Restructure as HF Dataset repository

- Remove Gradio app.py, requirements.txt, scripts
- Add comprehensive README.md (Dataset Card)
- Add .gitattributes
- Keep test query JSON files

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

.gitattributes

@@ -0,0 +1 @@
+*.json text eol=lf

README.md

@@ -0,0 +1,179 @@
+---
+tags:
+  - image-retrieval
+  - benchmark
+  - photobench
+  - vision-language
+license: mit
+size_categories:
+  - n<1K
+task_categories:
+  - image-retrieval
+language:
+  - en
+  - zh
+---
+
+# PhotoBench-Full
+
+PhotoBench-Full provides the **complete raw image collection** and **test queries** for open-ended image retrieval research.
+
+> **Leaderboard:** [PhotoBench-Full Leaderboard](https://huggingface.co/spaces/SorrowTea/PhotoBench-Full-LearderBoard)
+
+---
+
+## Dataset Description
+
+PhotoBench is an image retrieval benchmark with open-ended natural language queries.
+Unlike the protected version, PhotoBench-Full gives you unrestricted access to the raw images, allowing you to use your own embedding models, caption generators, or agent-based retrieval workflows.
+
+This dataset contains:
+
+- **Test queries** for 3 albums (English + Chinese)
+- **Raw images** for all 3 albums (available upon request; not included in this repository due to size)
+
+| Album | Images | Test Queries | Avg. GT per Query |
+|-------|--------|--------------|-------------------|
+| 1     | ~1,070 | 100          | ~4.5              |
+| 2     | ~1,470 | 100          | ~4.5              |
+| 3     | ~1,050 | 100          | ~4.5              |
+
+---
+
+## Data Format
+
+### Test Queries (`albumN_test.json`)
+
+Each file is a JSON array of query objects:
+
+```json
+[
+  {
+    "query_cn": "摆满的书桌",
+    "query_en": "cluttered desk"
+  },
+  {
+    "query_cn": "紫毛衣女孩",
+    "query_en": "girl in purple sweater"
+  }
+]
+```
+
+| Field      | Type   | Description                              |
+|------------|--------|------------------------------------------|
+| `query_cn` | string | Query text in Chinese                    |
+| `query_en` | string | Query text in English (primary language) |
+
+### Raw Images
+
+The raw images (`album1/`, `album2/`, `album3/`) contain the full-resolution original photos.
+
+**Total size:** ~11 GB  
+**Format:** JPEG  
+**Naming:** Original camera filenames (e.g., `IMG_1234.JPG`, `FullSizeRender.JPG`)
+
+> Raw images are not hosted in this repository due to size constraints. Please contact the authors or use the download instructions below.
+
+---
+
+## How to Use
+
+### 1. Download Test Queries
+
+The test query JSON files are available directly in this repository:
+
+```bash
+# Via huggingface_hub CLI
+huggingface-cli download SorrowTea/PhotoBench-Full-HF --repo-type dataset --local-dir ./photobench-full
+```
+
+Or browse and download individual files from the **Files** tab above.
+
+### 2. Download Raw Images
+
+Raw images are distributed separately. Contact the authors for access, or prepare the images according to the album structure:
+
+```
+raw_albums/
+├── album1/
+│   ├── IMG_0001.JPG
+│   ├── IMG_0002.JPG
+│   └── ...
+├── album2/
+└── album3/
+```
+
+### 3. Build Your Retrieval System
+
+With the raw images and test queries, you can:
+
+- Extract image embeddings with any vision encoder (CLIP, SigLIP, etc.)
+- Generate captions with any VLM (GPT-4V, Qwen-VL, etc.)
+- Design multi-step agent workflows
+- Evaluate with your own metrics
+
+### 4. Submit to Leaderboard
+
+Prepare your predictions in the submission format and upload to the [PhotoBench-Full Leaderboard](https://huggingface.co/spaces/SorrowTea/PhotoBench-Full-LearderBoard).
+
+**Submission format:**
+
+```json
+[
+  {
+    "album_id": "1",
+    "query_en": "cluttered desk",
+    "pred": ["IMG_1234.JPG", "IMG_5678.JPG", "IMG_9012.JPG"]
+  }
+]
+```
+
+Requirements:
+- `album_id`: "1", "2", or "3"
+- `query_en`: Must match the test query exactly
+- `pred`: Ordered list of predicted image filenames
+
+---
+
+## Evaluation
+
+The leaderboard computes the following metrics:
+
+| Metric     | Description                                        |
+|------------|----------------------------------------------------|
+| Recall@k   | Proportion of ground-truth images in top-k         |
+| NDCG@k     | Normalized Discounted Cumulative Gain at rank k    |
+
+Supported k values: **1, 5, 10, 20, 50, 100**
+
+Results are averaged per album, then averaged across albums for the final score.
+
+Only **full submissions** (all 3 albums, all queries) are eligible for public leaderboard ranking.
+
+---
+
+## Citation
+
+If you use PhotoBench-Full in your research, please cite:
+
+```bibtex
+@misc{photobench2026,
+  title={PhotoBench},
+  year={2026},
+  eprint={2603.01493},
+  archivePrefix={arXiv},
+  primaryClass={cs.CV}
+}
+```
+
+---
+
+## License
+
+This dataset is released under the MIT License.
+
+---
+
+## Contact
+
+For questions or data access requests, please open an issue on this repository or contact the authors.

