric": "macro_f1", "primary_metric_value": "0.006349206349206349", "target_variant": "future action label from windows.csv", "target_source_overlap": "false", "reason": "", "accuracy": "0.011527377521613832", "macro_f1": "0.006349206349206349", "balanced_accuracy": "0.003472222222222222", "num_classes": "19", "num_train": "810", "num_test": "347", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Next-Action Prediction" }, { "task": "next_action", "task_display": "Next-Action Prediction", "modality_group": "depth", "modality_display": "Depth", "status": "computed", "score": "0.001594896331738437", "primary_metric": "macro_f1", "primary_metric_value": "0.001594896331738437", "target_variant": "future action label from windows.csv", "target_source_overlap": "false", "reason": "", "accuracy": "0.002881844380403458", "macro_f1": "0.001594896331738437", "balanced_accuracy": "0.0008223684210526315", "num_classes": "19", "num_train": "810", "num_test": "347", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Next-Action Prediction" }, { "task": "next_action", "task_display": "Next-Action Prediction", "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "status": "computed", "score": "0.0", "primary_metric": "macro_f1", "primary_metric_value": "0.0", "target_variant": "future action label from windows.csv", "target_source_overlap": "false", "reason": "", "accuracy": "0.0", "macro_f1": "0.0", "balanced_accuracy": "0.0", "num_classes": "19", "num_train": "810", "num_test": "347", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Next-Action Prediction" }, { "task": "next_action", "task_display": "Next-Action Prediction", "modality_group": "motion_capture", "modality_display": "Motion Capture", "status": "computed", "score": "0.00322061191626409", "primary_metric": "macro_f1", "primary_metric_value": "0.00322061191626409", "target_variant": "future action label from windows.csv", "target_source_overlap": "false", "reason": "", "accuracy": "0.005763688760806916", "macro_f1": "0.00322061191626409", "balanced_accuracy": "0.001736111111111111", "num_classes": "19", "num_train": "810", "num_test": "347", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Next-Action Prediction" }, { "task": "next_action", "task_display": "Next-Action Prediction", "modality_group": "inertial", "modality_display": "Inertial", "status": "computed", "score": "0.00196078431372549", "primary_metric": "macro_f1", "primary_metric_value": "0.00196078431372549", "target_variant": "future action label from windows.csv", "target_source_overlap": "false", "reason": "", "accuracy": "0.002881844380403458", "macro_f1": "0.00196078431372549", "balanced_accuracy": "0.0010416666666666667", "num_classes": "19", "num_train": "810", "num_test": "347", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Next-Action Prediction" }, { "task": "next_action", "task_display": "Next-Action Prediction", "modality_group": "audio", "modality_display": "Audio", "status": "computed", "score": "0.004761904761904762", "primary_metric": "macro_f1", "primary_metric_value": "0.004761904761904762", "target_variant": "future action label from windows.csv", "target_source_overlap": "false", "reason": "", "accuracy": "0.008645533141210375", "macro_f1": "0.004761904761904762", "balanced_accuracy": "0.0026041666666666665", "num_classes": "19", "num_train": "810", "num_test": "347", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Next-Action Prediction" }, { "task": "next_action", "task_display": "Next-Action Prediction", "modality_group": "language", "modality_display": "Language", "status": "computed", "score": "0.04193971166448231", "primary_metric": "macro_f1", "primary_metric_value": "0.04193971166448231", "target_variant": "future action label from windows.csv", "target_source_overlap": "false", "reason": "", "accuracy": "0.1844380403458213", "macro_f1": "0.04193971166448231", "balanced_accuracy": "0.07142857142857142", "num_classes": "19", "num_train": "810", "num_test": "347", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Next-Action Prediction" }, { "task": "next_action", "task_display": "Next-Action Prediction", "modality_group": "no_language", "modality_display": "All Except Language", "status": "computed", "score": "0.004576659038901602", "primary_metric": "macro_f1", "primary_metric_value": "0.004576659038901602", "target_variant": "future action label from windows.csv", "target_source_overlap": "false", "reason": "", "accuracy": "0.008645533141210375", "macro_f1": "0.004576659038901602", "balanced_accuracy": "0.0024671052631578946", "num_classes": "19", "num_train": "810", "num_test": "347", "unseen_test_classes": "Place item on table|Wait/Prepare for pouring|Pour coffee|Pour milk into coffee", "unseen_test_class_count": "4", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Next-Action Prediction" }, { "task": "hand_trajectory_forecast", "task_display": "Future Hand Motion Forecasting", "modality_group": "all_features", "modality_display": "All Features", "status": "computed", "score": "0.11919872659288855", "primary_metric": "mae", "primary_metric_value": "7.3893513679504395", "target_variant": "future hand feature vector from shared_windows.npz", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "4377.7705078125", "mae": "7.3893513679504395", "r2": "-4562.174276179636", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Hand Trajectory Forecasting" }, { "task": "hand_trajectory_forecast", "task_display": "Future Hand Motion Forecasting", "modality_group": "video", "modality_display": "Video", "status": "computed", "score": "0.4956350584748486", "primary_metric": "mae", "primary_metric_value": "1.0176135301589966", "target_variant": "future hand feature vector from shared_windows.npz", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "1.7896661758422852", "mae": "1.0176135301589966", "r2": "-0.8654605965108897", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Hand Trajectory Forecasting" }, { "task": "hand_trajectory_forecast", "task_display": "Future Hand Motion Forecasting", "modality_group": "depth", "modality_display": "Depth", "status": "computed", "score": "0.04014931629731973", "primary_metric": "mae", "primary_metric_value": "23.907024383544922", "target_variant": "future hand feature vector from shared_windows.npz", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "72553.34375", "mae": "23.907024383544922", "r2": "-75625.0610993949", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Hand Trajectory Forecasting" }, { "task": "hand_trajectory_forecast", "task_display": "Future Hand Motion Forecasting", "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "status": "computed", "score": "0.5611809661721311", "primary_metric": "mae", "primary_metric_value": "0.7819563746452332", "target_variant": "future hand feature vector from shared_windows.npz", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "1.2600995302200317", "mae": "0.7819563746452332", "r2": "-0.3134661692106211", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Hand Trajectory Forecasting" }, { "task": "hand_trajectory_forecast", "task_display": "Future Hand Motion Forecasting", "modality_group": "motion_capture", "modality_display": "Motion Capture", "status": "computed", "score": "0.0839705207556719", "primary_metric": "mae", "primary_metric_value": "10.908941268920898", "target_variant": "future hand feature vector from shared_windows.npz", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "6293.8876953125", "mae": "10.908941268920898", "r2": "-6559.441194341517", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Hand Trajectory Forecasting" }, { "task": "hand_trajectory_forecast", "task_display": "Future Hand Motion Forecasting", "modality_group": "inertial", "modality_display": "Inertial", "status": "computed", "score": "0.5679183061202404", "primary_metric": "mae", "primary_metric_value": "0.7608166337013245", "target_variant": "future hand feature vector from shared_windows.npz", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "1.1916581392288208", "mae": "0.7608166337013245", "r2": "-0.24212624676650907", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Hand Trajectory Forecasting" }, { "task": "hand_trajectory_forecast", "task_display": "Future Hand Motion Forecasting", "modality_group": "audio", "modality_display": "Audio", "status": "computed", "score": "0.4677426495622507", "primary_metric": "mae", "primary_metric_value": "1.137927770614624", "target_variant": "future hand feature vector from shared_windows.npz", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "3.675895929336548", "mae": "1.137927770614624", "r2": "-2.831574197208222", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Hand