| |
| """ |
| Build a static interactive explorer for the single Xperience-10M sample episode. |
| |
| The explorer is generated from committed/exported artifacts only. Raw MP4/HDF5 |
| files are not embedded or redistributed. |
| """ |
|
|
| from __future__ import annotations |
|
|
| import argparse |
| import csv |
| import json |
| from datetime import datetime, timezone |
| from pathlib import Path |
|
|
| import numpy as np |
|
|
| from task_display import TASK_DISPLAY_NAMES, task_display_name |
|
|
|
|
| TASK_DISPLAY = { |
| "timeline_action": task_display_name("timeline_action"), |
| "timeline_subtask": task_display_name("timeline_subtask"), |
| "transition_detection": task_display_name("transition_detection"), |
| "next_action": task_display_name("next_action"), |
| "contact_prediction": task_display_name("contact_prediction"), |
| "object_relevance": task_display_name("object_relevance"), |
| } |
| TASK_DISPLAY_ALL = dict(TASK_DISPLAY_NAMES) |
|
|
|
|
| BLOCK_DISPLAY = { |
| "hand_left_joints": "Left Hand", |
| "hand_right_joints": "Right Hand", |
| "body_joints": "Body Joints", |
| "body_contacts": "Body Contacts", |
| "camera_translation": "Camera Translation", |
| "camera_rotation_matrix": "Camera Rotation", |
| "imu_accel_gyro": "IMU Accel/Gyro", |
| "depth_confidence": "Depth + Confidence", |
| "audio_fisheye_cam0_aac": "Audio", |
| "caption_objects_interaction_text": "Language Text", |
| "slam_point_cloud": "SLAM Point Cloud", |
| "calibration": "Calibration", |
| } |
|
|
|
|
| def parse_args() -> argparse.Namespace: |
| root = Path(__file__).resolve().parents[1] |
| parser = argparse.ArgumentParser(description="Build static single-episode explorer page.") |
| parser.add_argument("--workspace", type=Path, default=root) |
| parser.add_argument("--suite-dir", type=Path, default=root / "results/episode_task_suite") |
| parser.add_argument("--diagnostics-dir", type=Path, default=root / "results/single_episode_diagnostics") |
| parser.add_argument("--docs-dir", type=Path, default=root / "docs") |
| return parser.parse_args() |
|
|
|
|
| def read_csv(path: Path) -> list[dict]: |
| with path.open(newline="", encoding="utf-8") as fp: |
| return list(csv.DictReader(fp)) |
|
|
|
|
| def read_json(path: Path): |
| return json.loads(path.read_text(encoding="utf-8")) |
|
|
|
|
| def write_json(path: Path, data: dict) -> None: |
| path.parent.mkdir(parents=True, exist_ok=True) |
| path.write_text(json.dumps(data, indent=2, ensure_ascii=False) + "\n", encoding="utf-8") |
|
|
|
|
| def block_modality(name: str) -> str: |
| if name.startswith("video_"): |
| return "video" |
| if name.startswith("hand_") or name.startswith("body_"): |
| return "motion_capture" |
| if name.startswith("camera_") or name in {"slam_point_cloud", "calibration"}: |
| return "pose_slam" |
| if name.startswith("depth_"): |
| return "depth" |
| if name.startswith("imu_"): |
| return "inertial" |
| if name.startswith("audio_"): |
| return "audio" |
| if name.startswith("caption_"): |
| return "language" |
| return "other" |
|
|
|
|
| def load_predictions(suite_dir: Path) -> dict[str, dict[int, dict]]: |
| out: dict[str, dict[int, dict]] = {} |
| for task in TASK_DISPLAY: |
| path = suite_dir / task / "predictions.csv" |
| rows_by_window: dict[int, dict] = {} |
| if not path.exists(): |
| out[task] = rows_by_window |
| continue |
| for row in read_csv(path): |
| if "window_index" not in row: |
| continue |
| idx = int(row["window_index"]) |
| true_value = row.get("true_label") or row.get("true_objects") or row.get("true") or "" |
| pred_value = row.get("predicted_label") or row.get("predicted_objects") or row.get("predicted") or "" |
| if "correct" in row and row["correct"] != "": |
| correct = int(float(row["correct"])) |
| else: |
| correct = int(str(true_value) == str(pred_value)) |
| rows_by_window[idx] = { |