app.py

@@ -1,257 +0,0 @@
-import json
-from pathlib import Path
-
-import gradio as gr
-
-DATA_DIR = Path(__file__).parent / "data"
-TEST_DIR = DATA_DIR / "test"
-IMAGES_DIR = DATA_DIR / "images"
-PERSISTENT_IMAGES = Path("/data/images")
-
-LEADERBOARD_URL = "https://huggingface.co/spaces/SorrowTea/PhotoBench-Full-LearderBoard"
-
-TITLE = """
-<div style="text-align:center; padding: 48px 20px; background: linear-gradient(160deg, #faf8f5 0%, #f5f0eb 100%); border-radius: 20px; margin-bottom: 32px; border: 1px solid #e8e0d8;">
-    <h1 style="color:#1a1a1a; font-size:3em; font-weight:600; letter-spacing:-1px; margin:0; font-family:-apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;">
-        PhotoBench-Full
-    </h1>
-    <p style="color:#F37021; font-size:1.1em; font-weight:500; letter-spacing:2px; margin:12px 0 0 0;">
-        DATASET DOWNLOAD
-    </p>
-    <div style="width:60px; height:3px; background:#F37021; margin:20px auto; border-radius:2px;"></div>
-    <p style="color:#666; font-size:0.95em; margin-top:12px;">
-        Open-ended Image Retrieval Benchmark — Raw Images & Test Queries
-    </p>
-</div>
-"""
-
-INTRO = """
-<div style="text-align:center; max-width:720px; margin:0 auto 40px; color:#444; line-height:1.8;">
-
-PhotoBench-Full provides the <strong>complete raw image collection</strong> and <strong>test queries</strong> for unrestricted image retrieval research.
-
-<p style="margin-top:16px;">
-After downloading the data, you can develop your own retrieval approach and submit results to the
-<a href="{url}" target="_blank" style="color:#F37021; font-weight:600; text-decoration:none;">PhotoBench-Full Leaderboard ↗</a>.
-</p>
-
-</div>
-""".format(url=LEADERBOARD_URL)
-
-custom_css = """
-/* Apple-inspired modern clean theme */
-body {
-    background: #faf8f5 !important;
-    font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, 'Helvetica Neue', Arial, sans-serif !important;
-    font-size: 17px !important;
-}
-
-/* Primary buttons */
-.gr-button-primary {
-    background: #F37021 !important;
-    border: none !important;
-    border-radius: 10px !important;
-    color: #fff !important;
-    font-weight: 600 !important;
-    font-size: 0.95em !important;
-    padding: 12px 28px !important;
-    transition: all 0.25s ease !important;
-}
-
-.gr-button-primary:hover {
-    background: #e5631a !important;
-    transform: translateY(-1px) !important;
-    box-shadow: 0 6px 20px rgba(243,112,33,0.25) !important;
-}
-
-/* Markdown */
-.markdown-text {
-    max-width: 780px;
-    margin: 0 auto;
-    color: #333;
-    line-height: 1.8;
-    font-size: 1.05em;
-}
-
-/* Download cards */
-.download-card {
-    background: #fff;
-    border-radius: 14px;
-    border: 1px solid #e8e0d8;
-    padding: 24px;
-    margin-bottom: 16px;
-    box-shadow: 0 2px 16px rgba(0,0,0,0.04);
-}
-
-.download-card h3 {
-    margin-top: 0;
-    color: #1a1a1a;
-    font-weight: 600;
-    font-size: 1.1em;
-}
-
-.download-card p {
-    color: #666;
-    font-size: 0.95em;
-    margin: 8px 0 16px;
-}
-
-/* Form containers */
-.gr-form .gr-box {
-    border-radius: 14px !important;
-    background: #fff !important;
-    border: 1px solid #e8e0d8 !important;
-    padding: 24px !important;
-}
-
-/* Labels */
-.gr-input-label, .gr-dropdown-label {
-    font-weight: 500 !important;
-    color: #444 !important;
-    font-size: 0.9em !important;
-    margin-bottom: 6px !important;
-}
-"""
-
-
-def get_test_files():
-    """Return list of available test JSON files."""
-    files = []
-    if TEST_DIR.exists():
-        for f in sorted(TEST_DIR.glob("album*_test.json")):