Trajectory Forecasting" }, { "task": "hand_trajectory_forecast", "task_display": "Future Hand Motion Forecasting", "modality_group": "language", "modality_display": "Language", "status": "computed", "score": "0.451525705011023", "primary_metric": "mae", "primary_metric_value": "1.2147133350372314", "target_variant": "future hand feature vector from shared_windows.npz", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "2.3450045585632324", "mae": "1.2147133350372314", "r2": "-1.4443180759924243", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Hand Trajectory Forecasting" }, { "task": "hand_trajectory_forecast", "task_display": "Future Hand Motion Forecasting", "modality_group": "no_language", "modality_display": "All Except Language", "status": "computed", "score": "0.10781931019562568", "primary_metric": "mae", "primary_metric_value": "8.274776458740234", "target_variant": "future hand feature vector from shared_windows.npz", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "5327.0849609375", "mae": "8.274776458740234", "r2": "-5551.693117832905", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Hand Trajectory Forecasting" }, { "task": "contact_prediction", "task_display": "Contact State Prediction", "modality_group": "all_features", "modality_display": "All Features", "status": "computed", "score": "1.0", "primary_metric": "macro_f1", "primary_metric_value": "1.0", "target_variant": "contact proxy derived from body_contacts feature block", "target_source_overlap": "false", "reason": "", "accuracy": "1.0", "macro_f1": "1.0", "balanced_accuracy": "1.0", "num_classes": "1", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Contact State Prediction" }, { "task": "contact_prediction", "task_display": "Contact State Prediction", "modality_group": "video", "modality_display": "Video", "status": "computed", "score": "1.0", "primary_metric": "macro_f1", "primary_metric_value": "1.0", "target_variant": "contact proxy derived from body_contacts feature block", "target_source_overlap": "false", "reason": "", "accuracy": "1.0", "macro_f1": "1.0", "balanced_accuracy": "1.0", "num_classes": "1", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Contact State Prediction" }, { "task": "contact_prediction", "task_display": "Contact State Prediction", "modality_group": "depth", "modality_display": "Depth", "status": "computed", "score": "1.0", "primary_metric": "macro_f1", "primary_metric_value": "1.0", "target_variant": "contact proxy derived from body_contacts feature block", "target_source_overlap": "false", "reason": "", "accuracy": "1.0", "macro_f1": "1.0", "balanced_accuracy": "1.0", "num_classes": "1", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Contact State Prediction" }, { "task": "contact_prediction", "task_display": "Contact State Prediction", "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "status": "computed", "score": "1.0", "primary_metric": "macro_f1", "primary_metric_value": "1.0", "target_variant": "contact proxy derived from body_contacts feature block", "target_source_overlap": "false", "reason": "", "accuracy": "1.0", "macro_f1": "1.0", "balanced_accuracy": "1.0", "num_classes": "1", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Contact State Prediction" }, { "task": "contact_prediction", "task_display": "Contact State Prediction", "modality_group": "motion_capture", "modality_display": "Motion Capture", "status": "computed", "score": "1.0", "primary_metric": "macro_f1", "primary_metric_value": "1.0", "target_variant": "contact proxy derived from body_contacts feature block", "target_source_overlap": "false", "reason": "", "accuracy": "1.0", "macro_f1": "1.0", "balanced_accuracy": "1.0", "num_classes": "1", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Contact State Prediction" }, { "task": "contact_prediction", "task_display": "Contact State Prediction", "modality_group": "inertial", "modality_display": "Inertial", "status": "computed", "score": "1.0", "primary_metric": "macro_f1", "primary_metric_value": "1.0", "target_variant": "contact proxy derived from body_contacts feature block", "target_source_overlap": "false", "reason": "", "accuracy": "1.0", "macro_f1": "1.0", "balanced_accuracy": "1.0", "num_classes": "1", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Contact State Prediction" }, { "task": "contact_prediction", "task_display": "Contact State Prediction", "modality_group": "audio", "modality_display": "Audio", "status": "computed", "score": "1.0", "primary_metric": "macro_f1", "primary_metric_value": "1.0", "target_variant": "contact proxy derived from body_contacts feature block", "target_source_overlap": "false", "reason": "", "accuracy": "1.0", "macro_f1": "1.0", "balanced_accuracy": "1.0", "num_classes": "1", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Contact State Prediction" }, { "task": "contact_prediction", "task_display": "Contact State Prediction", "modality_group": "language", "modality_display": "Language", "status": "computed", "score": "1.0", "primary_metric": "macro_f1", "primary_metric_value": "1.0", "target_variant": "contact proxy derived from body_contacts feature block", "target_source_overlap": "false", "reason": "", "accuracy": "1.0", "macro_f1": "1.0", "balanced_accuracy": "1.0", "num_classes": "1", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Contact State Prediction" }, { "task": "contact_prediction", "task_display": "Contact State Prediction", "modality_group": "no_language", "modality_display": "All Except Language", "status": "computed", "score": "1.0", "primary_metric": "macro_f1", "primary_metric_value": "1.0", "target_variant": "contact proxy derived from body_contacts feature block", "target_source_overlap": "false", "reason": "", "accuracy": "1.0", "macro_f1": "1.0", "balanced_accuracy": "1.0", "num_classes": "1", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Contact State Prediction" }, { "task": "object_relevance", "task_display": "Relevant Object Prediction", "modality_group": "all_features", "modality_display": "All Features", "status": "computed", "score": "0.1830870279146141", "primary_metric": "micro_f1", "primary_metric_value": "0.1830870279146141", "target_variant": "object sets exported from annotation.hdf5 caption_frame_info_map", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "0.06426101593989779", "balanced_accuracy": "", "num_classes": "", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "0.1830870279146141", "exact_match": "0.031609195402298854", "precision": "0.20667284522706209", "recall": "0.1643330876934414", "num_objects": "34", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Object Relevance Prediction" }, { "task": "object_relevance", "task_display": "Relevant Object Prediction", "modality_group": "video", "modality_display": "Video", "status": "computed", "score": "0.14804270462633454", "primary_metric": "micro_f1", "primary_metric_value": "0.14804270462633454", "target_variant": "object sets exported from annotation.hdf5 caption_frame_info_map", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "0.04379950367755125", "balanced_accuracy": "", "num_classes": "", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "0.14804270462633454", "exact_match": "0.008620689655172414", "precision": "0.14315209910529939", "recall": "0.15327929255711129", "num_objects": "34", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Object Relevance Prediction" }, { "task": "object_relevance", "task_display": "Relevant Object Prediction", "modality_group": "depth", "modality_display": "Depth", "status": "computed", "score": "0.20134228187919462", "primary_metric": "micro_f1", "primary_metric_value": "0.20134228187919462", "target_variant": "object sets exported from annotation.hdf5 caption_frame_info_map", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "0.0649677953734521", "balanced_accuracy": "", "num_classes": "", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "0.20134228187919462", "exact_match": "0.011494252873563218", "precision": "0.18484288354898337", "recall": "0.2210759027266028", "num_objects": "34", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Object Relevance Prediction" }, { "task": "object_relevance", "task_display": "Relevant Object Prediction", "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "status": "computed", "score": "0.19528071602929212", "primary_metric": "micro_f1", "primary_metric_value": "0.19528071602929212", "target_variant": "object sets exported from annotation.hdf5 