| "true": true_value, |
| "predicted": pred_value, |
| "correct": correct, |
| "confidence": row.get("confidence", ""), |
| } |
| out[task] = rows_by_window |
| return out |
|
|
|
|
| def build_action_segments(windows: list[dict]) -> list[dict]: |
| segments = [] |
| if not windows: |
| return segments |
| current = windows[0]["action_label"] |
| start = int(windows[0]["start_frame"]) |
| start_idx = int(windows[0]["window_index"]) |
| last = windows[0] |
| for row in windows[1:]: |
| if row["action_label"] != current: |
| segments.append({ |
| "action": current, |
| "start_frame": start, |
| "end_frame": int(last["end_frame"]), |
| "start_window": start_idx, |
| "end_window": int(last["window_index"]), |
| }) |
| current = row["action_label"] |
| start = int(row["start_frame"]) |
| start_idx = int(row["window_index"]) |
| last = row |
| segments.append({ |
| "action": current, |
| "start_frame": start, |
| "end_frame": int(last["end_frame"]), |
| "start_window": start_idx, |
| "end_window": int(last["window_index"]), |
| }) |
| return segments |
|
|
|
|
| def build_data(args: argparse.Namespace) -> dict: |
| suite_dir = args.suite_dir |
| diagnostics_dir = args.diagnostics_dir |
| windows = read_csv(suite_dir / "windows.csv") |
| manifest = read_json(suite_dir / "feature_manifest.json") |
| summary = read_json(suite_dir / "summary_report.json") |
| provenance = read_json(diagnostics_dir / "provenance.json") |
| object_rows = {int(r["window_index"]): r for r in read_csv(diagnostics_dir / "object_labels/window_object_labels.csv")} |
| ablation_rows = read_csv(diagnostics_dir / "modality_ablation/ablation_metrics.csv") |
| for row in ablation_rows: |
| task = row.get("task") |
| if task in TASK_DISPLAY_ALL: |
| row["task_display_name"] = TASK_DISPLAY_ALL[task] |
| alignment_rows = read_csv(diagnostics_dir / "alignment_stress/alignment_shift_metrics.csv") |
| timeline_rows = read_csv(diagnostics_dir / "timeline_overlay/timeline_overlay.csv") |
| predictions = load_predictions(suite_dir) |
| X = np.load(suite_dir / "shared_windows.npz")["X"].astype(np.float32) |
|
|
| block_stats = {} |
| block_meta = [] |
| for block in manifest: |
| name = block["name"] |
| start, end = int(block["start"]), int(block["end"]) |
| values = X[:, start:end] |
| l2 = np.linalg.norm(values, axis=1) |
| mean_abs = np.mean(np.abs(values), axis=1) |
| max_l2 = float(max(l2.max(), 1e-8)) |
| block_stats[name] = { |
| "l2": l2, |
| "mean_abs": mean_abs, |
| "relative": l2 / max_l2, |
| } |
| block_meta.append({ |
| "name": name, |
| "display": BLOCK_DISPLAY.get(name, name.replace("_", " ").title()), |
| "modality": block_modality(name), |
| "start": start, |
| "end": end, |
| "dim": int(block["dim"]), |
| }) |
|
|
| explorer_windows = [] |
| for i, row in enumerate(windows): |
| idx = int(row["window_index"]) |
| obj = object_rows.get(idx, {}) |
| feature_stats = [] |
| for block in block_meta: |
| s = block_stats[block["name"]] |
| feature_stats.append({ |
| "name": block["name"], |
| "l2": round(float(s["l2"][i]), 6), |
| "mean_abs": round(float(s["mean_abs"][i]), 6), |
| "relative": round(float(s["relative"][i]), 6), |
| }) |
| task_predictions = {} |
| for task, rows_by_window in predictions.items(): |
| task_predictions[task] = rows_by_window.get(idx) |
| explorer_windows.append({ |
| "window_index": idx, |
| "start_frame": int(row["start_frame"]), |
| "end_frame": int(row["end_frame"]), |
| "center_frame": int(row["center_frame"]), |
| "action": row["action_label"], |
| "subtask": row["subtask_label"], |
| "objects": [x for x in obj.get("objects", "").split("|") if x], |
| "feature_stats": feature_stats, |
| "predictions": task_predictions, |
| }) |
|
|
| best_ablation = {} |
| for task in sorted({r["task"] for r in ablation_rows}): |
| computed = [r for r in ablation_rows if r["task"] == task and r["status"] == "computed" and r["score"]] |
