# Eval Pack: 18 scenes (3D-FRONT 8-vector)
#
# 对应评测结果: PSNR=17.6470  SSIM=0.6247  LPIPS=0.5402  (≈ 17.64 / 0.624 / 0.540)
# 来源 csv: nips26_data/3dfront_8vector_metrics_first100.csv
#
# 目录结构:
#   poses/<scene_id>.txt          评测用前 100 个相机位姿 (x y z qx qy qz qw, OpenGL camera-to-B)
#   gt_images/<scene_id>/         GT 透视图前 100 张 (000000.png ~ 000099.png, 512x512)
#   panorama/<scene_id>/          原始全景数据
#       frame_0000_rgb.png        全景 RGB (1024x512, equirectangular)
#       frame_0000_depth.png      全景 depth (1024x512, 16bit PNG, 单位 mm, scale=1000.0)
#       frame_0000_depth.npy     全景 depth (512x1024, float32, 单位 m)
#       trajectory.json          原始轨迹 (camera_locations 等)
#
# 渲染内参 (固定): 512x512, fovy=60deg, fx=fy=443.4050067, cx=cy=256
# 位姿坐标系: B 坐标系由 trajectory.json 的 camera_locations[0](原点) 和 [-1](-y方向) 定义, z=[0,0,1]
#
# 18 个场景 ID:
00110bde-f580-40be-b8bb-88715b338a2a
0018b6c8-c3b6-4fb8-a640-4b9b0b763254
1e16532d-1ec8-42c6-a66f-ef21e31c22c1
27127091-beea-4597-807f-83a71d05b3ac
312c3afe-ae86-4823-b67e-292660fc85f8
40eb416e-d9ae-4c25-8d4c-a2a2306e8a21
43921e8e-fbe4-4eca-b1be-99969dda2224
5a8a69a6-6c69-45a2-a0c6-75693d542b4c
65f862fe-7cee-4e29-9ff1-b9a73396a5ac
699eac12-e518-42e3-9c8b-4ecee18a9077
708218c2-ff5f-450f-9a8a-c82e6ad896a9
7ba559ee-a2d9-425c-9111-cf2a9c543c57
8463261b-999e-4506-a090-7fffcc106adb
9b8b8318-45e8-4518-ba42-6512352c5ca1
9ba61fbc-ec92-4421-932b-68f5adb9b08e
b8d5c14b-b931-4b1e-969f-b96f5c1f3bb5
bd3f0f42-c2a0-4c1a-b821-81336d57d247
c28564ac-4bcd-4200-bb5a-ea45fbf627bb