caption_frame_info_map", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "0.05592381693865655", "balanced_accuracy": "", "num_classes": "", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "0.19528071602929212", "exact_match": "0.0", "precision": "0.21798365122615804", "recall": "0.17686072218128224", "num_objects": "34", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Object Relevance Prediction" }, { "task": "object_relevance", "task_display": "Relevant Object Prediction", "modality_group": "motion_capture", "modality_display": "Motion Capture", "status": "computed", "score": "0.11607786589762077", "primary_metric": "micro_f1", "primary_metric_value": "0.11607786589762077", "target_variant": "object sets exported from annotation.hdf5 caption_frame_info_map", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "0.045395437036303915", "balanced_accuracy": "", "num_classes": "", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "0.11607786589762077", "exact_match": "0.0028735632183908046", "precision": "0.11362032462949895", "recall": "0.11864406779661017", "num_objects": "34", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Object Relevance Prediction" }, { "task": "object_relevance", "task_display": "Relevant Object Prediction", "modality_group": "inertial", "modality_display": "Inertial", "status": "computed", "score": "0.1716082659478886", "primary_metric": "micro_f1", "primary_metric_value": "0.1716082659478886", "target_variant": "object sets exported from annotation.hdf5 caption_frame_info_map", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "0.04806995854957751", "balanced_accuracy": "", "num_classes": "", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "0.1716082659478886", "exact_match": "0.0", "precision": "0.21979286536248563", "recall": "0.14075165806927045", "num_objects": "34", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Object Relevance Prediction" }, { "task": "object_relevance", "task_display": "Relevant Object Prediction", "modality_group": "audio", "modality_display": "Audio", "status": "computed", "score": "0.15894039735099336", "primary_metric": "micro_f1", "primary_metric_value": "0.15894039735099336", "target_variant": "object sets exported from annotation.hdf5 caption_frame_info_map", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "0.0640376063191102", "balanced_accuracy": "", "num_classes": "", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "0.15894039735099336", "exact_match": "0.005747126436781609", "precision": "0.16859504132231404", "recall": "0.1503316138540899", "num_objects": "34", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Object Relevance Prediction" }, { "task": "object_relevance", "task_display": "Relevant Object Prediction", "modality_group": "language", "modality_display": "Language", "status": "computed", "score": "0.23021032504780117", "primary_metric": "micro_f1", "primary_metric_value": "0.23021032504780117", "target_variant": "object sets exported from annotation.hdf5 caption_frame_info_map", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "0.0947530205484707", "balanced_accuracy": "", "num_classes": "", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "0.23021032504780117", "exact_match": "0.15229885057471265", "precision": "0.23926868044515104", "recall": "0.22181282240235814", "num_objects": "34", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Object Relevance Prediction" }, { "task": "object_relevance", "task_display": "Relevant Object Prediction", "modality_group": "no_language", "modality_display": "All Except Language", "status": "computed", "score": "0.15813953488372093", "primary_metric": "micro_f1", "primary_metric_value": "0.15813953488372093", "target_variant": "object sets exported from annotation.hdf5 caption_frame_info_map", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "0.05335536055344564", "balanced_accuracy": "", "num_classes": "", "num_train": "813", "num_test": "348", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "0.15813953488372093", "exact_match": "0.011494252873563218", "precision": "0.15368567454798332", "recall": "0.16285924834193072", "num_objects": "34", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Object Relevance Prediction" }, { "task": "caption_grounding", "task_display": "Language-to-Time Grounding", "modality_group": "all_features", "modality_display": "All Features", "status": "computed", "score": "0.20975713431835175", "primary_metric": "mrr", "primary_metric_value": "0.20975713431835175", "target_variant": "", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.20975713431835175", "top1_accuracy": "0.08908045977011494", "top5_accuracy": "0.32471264367816094", "top10_accuracy": "0.4454022988505747", "median_rank": "13.0", "mean_rank": "23.19827651977539", "num_queries": "348", "task_display_name": "Language Grounding" }, { "task": "caption_grounding", "task_display": "Language-to-Time Grounding", "modality_group": "video", "modality_display": "Video", "status": "computed", "score": "0.022670436650514603", "primary_metric": "mrr", "primary_metric_value": "0.022670436650514603", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.022670436650514603", "top1_accuracy": "0.0028735632183908046", "top5_accuracy": "0.02586206896551724", "top10_accuracy": "0.034482758620689655", "median_rank": "162.0", "mean_rank": "161.4770050048828", "num_queries": "348", "task_display_name": "Language Grounding" }, { "task": "caption_grounding", "task_display": "Language-to-Time Grounding", "modality_group": "depth", "modality_display": "Depth", "status": "computed", "score": "0.02443847246468067", "primary_metric": "mrr", "primary_metric_value": "0.02443847246468067", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.02443847246468067", "top1_accuracy": "0.0028735632183908046", "top5_accuracy": "0.020114942528735632", "top10_accuracy": "0.03735632183908046", "median_rank": "114.0", "mean_rank": "137.90805053710938", "num_queries": "348", "task_display_name": "Language Grounding" }, { "task": "caption_grounding", "task_display": "Language-to-Time Grounding", "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "status": "computed", "score": "0.02946249581873417", "primary_metric": "mrr", "primary_metric_value": "0.02946249581873417", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.02946249581873417", "top1_accuracy": "0.008620689655172414", "top5_accuracy": "0.028735632183908046", "top10_accuracy": "0.04597701149425287", "median_rank": "143.5", "mean_rank": "155.4712677001953", "num_queries": "348", "task_display_name": "Language Grounding" }, { "task": "caption_grounding", "task_display": "Language-to-Time Grounding", "modality_group": "motion_capture", "modality_display": "Motion Capture", "status": "computed", "score": "0.030569594353437424", "primary_metric": "mrr", "primary_metric_value": "0.030569594353437424", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.030569594353437424", "top1_accuracy": "0.008620689655172414", "top5_accuracy": "0.02586206896551724", "top10_accuracy": "0.04885057471264368", "median_rank": "110.5", "mean_rank": "130.32470703125", "num_queries": "348", "task_display_name": "Language Grounding" }, { "task": "caption_grounding", "task_display": "Language-to-Time Grounding", "modality_group": "inertial", "modality_display": "Inertial", "status": "computed", "score": "0.02470344305038452", "primary_metric": "mrr", "primary_metric_value": "0.02470344305038452", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.02470344305038452", "top1_accuracy": "0.0028735632183908046", "top5_accuracy": "0.022988505747126436", "top10_accuracy": "0.04597701149425287", "median_rank": "123.0", "mean_rank": "138.61207580566406", "num_queries": "348", "task_display_name": "Language Grounding" }, { "task": "caption_grounding", "task_display": "Language-to-Time Grounding", "modality_group": "audio", "modality_display": "Audio", "status": "computed", "score": "0.03902389109134674", "primary_metric": "mrr", "primary_metric_value": "0.03902389109134674", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.03902389109134674", "top1_accuracy": "0.011494252873563218", "top5_accuracy": "0.04885057471264368", "top10_accuracy": "0.07758620689655173", "median_rank": "141.0", "mean_rank": "152.14942932128906", "num_queries": "348", "task_display_name": "Language Grounding" }, { "task": "caption_grounding", "task_display": "Language-to-Time Grounding", "modality_group": "language", "modality_display": "Language", "status": "computed", "score": "0.24527303874492645", "primary_metric": "mrr", "primary_metric_value": "0.24527303874492645", "target_variant": "", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.24527303874492645", "top1_accuracy": "0.12643678160919541", "top5_accuracy": "0.34770114942528735", "top10_accuracy": "0.47126436781609193", "median_rank": "12.0", "mean_rank": "15.106322288513184", "num_queries": "348", "task_display_name": "Language