| if not computed: |
| continue |
| best = max(computed, key=lambda r: float(r["score"])) |
| non_overlap = [r for r in computed if r.get("target_source_overlap") == "false"] |
| best_non_overlap = max(non_overlap, key=lambda r: float(r["score"])) if non_overlap else None |
| best_ablation[task] = { |
| "task": task, |
| "task_display_name": TASK_DISPLAY.get(task, task_display_name(task)), |
| "best": { |
| "modality_group": best["modality_group"], |
| "modality_display": best["modality_display"], |
| "score": float(best["score"]), |
| "primary_metric": best["primary_metric"], |
| "target_source_overlap": best["target_source_overlap"], |
| }, |
| "best_non_overlap": None if best_non_overlap is None else { |
| "modality_group": best_non_overlap["modality_group"], |
| "modality_display": best_non_overlap["modality_display"], |
| "score": float(best_non_overlap["score"]), |
| "primary_metric": best_non_overlap["primary_metric"], |
| }, |
| } |
|
|
| return { |
| "meta": { |
| "generated_at": datetime.now(timezone.utc).isoformat(), |
| "window_count": len(explorer_windows), |
| "feature_dim": int(X.shape[1]), |
| "object_label_rows": len(object_rows), |
| "object_vocab_count": len(read_json(diagnostics_dir / "object_labels/object_vocab.json")["vocab"]), |
| "timeline_prediction_rows": len(timeline_rows), |
| "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": any("annotation.hdf5" in key for key in provenance["input_file_hashes"]), |
| "summary": { |
| "num_windows": summary.get("num_windows"), |
| "feature_dim": summary.get("feature_dim"), |
| "window_frames": summary.get("window_frames"), |
| "stride_frames": summary.get("stride_frames"), |
| }, |
| }, |
| "tasks": TASK_DISPLAY, |
| "task_display_names": TASK_DISPLAY_ALL, |
| "feature_blocks": block_meta, |
| "segments": build_action_segments(windows), |
| "windows": explorer_windows, |
| "ablation": { |
| "best_by_task": best_ablation, |
| "rows": ablation_rows, |
| }, |
| "alignment": alignment_rows, |
| } |
|
|
|
|
| HTML_TEMPLATE = """<!doctype html> |
| <html lang="en"> |
| <head> |
| <meta charset="utf-8"> |
| <meta name="viewport" content="width=device-width, initial-scale=1"> |
| <title>Single-Episode Explorer | Ropedia Xperience-10M</title> |
| <meta name="description" content="Interactive window-level explorer for the Ropedia Xperience-10M single-episode diagnostics."> |
| <meta name="theme-color" content="#020502"> |
| <link rel="icon" href="favicon.png" type="image/png" sizes="64x64"> |
| <link rel="preconnect" href="https://fonts.googleapis.com"> |
| <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin> |
| <link href="https://fonts.googleapis.com/css2?family=Inter+Tight:wght@400;500;600;700&family=Space+Grotesk:wght@400;500;600;700&display=swap" rel="stylesheet"> |
| <style> |
| :root { color-scheme: dark; --page:#020502; --panel:#071207; --surface:#0b1709; --ink:#f4f8ef; --muted:#a7b5a3; --line:rgba(204,255,160,.24); --soft:rgba(204,255,160,.14); --green:#ccffa0; --cyan:#7ae5c3; --blue:#9bdfff; --red:#ff8f7a; --amber:#d8f4a5; --card:rgba(5,10,6,.84); --pill:rgba(255,255,255,.05); --font-ui:"Inter Tight",ui-sans-serif,system-ui,-apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif; --font-copy:"Inter Tight",ui-sans-serif,system-ui,-apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif; --font-btn:"Space Grotesk",ui-sans-serif,system-ui,-apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif; --max:1400px; } |
| * { box-sizing:border-box; } |
| body { margin:0; background:var(--page); color:var(--ink); font-family:var(--font-copy); line-height:1.5; font-synthesis-weight:none; } |
| a { color:inherit; } |
| .wrap { width:min(var(--max), calc(100% - 42px)); margin:0 auto; } |
| header { position:sticky; top:0; z-index:10; background:rgba(2,5,2,.92); backdrop-filter:blur(16px); border-bottom:1px solid var(--soft); } |
