Add HF dataset viewer metadata

.gitignore

@@ -0,0 +1 @@
+.DS_Store

README.md

@@ -1,16 +1,22 @@
----
-license: mit
----
+---
+license: mit
+configs:
+  - config_name: default
+    data_dir: .
+    default: true
+---
 
 # AVGen-Bench Generated Videos Data Card
 
 ## Overview
 
-This data card describes the generated audio-video outputs organized under `generated_videos/` in AVGen-Bench.
+This data card describes the generated audio-video outputs stored directly in the repository root by model directory.
 
 The collection is intended for **benchmarking and qualitative/quantitative evaluation** of text-to-audio-video (T2AV) systems. It is not a training dataset. Each item is a model-generated video produced from a prompt defined in `prompts/*.json`.
 
-Code repository: https://github.com/microsoft/AVGen-Bench
+Code repository: https://github.com/microsoft/AVGen-Bench
+
+For Hugging Face Hub compatibility, the repository includes a root-level `metadata.parquet` file so the Dataset Viewer can expose each video as a structured row with prompt metadata instead of treating the repo as an unindexed file dump.
 
 ## What This Dataset Contains
 
@@ -23,28 +29,28 @@ The dataset is organized by:
 A typical top-level structure is:
 
 ```text
-AVGen-Bench/
-├── generated_videos/
-│   ├── Kling_2.6/
-│   ├── LTX-2/
-│   ├── LTX-2.3/
-│   ├── MOVA_360p_Emu3.5/
-│   ├── MOVA_360p_NanoBanana_2/
-│   ├── Ovi_11/
-│   ├── Seedance_1.5_pro/
-│   ├── Sora_2/
-│   ├── Veo_3.1_fast/
-│   ├── Veo_3.1_quality/
-│   ├── Wan_2.2_HunyuanVideo-Foley/
-│   └── Wan_2.6/
-├── prompts/
-└── generated_images/   # optional, depending on generation pipeline
+AVGen-Bench/
+├── Kling_2.6/
+├── LTX-2/
+├── LTX-2.3/
+├── MOVA_360p_Emu3.5/
+├── MOVA_360p_NanoBanana_2/
+├── Ovi_11/
+├── Seedance_1.5_pro/
+├── Sora_2/
+├── Veo_3.1_fast/
+├── Veo_3.1_quality/
+├── Wan_2.2_HunyuanVideo-Foley/
+├── Wan_2.6/
+├── metadata.parquet
+├── prompts/
+└── reference_image/   # optional, depending on generation pipeline
 ```
 