Grounding" }, { "task": "caption_grounding", "task_display": "Language-to-Time Grounding", "modality_group": "no_language", "modality_display": "All Except Language", "status": "computed", "score": "0.03208239749073982", "primary_metric": "mrr", "primary_metric_value": "0.03208239749073982", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.03208239749073982", "top1_accuracy": "0.0028735632183908046", "top5_accuracy": "0.040229885057471264", "top10_accuracy": "0.06896551724137931", "median_rank": "132.0", "mean_rank": "137.30746459960938", "num_queries": "348", "task_display_name": "Language Grounding" }, { "task": "cross_modal_retrieval", "task_display": "Cross-Modal Window Retrieval", "modality_group": "all_features", "modality_display": "All Features", "status": "computed", "score": "0.972757875919342", "primary_metric": "mrr", "primary_metric_value": "0.972757875919342", "target_variant": "", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.972757875919342", "top1_accuracy": "0.9683908045977011", "top5_accuracy": "0.9741379310344828", "top10_accuracy": "0.9798850574712644", "median_rank": "1.0", "mean_rank": "2.0862069129943848", "num_queries": "348", "task_display_name": "Cross-Modal Retrieval" }, { "task": "cross_modal_retrieval", "task_display": "Cross-Modal Window Retrieval", "modality_group": "video", "modality_display": "Video", "status": "computed", "score": "0.9701701402664185", "primary_metric": "mrr", "primary_metric_value": "0.9701701402664185", "target_variant": "", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.9701701402664185", "top1_accuracy": "0.9626436781609196", "top5_accuracy": "0.9798850574712644", "top10_accuracy": "0.9798850574712644", "median_rank": "1.0", "mean_rank": "3.844827651977539", "num_queries": "348", "task_display_name": "Cross-Modal Retrieval" }, { "task": "cross_modal_retrieval", "task_display": "Cross-Modal Window Retrieval", "modality_group": "depth", "modality_display": "Depth", "status": "computed", "score": "0.6656051278114319", "primary_metric": "mrr", "primary_metric_value": "0.6656051278114319", "target_variant": "", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.6656051278114319", "top1_accuracy": "0.5660919540229885", "top5_accuracy": "0.7902298850574713", "top10_accuracy": "0.8620689655172413", "median_rank": "1.0", "mean_rank": "5.729885101318359", "num_queries": "348", "task_display_name": "Cross-Modal Retrieval" }, { "task": "cross_modal_retrieval", "task_display": "Cross-Modal Window Retrieval", "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "status": "computed", "score": "0.42622581124305725", "primary_metric": "mrr", "primary_metric_value": "0.42622581124305725", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.42622581124305725", "top1_accuracy": "0.3017241379310345", "top5_accuracy": "0.5488505747126436", "top10_accuracy": "0.6551724137931034", "median_rank": "4.0", "mean_rank": "15.623562812805176", "num_queries": "348", "task_display_name": "Cross-Modal Retrieval" }, { "task": "cross_modal_retrieval", "task_display": "Cross-Modal Window Retrieval", "modality_group": "motion_capture", "modality_display": "Motion Capture", "status": "computed", "score": "0.2553335726261139", "primary_metric": "mrr", "primary_metric_value": "0.2553335726261139", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.2553335726261139", "top1_accuracy": "0.15804597701149425", "top5_accuracy": "0.35344827586206895", "top10_accuracy": "0.3994252873563218", "median_rank": "21.5", "mean_rank": "49.181034088134766", "num_queries": "348", "task_display_name": "Cross-Modal Retrieval" }, { "task": "cross_modal_retrieval", "task_display": "Cross-Modal Window Retrieval", "modality_group": "inertial", "modality_display": "Inertial", "status": "computed", "score": "0.2840072810649872", "primary_metric": "mrr", "primary_metric_value": "0.2840072810649872", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.2840072810649872", "top1_accuracy": "0.16379310344827586", "top5_accuracy": "0.3735632183908046", "top10_accuracy": "0.5229885057471264", "median_rank": "10.0", "mean_rank": "20.577587127685547", "num_queries": "348", "task_display_name": "Cross-Modal Retrieval" }, { "task": "cross_modal_retrieval", "task_display": "Cross-Modal Window Retrieval", "modality_group": "audio", "modality_display": "Audio", "status": "computed", "score": "0.02334633097052574", "primary_metric": "mrr", "primary_metric_value": "0.02334633097052574", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.02334633097052574", "top1_accuracy": "0.005747126436781609", "top5_accuracy": "0.014367816091954023", "top10_accuracy": "0.031609195402298854", "median_rank": "152.5", "mean_rank": "161.44540405273438", "num_queries": "348", "task_display_name": "Cross-Modal Retrieval" }, { "task": "cross_modal_retrieval", "task_display": "Cross-Modal Window Retrieval", "modality_group": "language", "modality_display": "Language", "status": "computed", "score": "0.031006580218672752", "primary_metric": "mrr", "primary_metric_value": "0.031006580218672752", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.031006580218672752", "top1_accuracy": "0.005747126436781609", "top5_accuracy": "0.031609195402298854", "top10_accuracy": "0.05747126436781609", "median_rank": "138.0", "mean_rank": "146.83045959472656", "num_queries": "348", "task_display_name": "Cross-Modal Retrieval" }, { "task": "cross_modal_retrieval", "task_display": "Cross-Modal Window Retrieval", "modality_group": "no_language", "modality_display": "All Except Language", "status": "computed", "score": "0.9725185036659241", "primary_metric": "mrr", "primary_metric_value": "0.9725185036659241", "target_variant": "", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "0.9725185036659241", "top1_accuracy": "0.9683908045977011", "top5_accuracy": "0.9741379310344828", "top10_accuracy": "0.9770114942528736", "median_rank": "1.0", "mean_rank": "2.181034564971924", "num_queries": "348", "task_display_name": "Cross-Modal Retrieval" }, { "task": "modality_reconstruction", "task_display": "Sensor-to-Visual Reconstruction", "modality_group": "all_features", "modality_display": "All Features", "status": "computed", "score": "0.20013932390163613", "primary_metric": "mae", "primary_metric_value": "3.9965193271636963", "target_variant": "", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "4488.36962890625", "mae": "3.9965193271636963", "r2": "0.4789615393741845", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Cross-Modal Reconstruction" }, { "task": "modality_reconstruction", "task_display": "Sensor-to-Visual Reconstruction", "modality_group": "video", "modality_display": "Video", "status": "computed", "score": "0.611318891594774", "primary_metric": "mae", "primary_metric_value": "0.635807454586029", "target_variant": "", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "8679.7548828125", "mae": "0.635807454586029", "r2": "-0.007601057526781085", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Cross-Modal Reconstruction" }, { "task": "modality_reconstruction", "task_display": "Sensor-to-Visual Reconstruction", "modality_group": "depth", "modality_display": "Depth", "status": "computed", "score": "0.062215385980961393", "primary_metric": "mae", "primary_metric_value": "15.07319450378418", "target_variant": "", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "38000.71875", "mae": "15.07319450378418", "r2": "-3.4113648334674167", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Cross-Modal Reconstruction" }, { "task": "modality_reconstruction", "task_display": "Sensor-to-Visual Reconstruction", "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "status": "computed", "score": "0.5359235021455191", "primary_metric": "mae", "primary_metric_value": "0.8659379482269287", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "8678.9697265625", "mae": "0.8659379482269287", "r2": "-0.007509963078260462", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Cross-Modal Reconstruction" }, { "task": "modality_reconstruction", "task_display": "Sensor-to-Visual Reconstruction", "modality_group": "motion_capture", "modality_display": "Motion Capture", "status": "computed", "score": "0.07724422027114182", "primary_metric": "mae", "primary_metric_value": "11.945952415466309", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "16462.224609375", "mae": "11.945952415466309", "r2": "-0.911039534454414", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Cross-Modal Reconstruction" }, { "task": "modality_reconstruction", "task_display": "Sensor-to-Visual