| .nav { height:64px; display:flex; align-items:center; justify-content:space-between; gap:18px; } |
| .brand { display:flex; gap:11px; align-items:center; text-decoration:none; font-family:var(--font-ui); font-weight:700; } |
| .brand img { width:38px; height:38px; border:1px solid rgba(204,255,160,.42); border-radius:8px; background:#061006; } |
| .nav-links { display:flex; gap:10px; color:#c9d5c5; font-family:var(--font-btn); font-size:14px; } |
| .nav-links a { min-height:36px; display:inline-flex; align-items:center; justify-content:center; padding:0 14px; border:1px solid transparent; border-radius:999px; background:var(--pill); text-decoration:none; } |
| .nav-links a:hover { border-color:var(--green); color:var(--green); background:rgba(255,255,255,.08); } |
| .hero { padding:78px 0 38px; border-bottom:1px solid var(--soft); background:linear-gradient(90deg, rgba(2,5,2,.95) 0%, rgba(2,5,2,.82) 55%, rgba(2,5,2,.70) 100%), linear-gradient(180deg, rgba(2,5,2,.08), rgba(5,6,11,.96)), url("assets/modalities/video.jpg") center right / cover no-repeat; } |
| h1 { max-width:900px; margin:0; font-family:var(--font-ui); font-size:clamp(42px, 6vw, 76px); line-height:.98; letter-spacing:0; } |
| .hero p { max-width:820px; margin:22px 0 0; color:#c7d1c3; font-size:18px; line-height:1.62; } |
| .language-strip { max-width:900px; display:flex; flex-wrap:wrap; gap:8px; margin-top:22px; font-family:var(--font-btn); font-size:12px; } |
| .language-strip a { min-height:32px; display:inline-flex; align-items:center; justify-content:center; padding:0 10px; border:1px solid rgba(204,255,160,.18); border-radius:999px; background:rgba(255,255,255,.05); color:#dfeadc; text-decoration:none; } |
| .language-strip a:hover { color:var(--green); border-color:var(--green); background:rgba(204,255,160,.08); } |
| .stats { display:grid; grid-template-columns:repeat(4,minmax(0,1fr)); gap:10px; margin-top:28px; max-width:900px; } |
| .stat { border:1px solid rgba(204,255,160,.18); border-radius:8px; background:var(--card); padding:13px 14px; } |
| .stat strong { display:block; font-family:var(--font-ui); font-size:24px; line-height:1; } |
| .stat span { display:block; margin-top:6px; color:var(--muted); font-size:12px; } |
| main { padding:26px 0 70px; } |
| .shell { display:grid; grid-template-columns:330px minmax(0,1fr); gap:18px; align-items:start; } |
| .panel { border:1px solid rgba(204,255,160,.18); border-radius:8px; background:var(--card); box-shadow:0 18px 48px rgba(0,0,0,.32); } |
| .side { position:sticky; top:84px; padding:18px; } |
| label { display:block; color:var(--muted); font-size:12px; font-family:var(--font-btn); font-weight:700; margin:14px 0 7px; } |
| input[type=range] { width:100%; accent-color:var(--green); } |
| select, input[type=search] { width:100%; min-height:40px; border:1px solid var(--soft); border-radius:999px; background:#020802; color:var(--ink); padding:9px 12px; font:inherit; } |
| .button-row { display:grid; grid-template-columns:1fr 1fr; gap:8px; margin-top:12px; } |
| button { border:1px solid rgba(255,255,255,.12); border-radius:999px; background:var(--pill); color:var(--ink); min-height:38px; font:700 13px var(--font-btn); cursor:pointer; } |
| button:hover { border-color:var(--green); color:var(--green); background:rgba(255,255,255,.08); } |
| .timeline { padding:18px; margin-bottom:18px; } |
| .timeline-strip { position:relative; height:60px; border:1px solid var(--soft); border-radius:8px; overflow:hidden; background:#030803; } |
| .segment { position:absolute; top:0; bottom:0; border-right:1px solid rgba(2,5,2,.45); opacity:.92; } |
| .marker { position:absolute; top:0; bottom:0; width:3px; background:var(--ink); box-shadow:0 0 0 2px rgba(2,5,2,.9), 0 0 18px rgba(204, 255, 160,.6); } |
| .timeline-meta { display:flex; justify-content:space-between; gap:12px; margin-top:10px; color:var(--muted); font-size:12px; } |