 Within each model directory, videos are grouped by category, for example:
 
 ```text
-generated_videos/veo3.1_fast/
+Veo_3.1_fast/
 ├── ads/
 ├── animals/
 ├── asmr/
@@ -90,7 +96,7 @@ Each generated item is typically:
 
 - A single `.mp4` file
 - Containing model-generated video and, when supported by the model/pipeline, synthesized audio
-- Stored under `generated_videos/<model>/<category>/`
+- Stored under `<model>/<category>/`
 
 The filename is usually derived from prompt content after sanitization. Exact naming may vary by generation script or provider wrapper.
 In the standard export pipeline, the filename is derived from the prompt's `content` field using the following logic:
@@ -109,9 +115,18 @@ def safe_filename(name: str, max_len: int = 180) -> str:
 
 So the expected output path pattern is:
 
-```text
-generated_videos/<model>/<category>/<safe_filename(content)>.mp4
-```
+```text
+<model>/<category>/<safe_filename(content)>.mp4
+```
+
+For Dataset Viewer indexing, `metadata.parquet` stores one row per exported video with:
+
+- `file_name`: relative path to the `.mp4`
+- `model`: model directory name
+- `category`: benchmark category
+- `content`: prompt short name
+- `prompt`: full generation prompt
+- `prompt_id`: index inside `prompts/<category>.json`
 
 ## How The Data Was Produced
 

metadata.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:236f73fb879f176c4999a60838f53d1dae7a8904b070344d60802f58ecb449c9
+size 112572

scripts/build_hf_metadata.py

@@ -0,0 +1,123 @@
+#!/usr/bin/env python3
+
+from __future__ import annotations
+
+import argparse
+import json
+import re
+import unicodedata
+from collections import defaultdict
+from pathlib import Path
+
+import pyarrow as pa
+import pyarrow.parquet as pq
+
+
+INVALID_FILENAME_CHARS = r'[/\\:*?"<>|\n\r\t]'
+
+
+def normalize_text(value: str) -> str:
+    return unicodedata.normalize("NFC", str(value).strip())
+
+
+def safe_filename(name: str, max_len: int = 180) -> str:
+    name = normalize_text(name)
+    name = re.sub(INVALID_FILENAME_CHARS, "_", name)
+    name = re.sub(r"\s+", " ", name).strip()
+    if not name:
+        name = "untitled"
+    if len(name) > max_len:
+        name = name[:max_len].rstrip()
+    return name
+
+
+def load_prompt_index(prompts_dir: Path) -> dict[str, dict[str, dict[str, object]]]:
+    prompt_index: dict[str, dict[str, dict[str, object]]] = defaultdict(dict)
+    duplicates: list[tuple[str, str]] = []
+
+    for prompt_file in sorted(prompts_dir.glob("*.json")):
+        category = prompt_file.stem
+        records = json.loads(prompt_file.read_text(encoding="utf-8"))
+        for prompt_id, record in enumerate(records):
+            content = normalize_text(record["content"])
+            key = safe_filename(content)
+            if key in prompt_index[category]:
+                duplicates.append((category, key))
+            prompt_index[category][key] = {
+                "content": content,
+                "prompt": normalize_text(record["prompt"]),
+                "prompt_id": prompt_id,
+            }
+
+    if duplicates:
+        duplicate_lines = "\n".join(f"- {category}: {key}" for category, key in duplicates[:20])
+        raise ValueError(f"Duplicate prompt filenames detected:\n{duplicate_lines}")
+
+    return prompt_index
+
+
+def build_rows(repo_root: Path) -> list[dict[str, object]]:
+    prompt_index = load_prompt_index(repo_root / "prompts")
+    rows: list[dict[str, object]] = []
+
+    for video_path in sorted(repo_root.glob("*/*/*.mp4")):
+        model = video_path.parts[-3]
+        category = video_path.parts[-2]
+        filename = normalize_text(video_path.stem)
+        prompt_record = prompt_index.get(category, {}).get(filename)
+        rel_path = video_path.relative_to(repo_root).as_posix()
+
+        rows.append(
+            {
+                "file_name": rel_path,
+                "model": model,
+                "category": category,
+                "content": prompt_record["content"] if prompt_record else filename,
+                "prompt": prompt_record["prompt"] if prompt_record else None,
+                "prompt_id": prompt_record["prompt_id"] if prompt_record else None,
+                "matched_prompt": prompt_record is not None,
+            }
+        )
+
+    return rows
+
+
+def write_parquet(rows: list[dict[str, object]], output_path: Path) -> None:
+    table = pa.Table.from_pylist(rows)
+    pq.write_table(table, output_path)
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        description="Build metadata.parquet for Hugging Face Dataset Viewer."
+    )
+    parser.add_argument(
+        "--repo-root",
+        type=Path,
+        default=Path(__file__).resolve().parents[1],
+        help="Path to the dataset repository root.",
+    )
+    parser.add_argument(
+        "--output",
+        type=Path,
+        default=None,
+        help="Output parquet path. Defaults to <repo-root>/metadata.parquet.",
+    )
+    args = parser.parse_args()
+
+    repo_root = args.repo_root.resolve()
+    output_path = args.output.resolve() if args.output else repo_root / "metadata.parquet"
+
+    rows = build_rows(repo_root)
+    if not rows:
+        raise ValueError(f"No MP4 files found under {repo_root}")
+
+    write_parquet(rows, output_path)
+
+    matched = sum(1 for row in rows if row["matched_prompt"])
+    print(f"Wrote {len(rows)} rows to {output_path}")
+    print(f"Matched prompts: {matched}/{len(rows)}")
+
+
+if __name__ == "__main__":
+    main()