Reconstruction", "modality_group": "inertial", "modality_display": "Inertial", "status": "computed", "score": "0.5185351442505587", "primary_metric": "mae", "primary_metric_value": "0.9285095930099487", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "8680.1376953125", "mae": "0.9285095930099487", "r2": "-0.007645498747803181", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Cross-Modal Reconstruction" }, { "task": "modality_reconstruction", "task_display": "Sensor-to-Visual Reconstruction", "modality_group": "audio", "modality_display": "Audio", "status": "computed", "score": "0.4237932455142579", "primary_metric": "mae", "primary_metric_value": "1.359641194343567", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "8681.8916015625", "mae": "1.359641194343567", "r2": "-0.007849137404613904", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Cross-Modal Reconstruction" }, { "task": "modality_reconstruction", "task_display": "Sensor-to-Visual Reconstruction", "modality_group": "language", "modality_display": "Language", "status": "computed", "score": "0.411308516935754", "primary_metric": "mae", "primary_metric_value": "1.4312649965286255", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "8681.390625", "mae": "1.4312649965286255", "r2": "-0.00779095493123938", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Cross-Modal Reconstruction" }, { "task": "modality_reconstruction", "task_display": "Sensor-to-Visual Reconstruction", "modality_group": "no_language", "modality_display": "All Except Language", "status": "computed", "score": "0.198028246205017", "primary_metric": "mae", "primary_metric_value": "4.0497846603393555", "target_variant": "", "target_source_overlap": "true", "reason": "", "accuracy": "", "macro_f1": "", "balanced_accuracy": "", "num_classes": "", "num_train": "", "num_test": "", "unseen_test_classes": "", "unseen_test_class_count": "", "mse": "4352.54638671875", "mae": "4.0497846603393555", "r2": "0.49472875505192115", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Cross-Modal Reconstruction" }, { "task": "temporal_order", "task_display": "Temporal Order Verification", "modality_group": "all_features", "modality_display": "All Features", "status": "computed", "score": "0.5172413793103449", "primary_metric": "macro_f1", "primary_metric_value": "0.5172413793103449", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.5172413793103449", "macro_f1": "0.5172413793103449", "balanced_accuracy": "0.5172413793103449", "num_classes": "2", "num_train": "1624", "num_test": "696", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Temporal Order Verification" }, { "task": "temporal_order", "task_display": "Temporal Order Verification", "modality_group": "video", "modality_display": "Video", "status": "computed", "score": "0.5172413793103449", "primary_metric": "macro_f1", "primary_metric_value": "0.5172413793103449", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.5172413793103449", "macro_f1": "0.5172413793103449", "balanced_accuracy": "0.5172413793103449", "num_classes": "2", "num_train": "1624", "num_test": "696", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Temporal Order Verification" }, { "task": "temporal_order", "task_display": "Temporal Order Verification", "modality_group": "depth", "modality_display": "Depth", "status": "computed", "score": "0.49424869738982513", "primary_metric": "macro_f1", "primary_metric_value": "0.49424869738982513", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.4942528735632184", "macro_f1": "0.49424869738982513", "balanced_accuracy": "0.49425287356321834", "num_classes": "2", "num_train": "1624", "num_test": "696", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Temporal Order Verification" }, { "task": "temporal_order", "task_display": "Temporal Order Verification", "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "status": "computed", "score": "0.5258620689655172", "primary_metric": "macro_f1", "primary_metric_value": "0.5258620689655172", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.5258620689655172", "macro_f1": "0.5258620689655172", "balanced_accuracy": "0.5258620689655172", "num_classes": "2", "num_train": "1624", "num_test": "696", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Temporal Order Verification" }, { "task": "temporal_order", "task_display": "Temporal Order Verification", "modality_group": "motion_capture", "modality_display": "Motion Capture", "status": "computed", "score": "0.4942528735632184", "primary_metric": "macro_f1", "primary_metric_value": "0.4942528735632184", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.4942528735632184", "macro_f1": "0.4942528735632184", "balanced_accuracy": "0.4942528735632184", "num_classes": "2", "num_train": "1624", "num_test": "696", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Temporal Order Verification" }, { "task": "temporal_order", "task_display": "Temporal Order Verification", "modality_group": "inertial", "modality_display": "Inertial", "status": "computed", "score": "0.5", "primary_metric": "macro_f1", "primary_metric_value": "0.5", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.5", "macro_f1": "0.5", "balanced_accuracy": "0.5", "num_classes": "2", "num_train": "1624", "num_test": "696", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Temporal Order Verification" }, { "task": "temporal_order", "task_display": "Temporal Order Verification", "modality_group": "audio", "modality_display": "Audio", "status": "computed", "score": "0.4425287356321839", "primary_metric": "macro_f1", "primary_metric_value": "0.4425287356321839", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.4425287356321839", "macro_f1": "0.4425287356321839", "balanced_accuracy": "0.4425287356321839", "num_classes": "2", "num_train": "1624", "num_test": "696", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Temporal Order Verification" }, { "task": "temporal_order", "task_display": "Temporal Order Verification", "modality_group": "language", "modality_display": "Language", "status": "computed", "score": "0.4236751152073733", "primary_metric": "macro_f1", "primary_metric_value": "0.4236751152073733", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.47126436781609193", "macro_f1": "0.4236751152073733", "balanced_accuracy": "0.47126436781609193", "num_classes": "2", "num_train": "1624", "num_test": "696", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Temporal Order Verification" }, { "task": "temporal_order", "task_display": "Temporal Order Verification", "modality_group": "no_language", "modality_display": "All Except Language", "status": "computed", "score": "0.5057471264367817", "primary_metric": "macro_f1", "primary_metric_value": "0.5057471264367817", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.5057471264367817", "macro_f1": "0.5057471264367817", "balanced_accuracy": "0.5057471264367817", "num_classes": "2", "num_train": "1624", "num_test": "696", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Temporal Order Verification" }, { "task": "misalignment_detection", "task_display": "Cross-Modal Misalignment Detection", "modality_group": "all_features", "modality_display": "All Features", "status": "computed", "score": "0.4159692457248906", "primary_metric": "macro_f1", "primary_metric_value": "0.4159692457248906", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.4479768786127168", "macro_f1": "0.4159692457248906", "balanced_accuracy": "0.44797687861271673", "num_classes": "2", "num_train": "1614", "num_test": "692", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Multimodal Synchronization Detection" }, { "task": "misalignment_detection", "task_display": "Cross-Modal Misalignment Detection", "modality_group": "video", "modality_display": "Video", "status": "computed", "score": "0.49488307322727143", "primary_metric": "macro_f1", "primary_metric_value": "0.49488307322727143", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.4985549132947977", "macro_f1": "0.49488307322727143", "balanced_accuracy": "0.4985549132947977", "num_classes": "2", "num_train": "1614", "num_test": "692", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Multimodal Synchronization Detection" }, { "task": "misalignment_detection", "task_display": "Cross-Modal Misalignment Detection", "modality_group": "depth", "modality_display": "Depth", "status": "computed", "score": "0.46659963973021656", "primary_metric": "macro_f1", "primary_metric_value": "0.46659963973021656", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.4682080924855491", "macro_f1": "0.46659963973021656", "balanced_accuracy": "0.4682080924855491", "num_classes": "2", "num_train": "1614", "num_test": "692", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Multimodal