| .content { display:grid; gap:18px; } |
| .window-panel { padding:22px; } |
| .window-head { display:grid; grid-template-columns:minmax(0,1fr) auto; gap:18px; align-items:start; border-bottom:1px solid var(--soft); padding-bottom:18px; } |
| h2 { margin:0; font-family:var(--font-ui); font-size:30px; line-height:1.1; } |
| .frame-pill { border:1px solid var(--line); border-radius:999px; padding:8px 10px; color:var(--green); font-family:var(--font-btn); font-size:13px; font-weight:700; white-space:nowrap; background:var(--pill); } |
| .subtle { color:var(--muted); } |
| .chips { display:flex; flex-wrap:wrap; gap:7px; margin-top:12px; } |
| .chip { border:1px solid var(--soft); background:var(--pill); color:#e7f2df; border-radius:999px; padding:5px 8px; font-size:12px; } |
| .grid { display:grid; gap:12px; } |
| .pred-grid { grid-template-columns:repeat(3,minmax(0,1fr)); margin-top:18px; } |
| .pred { border:1px solid rgba(204,255,160,.18); border-radius:8px; background:rgba(5,10,6,.72); padding:13px; min-height:118px; } |
| .pred h3 { margin:0 0 8px; font-family:var(--font-ui); font-size:15px; } |
| .pred p { margin:4px 0; color:#cdd8c8; font-size:13px; overflow-wrap:anywhere; } |
| .pred .ok { color:var(--green); font-weight:700; } |
| .pred .bad { color:var(--red); font-weight:700; } |
| .feature-grid { grid-template-columns:repeat(2,minmax(0,1fr)); } |
| .feature { display:grid; grid-template-columns:130px 1fr 62px; gap:10px; align-items:center; border-bottom:1px solid var(--soft); padding:10px 0; } |
| .feature:last-child { border-bottom:0; } |
| .feature-name { font-family:var(--font-ui); font-size:13px; color:#edf6e8; } |
| .bar { height:10px; border-radius:999px; background:rgba(204,255,160,.16); overflow:hidden; } |
| .bar span { display:block; height:100%; width:calc(var(--w) * 1%); background:linear-gradient(90deg,var(--green),rgba(204,255,160,.45)); } |
| .num { text-align:right; color:var(--muted); font-size:12px; font-variant-numeric:tabular-nums; } |
| .analysis-grid { display:grid; grid-template-columns:1fr 1fr; gap:18px; } |
| .analysis { padding:18px; } |
| .analysis h3 { margin:0 0 12px; font-family:var(--font-ui); font-size:20px; } |
| .rows { display:grid; gap:8px; } |
| .row { display:grid; grid-template-columns:1fr auto; gap:12px; border-bottom:1px solid var(--soft); padding:8px 0; color:#d8e4d3; font-size:13px; } |
| .row strong { color:var(--green); font-variant-numeric:tabular-nums; } |
| .note { margin-top:12px; color:var(--muted); font-size:12px; line-height:1.55; } |
| @media (max-width: 980px) { .shell,.analysis-grid { grid-template-columns:1fr; } .side { position:static; } .pred-grid,.feature-grid,.stats { grid-template-columns:1fr; } .window-head { grid-template-columns:1fr; } .nav-links { display:none; } } |
| </style> |
| </head> |
| <body> |
| <header> |
| <div class="wrap nav"> |
| <a class="brand" href="index.html"><img src="assets/brand/xperience10m-logo-mark-192.png" alt=""><span>Ropedia Xperience-10M</span></a> |
| <nav class="nav-links"><a href="index.html">Project</a><a href="index.html#raw-sample">Raw Sample</a><a href="single_episode_explorer.html">Explorer</a><a href="data/language_versions.json">Languages</a><a href="data/single_episode_explorer.json">Data JSON</a></nav> |
| </div> |
| </header> |
| <section class="hero"> |
| <div class="wrap"> |
| <h1>Single-Episode Research Explorer</h1> |
| <p>Inspect the exported Xperience-10M sample windows, real object labels, model predictions, feature-block statistics, and diagnostic scores from one aligned episode.</p> |
| <div class="language-strip" aria-label="README language versions"> |
| <a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/blob/main/README.md">English</a> |
| <a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/blob/main/README.zh.md">中文</a> |