Synchronization Detection" }, { "task": "misalignment_detection", "task_display": "Cross-Modal Misalignment Detection", "modality_group": "pose_slam", "modality_display": "Pose + SLAM", "status": "computed", "score": "0.4929686094043242", "primary_metric": "macro_f1", "primary_metric_value": "0.4929686094043242", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.5057803468208093", "macro_f1": "0.4929686094043242", "balanced_accuracy": "0.5057803468208092", "num_classes": "2", "num_train": "1614", "num_test": "692", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Multimodal Synchronization Detection" }, { "task": "misalignment_detection", "task_display": "Cross-Modal Misalignment Detection", "modality_group": "motion_capture", "modality_display": "Motion Capture", "status": "computed", "score": "0.4133918268956141", "primary_metric": "macro_f1", "primary_metric_value": "0.4133918268956141", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.4638728323699422", "macro_f1": "0.4133918268956141", "balanced_accuracy": "0.4638728323699422", "num_classes": "2", "num_train": "1614", "num_test": "692", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Multimodal Synchronization Detection" }, { "task": "misalignment_detection", "task_display": "Cross-Modal Misalignment Detection", "modality_group": "inertial", "modality_display": "Inertial", "status": "computed", "score": "0.48899072503396884", "primary_metric": "macro_f1", "primary_metric_value": "0.48899072503396884", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.49421965317919075", "macro_f1": "0.48899072503396884", "balanced_accuracy": "0.49421965317919075", "num_classes": "2", "num_train": "1614", "num_test": "692", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Multimodal Synchronization Detection" }, { "task": "misalignment_detection", "task_display": "Cross-Modal Misalignment Detection", "modality_group": "audio", "modality_display": "Audio", "status": "computed", "score": "0.5152624305034046", "primary_metric": "macro_f1", "primary_metric_value": "0.5152624305034046", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.5187861271676301", "macro_f1": "0.5152624305034046", "balanced_accuracy": "0.5187861271676301", "num_classes": "2", "num_train": "1614", "num_test": "692", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Multimodal Synchronization Detection" }, { "task": "misalignment_detection", "task_display": "Cross-Modal Misalignment Detection", "modality_group": "language", "modality_display": "Language", "status": "computed", "score": "0.4942161609504254", "primary_metric": "macro_f1", "primary_metric_value": "0.4942161609504254", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.5", "macro_f1": "0.4942161609504254", "balanced_accuracy": "0.5", "num_classes": "2", "num_train": "1614", "num_test": "692", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Multimodal Synchronization Detection" }, { "task": "misalignment_detection", "task_display": "Cross-Modal Misalignment Detection", "modality_group": "no_language", "modality_display": "All Except Language", "status": "computed", "score": "0.4136154066851563", "primary_metric": "macro_f1", "primary_metric_value": "0.4136154066851563", "target_variant": "", "target_source_overlap": "false", "reason": "", "accuracy": "0.44942196531791906", "macro_f1": "0.4136154066851563", "balanced_accuracy": "0.4494219653179191", "num_classes": "2", "num_train": "1614", "num_test": "692", "unseen_test_classes": "", "unseen_test_class_count": "0", "mse": "", "mae": "", "r2": "", "micro_f1": "", "exact_match": "", "precision": "", "recall": "", "num_objects": "", "mrr": "", "top1_accuracy": "", "top5_accuracy": "", "top10_accuracy": "", "median_rank": "", "mean_rank": "", "num_queries": "", "task_display_name": "Multimodal Synchronization Detection" } ] }, "alignment": [ { "query_group": "motion_capture", "query_display": "Motion Capture", "target_group": "depth_plus_video", "shift_windows": "-40", "shift_frames": "-200", "status": "derived_perturbation", "mrr": "0.02984776347875595", "top1_accuracy": "0.006493506493506494", "top5_accuracy": "0.032467532467532464", "top10_accuracy": "0.048701298701298704", "median_rank": "119.0", "mean_rank": "125.72402954101562", "num_queries": "308" }, { "query_group": "motion_capture", "query_display": "Motion Capture", "target_group": "depth_plus_video", "shift_windows": "-20", "shift_frames": "-100", "status": "derived_perturbation", "mrr": "0.019169922918081284", "top1_accuracy": "0.003048780487804878", "top5_accuracy": "0.003048780487804878", "top10_accuracy": "0.018292682926829267", "median_rank": "153.0", "mean_rank": "154.24085998535156", "num_queries": "328" }, { "query_group": "motion_capture", "query_display": "Motion Capture", "target_group": "depth_plus_video", "shift_windows": "-10", "shift_frames": "-50", "status": "derived_perturbation", "mrr": "0.01829293556511402", "top1_accuracy": "0.0", "top5_accuracy": "0.0029585798816568047", "top10_accuracy": "0.023668639053254437", "median_rank": "106.0", "mean_rank": "130.5325469970703", "num_queries": "338" }, { "query_group": "motion_capture", "query_display": "Motion Capture", "target_group": "depth_plus_video", "shift_windows": "-5", "shift_frames": "-25", "status": "derived_perturbation", "mrr": "0.028951412066817284", "top1_accuracy": "0.0", "top5_accuracy": "0.014577259475218658", "top10_accuracy": "0.05830903790087463", "median_rank": "67.0", "mean_rank": "89.93585968017578", "num_queries": "343" }, { "query_group": "motion_capture", "query_display": "Motion Capture", "target_group": "depth_plus_video", "shift_windows": "0", "shift_frames": "0", "status": "derived_perturbation", "mrr": "0.2553335726261139", "top1_accuracy": "0.15804597701149425", "top5_accuracy": "0.35344827586206895", "top10_accuracy": "0.3994252873563218", "median_rank": "21.5", "mean_rank": "49.181034088134766", "num_queries": "348" }, { "query_group": "motion_capture", "query_display": "Motion Capture", "target_group": "depth_plus_video", "shift_windows": "5", "shift_frames": "25", "status": "derived_perturbation", "mrr": "0.04436318948864937", "top1_accuracy": "0.008746355685131196", "top5_accuracy": "0.037900874635568516", "top10_accuracy": "0.08746355685131195", "median_rank": "64.0", "mean_rank": "83.76384735107422", "num_queries": "343" }, { "query_group": "motion_capture", "query_display": "Motion Capture", "target_group": "depth_plus_video", "shift_windows": "10", "shift_frames": "50", "status": "derived_perturbation", "mrr": "0.026273079216480255", "top1_accuracy": "0.0", "top5_accuracy": "0.014792899408284023", "top10_accuracy": "0.047337278106508875", "median_rank": "77.0", "mean_rank": "106.73668670654297", "num_queries": "338" }, { "query_group": "motion_capture", "query_display": "Motion Capture", "target_group": "depth_plus_video", "shift_windows": "20", "shift_frames": "100", "status": "derived_perturbation", "mrr": "0.023496314883232117", "top1_accuracy": "0.003048780487804878", "top5_accuracy": "0.018292682926829267", "top10_accuracy": "0.04573170731707317", "median_rank": "108.5", "mean_rank": "137.9176788330078", "num_queries": "328" }, { "query_group": "motion_capture", "query_display": "Motion Capture", "target_group": "depth_plus_video", "shift_windows": "40", "shift_frames": "200", "status": "derived_perturbation", "mrr": "0.02917252667248249", "top1_accuracy": "0.006493506493506494", "top5_accuracy": "0.025974025974025976", "top10_accuracy": "0.05519480519480519", "median_rank": "110.5", "mean_rank": "121.33441925048828", "num_queries": "308" }, { "query_group": "pose_slam", "query_display": "Pose + SLAM", "target_group": "depth_plus_video", "shift_windows": "-40", "shift_frames": "-200", "status": "derived_perturbation", "mrr": "0.03615332394838333", "top1_accuracy": "0.003246753246753247", "top5_accuracy": "0.03571428571428571", "top10_accuracy": "0.07467532467532467", "median_rank": "98.0", "mean_rank": "114.43506622314453", "num_queries": "308" }, { "query_group": "pose_slam", "query_display": "Pose + SLAM", "target_group": "depth_plus_video", "shift_windows": "-20", "shift_frames": "-100", "status": "derived_perturbation", "mrr": "0.02059117704629898", "top1_accuracy": "0.003048780487804878", "top5_accuracy": "0.012195121951219513", "top10_accuracy": "0.024390243902439025", "median_rank": "109.5", "mean_rank": "137.0731658935547", "num_queries": "328" }, { "query_group": "pose_slam", "query_display": "Pose + SLAM", "target_group": "depth_plus_video", "shift_windows": "-10", "shift_frames": "-50", "status": "derived_perturbation", "mrr": "0.04128313437104225", "top1_accuracy": "0.005917159763313609", "top5_accuracy": "0.038461538461538464", "top10_accuracy": "0.07692307692307693", "median_rank": "72.0", "mean_rank": "103.94674682617188", "num_queries": "338" }, { "query_group": "pose_slam", "query_display": "Pose + SLAM", "target_group": "depth_plus_video", "shift_windows": "-5", "shift_frames": "-25", "status": "derived_perturbation", "mrr": "0.05835483595728874", "top1_accuracy": "0.011661807580174927", "top5_accuracy": "0.061224489795918366", "top10_accuracy": "0.119533527696793", "median_rank": "43.0", "mean_rank": "58.218658447265625", "num_queries": "343" }, { "query_group": "pose_slam", "query_display": "Pose + SLAM", "target_group": "depth_plus_video", "shift_windows": "0", "shift_frames": "0", "status": "derived_perturbation", "mrr": "0.42622581124305725", "top1_accuracy": "0.3017241379310345", "top5_accuracy": "0.5488505747126436", "top10_accuracy": "0.6551724137931034", "median_rank": "4.0", "mean_rank": "15.623562812805176", "num_queries": "348" }, { "query_group": "pose_slam", "query_display": "Pose + SLAM", "target_group": "depth_plus_video", "shift_windows": "5", "shift_frames": "25", "status": "derived_perturbation", "mrr": "0.04654298722743988", "top1_accuracy": "0.0058309037900874635", "top5_accuracy": "0.04956268221574344", "top10_accuracy": "0.11661807580174927", "median_rank": "55.0", "mean_rank": "66.43148803710938", "num_queries": "343" }, { "query_group": "pose_slam", "query_display": "Pose + SLAM", "target_group": "depth_plus_video", "shift_windows": "10", "shift_frames": "50", "status": "derived_perturbation", "mrr": "0.034309279173612595", "top1_accuracy": "0.005917159763313609", "top5_accuracy": "0.023668639053254437", "top10_accuracy": "0.05621301775147929", "median_rank": "71.0", "mean_rank": "100.44082641601562", "num_queries": "338" }, { "query_group": "pose_slam", "query_display": "Pose + SLAM", "target_group": "depth_plus_video", "shift_windows": "20", "shift_frames": "100", "status": "derived_perturbation", "mrr": "0.03287472575902939", "top1_accuracy": "0.006097560975609756", "top5_accuracy": "0.03048780487804878", "top10_accuracy": "0.06097560975609756", "median_rank": "97.5", "mean_rank": "127.41158294677734", "num_queries": "328" }, { "query_group": "pose_slam", "query_display": "Pose + SLAM", "target_group": "depth_plus_video", "shift_windows": "40", "shift_frames": "200", "status": "derived_perturbation", "mrr": "0.024975253269076347", "top1_accuracy": "0.003246753246753247", "top5_accuracy": "0.016233766233766232", "top10_accuracy": "0.03571428571428571", "median_rank": "88.5", "mean_rank": "116.36363983154297", "num_queries": "308" }, { "query_group": "inertial", "query_display": "Inertial", "target_group": "depth_plus_video", "shift_windows": "-40", "shift_frames": "-200", "status": "derived_perturbation", "mrr": "0.042965441942214966", "top1_accuracy": "0.00974025974025974", "top5_accuracy": "0.045454545454545456", "top10_accuracy": "0.09090909090909091", "median_rank": "90.0", "mean_rank": "116.86363983154297", "num_queries": "308" }, { "query_group": "inertial", "query_display": "Inertial", "target_group": "depth_plus_video", "shift_windows": "-20", "shift_frames": "-100", "status": "derived_perturbation", "mrr": "0.019861916080117226", "top1_accuracy": "0.003048780487804878", "top5_accuracy": "0.009146341463414634", "top10_accuracy": "0.01524390243902439", "median_rank": "112.0", "mean_rank": "135.9573211669922", "num_queries": "328" }, { "query_group": "inertial", "query_display": "Inertial", "target_group": "depth_plus_video", "shift_windows": "-10", "shift_frames": "-50", "status": "derived_perturbation", "mrr": "0.04950016736984253", "top1_accuracy": "0.011834319526627219", "top5_accuracy": "0.05325443786982249", "top10_accuracy": "0.10946745562130178", "median_rank": "74.0", "mean_rank": "102.37574005126953", "num_queries": "338" }, { "query_group": "inertial", "query_display": "Inertial", "target_group": "depth_plus_video", "shift_windows": "-5", "shift_frames": "-25", "status": "derived_perturbation", "mrr": "0.05499911680817604", "top1_accuracy": "0.0029154518950437317", "top5_accuracy": "0.05830903790087463", "top10_accuracy": "0.13994169096209913", "median_rank": "39.0", "mean_rank": "63.70845413208008", "num_queries": "343" }, { "query_group": "inertial", "query_display": "Inertial", "target_group": "depth_plus_video", "shift_windows": "0", "shift_frames": "0", "status": "derived_perturbation", "mrr": "0.2840072810649872", "top1_accuracy": "0.16379310344827586", "top5_accuracy": "0.3735632183908046", "top10_accuracy": "0.5229885057471264", "median_rank": "10.0", "mean_rank": "20.577587127685547", "num_queries": "348" }, { "query_group": "inertial", "query_display": "Inertial", "target_group": "depth_plus_video", "shift_windows": "5", "shift_frames": "25", "status": "derived_perturbation", "mrr": "0.054161082953214645", "top1_accuracy": "0.014577259475218658", "top5_accuracy": "0.04956268221574344", "top10_accuracy": "0.10495626822157435", "median_rank": "53.0", "mean_rank": "68.86006164550781", "num_queries": "343" }, { "query_group": "inertial", "query_display": "Inertial", "target_group": "depth_plus_video", "shift_windows": "10", "shift_frames": "50", "status": "derived_perturbation", "mrr": "0.03178250044584274", "top1_accuracy": "0.005917159763313609", "top5_accuracy": "0.008875739644970414", "top10_accuracy": "0.05917159763313609", "median_rank": "76.0", "mean_rank": "103.02071380615234", "num_queries": "338" }, { "query_group": "inertial", "query_display": "Inertial", "target_group": "depth_plus_video", "shift_windows": "20", "shift_frames": "100", "status": "derived_perturbation", "mrr": "0.03213934600353241", "top1_accuracy": "0.009146341463414634", "top5_accuracy": "0.021341463414634148", "top10_accuracy": "0.042682926829268296", "median_rank": "93.5", "mean_rank": "125.67987823486328", "num_queries": "328" }, { "query_group": "inertial", "query_display": "Inertial", "target_group": "depth_plus_video", "shift_windows": "40", "shift_frames": "200", "status": "derived_perturbation", "mrr": "0.031400587409734726", "top1_accuracy": "0.003246753246753247", "top5_accuracy": "0.03896103896103896", "top10_accuracy": "0.05194805194805195", "median_rank": "91.0", "mean_rank": "115.68830871582031", "num_queries": "308" }, { "query_group": "language", "query_display": "Language", "target_group": "depth_plus_video", "shift_windows": "-40", "shift_frames": "-200", "status": "derived_perturbation", "mrr": "0.015811588615179062", "top1_accuracy": "0.0", "top5_accuracy": "0.003246753246753247", "top10_accuracy": "0.016233766233766232", "median_rank": "145.5", "mean_rank": "141.49026489257812", "num_queries": "308" }, { "query_group": "language", "query_display": "Language", "target_group": "depth_plus_video", "shift_windows": "-20", "shift_frames": "-100", "status": "derived_perturbation", "mrr": "0.027325116097927094", "top1_accuracy": "0.006097560975609756", "top5_accuracy": "0.024390243902439025", "top10_accuracy": "0.06097560975609756", "median_rank": "174.0", "mean_rank": "162.0792694091797", "num_queries": "328" }, { "query_group": "language", "query_display": "Language", "target_group": "depth_plus_video", "shift_windows": "-10", "shift_frames": "-50", "status": "derived_perturbation", "mrr": "0.02521640993654728", "top1_accuracy": "0.0029585798816568047", "top5_accuracy": "0.023668639053254437", "top10_accuracy": "0.05621301775147929", "median_rank": "165.0", "mean_rank": "162.10354614257812", "num_queries": "338" }, { "query_group": "language", "query_display": "Language", "target_group": "depth_plus_video", "shift_windows": "-5", "shift_frames": "-25", "status": "derived_perturbation", "mrr": "0.02469729632139206", "top1_accuracy": "0.0029154518950437317", "top5_accuracy": "0.02040816326530612", "top10_accuracy": "0.04956268221574344", "median_rank": "165.0", "mean_rank": "158.99708557128906", "num_queries": "343" }, { "query_group": "language", "query_display": "Language", "target_group": "depth_plus_video", "shift_windows": "0", "shift_frames": "0", "status": "derived_perturbation", "mrr": "0.031006580218672752", "top1_accuracy": "0.005747126436781609", "top5_accuracy": "0.031609195402298854", "top10_accuracy": "0.05747126436781609", "median_rank": "138.0", "mean_rank": "146.83045959472656", "num_queries": "348" }, { "query_group": "language", "query_display": "Language", "target_group": "depth_plus_video", "shift_windows": "5", "shift_frames": "25", "status": "derived_perturbation", "mrr": "0.04090346768498421", "top1_accuracy": "0.008746355685131196", "top5_accuracy": "0.04956268221574344", "top10_accuracy": "0.08454810495626822", "median_rank": "102.0", "mean_rank": "135.07289123