| <a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/blob/main/README.es.md">Español</a> |
| <a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/blob/main/README.fr.md">Français</a> |
| <a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/blob/main/README.de.md">Deutsch</a> |
| <a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/blob/main/README.ja.md">日本語</a> |
| <a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/blob/main/README.ko.md">한국어</a> |
| <a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/blob/main/README.pt.md">Português</a> |
| <a href="data/language_versions.json">Manifest</a> |
| </div> |
| <div class="stats"> |
| <div class="stat"><strong id="statWindows">-</strong><span>windows</span></div> |
| <div class="stat"><strong id="statDim">-</strong><span>feature dimensions</span></div> |
| <div class="stat"><strong id="statObjects">-</strong><span>object labels</span></div> |
| <div class="stat"><strong id="statPreds">-</strong><span>prediction rows</span></div> |
| </div> |
| </div> |
| </section> |
| <main> |
| <div class="wrap shell"> |
| <aside class="panel side"> |
| <label for="windowRange">Window</label> |
| <input id="windowRange" type="range" min="0" max="0" value="0"> |
| <div class="button-row"><button id="prevWindow" type="button">Previous</button><button id="nextWindow" type="button">Next</button></div> |
| <label for="taskSelect">Task Focus</label> |
| <select id="taskSelect"></select> |
| <label for="searchBox">Search Action or Object</label> |
| <input id="searchBox" type="search" placeholder="e.g. Pour coffee, kettle"> |
| <div class="button-row"><button id="firstMatch" type="button">First Match</button><button id="firstPred" type="button">First Predicted</button></div> |
| <p class="note">This explorer embeds window-level exported artifacts. Open the Raw Sample page to play browser-preview clips, access the full raw MP4/HDF5/RRD source files, and inspect the HDF5/RRD file organization.</p> |
| </aside> |
| <section class="content"> |
| <div class="panel timeline"> |
| <div class="timeline-strip" id="timelineStrip"></div> |
| <div class="timeline-meta"><span id="timelineLeft"></span><span id="timelineRight"></span></div> |
| </div> |
| <section class="panel window-panel"> |
| <div class="window-head"> |
| <div> |
| <h2 id="windowTitle">Window</h2> |
| <p id="windowSubtitle" class="subtle"></p> |
| <div class="chips" id="objectChips"></div> |
| </div> |
| <div class="frame-pill" id="framePill"></div> |
| </div> |
| <div class="grid pred-grid" id="predictionGrid"></div> |
| </section> |
| <section class="analysis-grid"> |
| <div class="panel analysis"> |
| <h3>Feature Blocks</h3> |
| <div class="grid feature-grid" id="featureGrid"></div> |
| </div> |
| <div class="panel analysis"> |
| <h3>Diagnostics</h3> |
| <div class="rows" id="diagnosticRows"></div> |
| <p class="note" id="diagnosticNote"></p> |
| </div> |
| </section> |
| </section> |
| </div> |
| </main> |
| <script id="explorer-data" type="application/json">__DATA__</script> |
| <script> |
| const DATA = JSON.parse(document.getElementById("explorer-data").textContent); |
| function hasPrediction(windowRecord, taskKey) { |
| return taskKey === "all" ? Object.values(windowRecord.predictions).some(Boolean) : Boolean(windowRecord.predictions[taskKey]); |
| } |
| function defaultWindowIndex() { |
| let best = 0; |
| let bestCount = -1; |
| DATA.windows.forEach((w) => { |
| const count = Object.values(w.predictions).filter(Boolean).length; |
| if (count > bestCount) { best = w.window_index; bestCount = count; } |
| }); |
| return best; |
| } |
| const state = { index: defaultWindowIndex(), task: "all" }; |
| const range = document.getElementById("windowRange"); |
| const taskSelect = document.getElementById("taskSelect"); |
| const searchBox = document.getElementById("searchBox"); |
| const colors = ["#5ccf7d", "#7ae5c3", "#9bdfff", "#d8f4a5", "#f0a45e", "#cba8ff", "#ff8f7a"]; |
| document.getElementById("statWindows").textContent = DATA.meta.window_count; |
| document.getElementById("statDim").textContent = DATA.meta.feature_dim; |
| document.getElementById("statObjects").textContent = DATA.meta.object_vocab_count; |
| document.getElementById("statPreds").textContent = DATA.meta.timeline_prediction_rows; |
| range.max = DATA.windows.length - 1; |
| for (const [key, label] of Object.entries(DATA.tasks)) { |
| const option = document.createElement("option"); |
| option.value = key; |
| option.textContent = label; |
| taskSelect.appendChild(option); |
| } |
| const allOption = document.createElement("option"); |
| allOption.value = "all"; |
| allOption.textContent = "All Prediction Cards"; |
| taskSelect.insertBefore(allOption, taskSelect.firstChild); |
| taskSelect.value = state.task; |
| function pct(value, min, max) { return ((value - min) / Math.max(1, max - min)) * 100; } |
| function splitObjects(value) { return String(value || "").split("|").filter(Boolean); } |
| function renderTimeline() { |
| const strip = document.getElementById("timelineStrip"); |
| strip.innerHTML = ""; |
| const minFrame = DATA.windows[0].start_frame; |
| const maxFrame = DATA.windows[DATA.windows.length - 1].end_frame; |
| DATA.segments.forEach((seg, i) => { |
| const el = document.createElement("div"); |
| el.className = "segment"; |
| el.style.left = pct(seg.start_frame, minFrame, maxFrame) + "%"; |
| el.style.width = Math.max(0.3, pct(seg.end_frame, minFrame, maxFrame) - pct(seg.start_frame, minFrame, maxFrame)) + "%"; |
| el.style.background = colors[i % colors.length]; |
| el.title = `${seg.action} (${seg.start_frame}-${seg.end_frame})`; |
| el.addEventListener("click", () => { state.index = seg.start_window; render(); }); |
| strip.appendChild(el); |
| }); |
| const marker = document.createElement("div"); |
| marker.className = "marker"; |
| marker.style.left = pct(DATA.windows[state.index].center_frame, minFrame, maxFrame) + "%"; |
| strip.appendChild(marker); |
| document.getElementById("timelineLeft").textContent = `frame ${minFrame}`; |
| document.getElementById("timelineRight").textContent = `frame ${maxFrame}`; |
| } |
| function renderPredictions(w) { |
| const grid = document.getElementById("predictionGrid"); |
| grid.innerHTML = ""; |
| const taskEntries = Object.entries(DATA.tasks).filter(([key]) => state.task === "all" || key === state.task); |
| for (const [key, label] of taskEntries) { |
| const pred = w.predictions[key]; |
| const card = document.createElement("article"); |
| card.className = "pred"; |
| let body = ""; |
| if (!pred) { |
| body = `<p class="subtle">No held-out prediction row for this window.</p>`; |
| } else { |
| const status = pred.correct ? `<span class="ok">correct</span>` : `<span class="bad">mismatch</span>`; |
| body = `<p>${status}</p><p><strong>true</strong>: ${escapeHtml(pred.true || "")}</p><p><strong>pred</strong>: ${escapeHtml(pred.predicted || "")}</p>`; |
| if (pred.confidence) body += `<p><strong>confidence</strong>: ${Number(pred.confidence).toFixed(3)}</p>`; |
| } |
| card.innerHTML = `<h3>${escapeHtml(label)}</h3>${body}`; |
| grid.appendChild(card); |
| } |
| } |
| function renderFeatures(w) { |
| const grid = document.getElementById("featureGrid"); |
| grid.innerHTML = ""; |
| for (const stat of w.feature_stats) { |
| const block = DATA.feature_blocks.find((b) => b.name === stat.name); |
| const row = document.createElement("div"); |
| row.className = "feature"; |
| row.innerHTML = `<span class="feature-name">${escapeHtml(block.display)}</span><span class="bar"><span style="--w:${Math.round(stat.relative * 100)}"></span></span><span class="num">${stat.l2.toFixed(2)}</span>`; |
| grid.appendChild(row); |
| } |
| } |