535156", "num_queries": "343" }, { "query_group": "language", "query_display": "Language", "target_group": "depth_plus_video", "shift_windows": "10", "shift_frames": "50", "status": "derived_perturbation", "mrr": "0.0362100675702095", "top1_accuracy": "0.008875739644970414", "top5_accuracy": "0.03254437869822485", "top10_accuracy": "0.07692307692307693", "median_rank": "101.5", "mean_rank": "131.18934631347656", "num_queries": "338" }, { "query_group": "language", "query_display": "Language", "target_group": "depth_plus_video", "shift_windows": "20", "shift_frames": "100", "status": "derived_perturbation", "mrr": "0.03773954510688782", "top1_accuracy": "0.009146341463414634", "top5_accuracy": "0.036585365853658534", "top10_accuracy": "0.08231707317073171", "median_rank": "111.0", "mean_rank": "137.8353729248047", "num_queries": "328" }, { "query_group": "language", "query_display": "Language", "target_group": "depth_plus_video", "shift_windows": "40", "shift_frames": "200", "status": "derived_perturbation", "mrr": "0.037675727158784866", "top1_accuracy": "0.00974025974025974", "top5_accuracy": "0.04220779220779221", "top10_accuracy": "0.08116883116883117", "median_rank": "139.5", "mean_rank": "144.4837646484375", "num_queries": "308" }, { "query_group": "motion_pose_inertial", "query_display": "Motion + Pose + IMU", "target_group": "depth_plus_video", "shift_windows": "-40", "shift_frames": "-200", "status": "derived_perturbation", "mrr": "0.05048111826181412", "top1_accuracy": "0.016233766233766232", "top5_accuracy": "0.05519480519480519", "top10_accuracy": "0.1038961038961039", "median_rank": "99.5", "mean_rank": "116.717529296875", "num_queries": "308" }, { "query_group": "motion_pose_inertial", "query_display": "Motion + Pose + IMU", "target_group": "depth_plus_video", "shift_windows": "-20", "shift_frames": "-100", "status": "derived_perturbation", "mrr": "0.023761091753840446", "top1_accuracy": "0.003048780487804878", "top5_accuracy": "0.01524390243902439", "top10_accuracy": "0.021341463414634148", "median_rank": "111.0", "mean_rank": "140.868896484375", "num_queries": "328" }, { "query_group": "motion_pose_inertial", "query_display": "Motion + Pose + IMU", "target_group": "depth_plus_video", "shift_windows": "-10", "shift_frames": "-50", "status": "derived_perturbation", "mrr": "0.039904821664094925", "top1_accuracy": "0.008875739644970414", "top5_accuracy": "0.03550295857988166", "top10_accuracy": "0.07988165680473373", "median_rank": "81.5", "mean_rank": "107.72189331054688", "num_queries": "338" }, { "query_group": "motion_pose_inertial", "query_display": "Motion + Pose + IMU", "target_group": "depth_plus_video", "shift_windows": "-5", "shift_frames": "-25", "status": "derived_perturbation", "mrr": "0.051686227321624756", "top1_accuracy": "0.008746355685131196", "top5_accuracy": "0.043731778425655975", "top10_accuracy": "0.12244897959183673", "median_rank": "47.0", "mean_rank": "63.40524673461914", "num_queries": "343" }, { "query_group": "motion_pose_inertial", "query_display": "Motion + Pose + IMU", "target_group": "depth_plus_video", "shift_windows": "0", "shift_frames": "0", "status": "derived_perturbation", "mrr": "0.38971078395843506", "top1_accuracy": "0.28448275862068967", "top5_accuracy": "0.4827586206896552", "top10_accuracy": "0.5718390804597702", "median_rank": "6.0", "mean_rank": "25.27011489868164", "num_queries": "348" }, { "query_group": "motion_pose_inertial", "query_display": "Motion + Pose + IMU", "target_group": "depth_plus_video", "shift_windows": "5", "shift_frames": "25", "status": "derived_perturbation", "mrr": "0.05908266454935074", "top1_accuracy": "0.014577259475218658", "top5_accuracy": "0.0641399416909621", "top10_accuracy": "0.13119533527696792", "median_rank": "48.0", "mean_rank": "64.80757904052734", "num_queries": "343" }, { "query_group": "motion_pose_inertial", "query_display": "Motion + Pose + IMU", "target_group": "depth_plus_video", "shift_windows": "10", "shift_frames": "50", "status": "derived_perturbation", "mrr": "0.03273069113492966", "top1_accuracy": "0.0029585798816568047", "top5_accuracy": "0.01775147928994083", "top10_accuracy": "0.06804733727810651", "median_rank": "63.0", "mean_rank": "96.31952667236328", "num_queries": "338" }, { "query_group": "motion_pose_inertial", "query_display": "Motion + Pose + IMU", "target_group": "depth_plus_video", "shift_windows": "20", "shift_frames": "100", "status": "derived_perturbation", "mrr": "0.028844518586993217", "top1_accuracy": "0.006097560975609756", "top5_accuracy": "0.024390243902439025", "top10_accuracy": "0.04573170731707317", "median_rank": "81.5", "mean_rank": "131.6280517578125", "num_queries": "328" }, { "query_group": "motion_pose_inertial", "query_display": "Motion + Pose + IMU", "target_group": "depth_plus_video", "shift_windows": "40", "shift_frames": "200", "status": "derived_perturbation", "mrr": "0.033500298857688904", "top1_accuracy": "0.00974025974025974", "top5_accuracy": "0.032467532467532464", "top10_accuracy": "0.048701298701298704", "median_rank": "99.0", "mean_rank": "120.70130157470703", "num_queries": "308" }, { "query_group": "audio", "query_display": "Audio", "target_group": "depth_plus_video", "shift_windows": "-40", "shift_frames": "-200", "status": "derived_perturbation", "mrr": "0.0169698353856802", "top1_accuracy": "0.0", "top5_accuracy": "0.012987012987012988", "top10_accuracy": "0.01948051948051948", "median_rank": "174.5", "mean_rank": "164.24349975585938", "num_queries": "308" }, { "query_group": "audio", "query_display": "Audio", "target_group": "depth_plus_video", "shift_windows": "-20", "shift_frames": "-100", "status": "derived_perturbation", "mrr": "0.020251398906111717", "top1_accuracy": "0.0", "top5_accuracy": "0.012195121951219513", "top10_accuracy": "0.036585365853658534", "median_rank": "165.0", "mean_rank": "155.72256469726562", "num_queries": "328" }, { "query_group": "audio", "query_display": "Audio", "target_group": "depth_plus_video", "shift_windows": "-10", "shift_frames": "-50", "status": "derived_perturbation", "mrr": "0.017801525071263313", "top1_accuracy": "0.0", "top5_accuracy": "0.020710059171597635", "top10_accuracy": "0.029585798816568046", "median_rank": "158.0", "mean_rank": "167.45266723632812", "num_queries": "338" }, { "query_group": "audio", "query_display": "Audio", "target_group": "depth_plus_video", "shift_windows": "-5", "shift_frames": "-25", "status": "derived_perturbation", "mrr": "0.03099018894135952", "top1_accuracy": "0.011661807580174927", "top5_accuracy": "0.029154518950437316", "top10_accuracy": "0.04664723032069971", "median_rank": "139.0", "mean_rank": "157.4344024658203", "num_queries": "343" }, { "query_group": "audio", "query_display": "Audio", "target_group": "depth_plus_video", "shift_windows": "0", "shift_frames": "0", "status": "derived_perturbation", "mrr": "0.02334633097052574", "top1_accuracy": "0.005747126436781609", "top5_accuracy": "0.014367816091954023", "top10_accuracy": "0.031609195402298854", "median_rank": "152.5", "mean_rank": "161.44540405273438", "num_queries": "348" }, { "query_group": "audio", "query_display": "Audio", "target_group": "depth_plus_video", "shift_windows": "5", "shift_frames": "25", "status": "derived_perturbation", "mrr": "0.019279710948467255", "top1_accuracy": "0.0029154518950437317", "top5_accuracy": "0.011661807580174927", "top10_accuracy": "0.02040816326530612", "median_rank": "142.0", "mean_rank": "152.10494995117188", "num_queries": "343" }, { "query_group": "audio", "query_display": "Audio", "target_group": "depth_plus_video", "shift_windows": "10", "shift_frames": "50", "status": "derived_perturbation", "mrr": "0.025376494973897934", "top1_accuracy": "0.0029585798816568047", "top5_accuracy": "0.023668639053254437", "top10_accuracy": "0.05621301775147929", "median_rank": "139.0", "mean_rank": "148.3550262451172", "num_queries": "338" }, { "query_group": "audio", "query_display": "Audio", "target_group": "depth_plus_video", "shift_windows": "20", "shift_frames": "100", "status": "derived_perturbation", "mrr": "0.020232554525136948", "top1_accuracy": "0.003048780487804878", "top5_accuracy": "0.012195121951219513", "top10_accuracy": "0.03048780487804878", "median_rank": "135.5", "mean_rank": "148.68597412109375", "num_queries": "328" }, { "query_group": "audio", "query_display": "Audio", "target_group": "depth_plus_video", "shift_windows": "40", "shift_frames": "200", "status": "derived_perturbation", "mrr": "0.013224045746028423", "top1_accuracy": "0.0", "top5_accuracy": "0.006493506493506494", "top10_accuracy": "0.012987012987012988", "median_rank": "168.0", "mean_rank": "168.0779266357422", "num_queries": "308" } ] }