| function renderDiagnostics() { |
| const rows = document.getElementById("diagnosticRows"); |
| rows.innerHTML = ""; |
| const task = state.task === "all" ? "object_relevance" : state.task; |
| const diag = DATA.ablation.best_by_task[task]; |
| if (diag) { |
| rows.innerHTML += `<div class="row"><span>Best modality for ${escapeHtml(DATA.tasks[task] || task)}</span><strong>${escapeHtml(diag.best.modality_display)} ${diag.best.score.toFixed(3)}</strong></div>`; |
| if (diag.best_non_overlap) rows.innerHTML += `<div class="row"><span>Best non-overlap modality</span><strong>${escapeHtml(diag.best_non_overlap.modality_display)} ${diag.best_non_overlap.score.toFixed(3)}</strong></div>`; |
| } |
| const zeroRows = DATA.alignment.filter((r) => Number(r.shift_windows) === 0); |
| zeroRows.slice(0, 5).forEach((r) => { |
| rows.innerHTML += `<div class="row"><span>${escapeHtml(r.query_display)} zero-shift retrieval MRR</span><strong>${Number(r.mrr).toFixed(3)}</strong></div>`; |
| }); |
| document.getElementById("diagnosticNote").textContent = DATA.meta.source_policy; |
| } |
| function renderWindow() { |
| const w = DATA.windows[state.index]; |
| range.value = state.index; |
| document.getElementById("windowTitle").textContent = `Window ${w.window_index}: ${w.action || "unlabeled action"}`; |
| document.getElementById("windowSubtitle").textContent = w.subtask || "No subtask label"; |
| document.getElementById("framePill").textContent = `frames ${w.start_frame}-${w.end_frame}`; |
| const chips = document.getElementById("objectChips"); |
| chips.innerHTML = ""; |
| (w.objects.length ? w.objects : ["no object label"]).forEach((obj) => { |
| const chip = document.createElement("span"); |
| chip.className = "chip"; |
| chip.textContent = obj; |
| chips.appendChild(chip); |
| }); |
| renderPredictions(w); |
| renderFeatures(w); |
| renderDiagnostics(); |
| } |
| function render() { renderTimeline(); renderWindow(); } |
| function escapeHtml(s) { return String(s).replace(/[&<>"']/g, (c) => ({ "&":"&", "<":"<", ">":">", '"':""", "'":"'" }[c])); } |
| range.addEventListener("input", () => { state.index = Number(range.value); render(); }); |
| taskSelect.addEventListener("change", () => { state.task = taskSelect.value; render(); }); |
| document.getElementById("prevWindow").addEventListener("click", () => { state.index = Math.max(0, state.index - 1); render(); }); |
| document.getElementById("nextWindow").addEventListener("click", () => { state.index = Math.min(DATA.windows.length - 1, state.index + 1); render(); }); |
| document.getElementById("firstPred").addEventListener("click", () => { |
| const found = DATA.windows.find((w) => hasPrediction(w, state.task)); |
| if (found) { state.index = found.window_index; render(); } |
| }); |
| document.getElementById("firstMatch").addEventListener("click", () => { |
| const q = searchBox.value.trim().toLowerCase(); |
| if (!q) return; |
| const found = DATA.windows.find((w) => [w.action, w.subtask, ...w.objects].join(" ").toLowerCase().includes(q)); |
| if (found) { state.index = found.window_index; render(); } |
| }); |
| render(); |
| </script> |
| </body> |
| </html> |
| """ |
|
|
|
|
| def write_html(path: Path, data: dict) -> None: |
| path.parent.mkdir(parents=True, exist_ok=True) |
| payload = json.dumps(data, ensure_ascii=False).replace("</script", "<\\/script") |
| path.write_text(HTML_TEMPLATE.replace("__DATA__", payload), encoding="utf-8") |
|
|
|
|
| def main() -> None: |
| args = parse_args() |
| data = build_data(args) |
| write_json(args.docs_dir / "data/single_episode_explorer.json", data) |
| write_html(args.docs_dir / "single_episode_explorer.html", data) |
| print(f"Wrote {args.docs_dir / 'data/single_episode_explorer.json'}") |
| print(f"Wrote {args.docs_dir / 'single_episode_explorer.html'}") |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