-            files.append(str(f))
-    return files
-
-
-def get_image_archives():
-    """Return list of available image zip archives."""
-    archives = []
-    for d in [IMAGES_DIR, PERSISTENT_IMAGES]:
-        if d.exists():
-            for f in sorted(d.glob("album*.zip")):
-                archives.append(str(f))
-    return archives
-
-
-def load_test_preview(filepath):
-    """Load first few entries of a test JSON for preview."""
-    try:
-        with open(filepath, "r", encoding="utf-8") as f:
-            data = json.load(f)
-        return data[:5]
-    except Exception as e:
-        return {"error": str(e)}
-
-
-# Build UI
-with gr.Blocks(css=custom_css, title="PhotoBench-Full Download") as demo:
-    gr.HTML(TITLE)
-    gr.Markdown(INTRO, elem_classes="markdown-text")
-
-    # ---- Test Queries Section ----
-    gr.Markdown("""
-    ### Test Queries
-
-    Each album contains English and Chinese test queries. Download the JSON files to build your retrieval system.
-    """)
-
-    test_files = get_test_files()
-    if test_files:
-        for tf in test_files:
-            name = Path(tf).name
-            album_id = name.replace("album", "").replace("_test.json", "")
-            with gr.Row():
-                with gr.Column(scale=1):
-                    pass
-                with gr.Column(scale=3):
-                    with gr.Row():
-                        with gr.Column():
-                            gr.Markdown(f"""
-                            <div class="download-card">
-                                <h3>Album {album_id}</h3>
-                                <p>Test queries ({name}) — contains query_en and query_cn for each test case.</p>
-                            </div>
-                            """)
-                        with gr.Column():
-                            gr.File(value=tf, label=f"Download {name}", interactive=False)
-                with gr.Column(scale=1):
-                    pass
-    else:
-        gr.Markdown("*Test query files not found. Please ensure data/test/album*_test.json exist.*")
-
-    # ---- Images Section ----
-    gr.Markdown("""
-    ### Raw Images
-
-    The raw image collections for all 3 albums. These are the original photos used in the benchmark.
-    """)
-
-    image_archives = get_image_archives()
-    if image_archives:
-        for af in image_archives:
-            name = Path(af).name
-            album_id = name.replace("album", "").replace(".zip", "")
-            with gr.Row():
-                with gr.Column(scale=1):
-                    pass
-                with gr.Column(scale=3):
-                    with gr.Row():
-                        with gr.Column():
-                            size_mb = Path(af).stat().st_size / (1024 * 1024)
-                            gr.Markdown(f"""
-                            <div class="download-card">
-                                <h3>Album {album_id} Images</h3>
-                                <p>Raw image collection ({name}) — {size_mb:.1f} MB</p>
-                            </div>
-                            """)
-                        with gr.Column():
-                            gr.File(value=af, label=f"Download {name}", interactive=False)
-                with gr.Column(scale=1):
-                    pass
-    else:
-        gr.Markdown("""
-        <div style="text-align:center; padding: 32px; background:#fff; border-radius:14px; border:1px solid #e8e0d8; max-width:720px; margin:0 auto;">
-            <p style="color:#666; font-size:1em; line-height:1.8;">
-                Raw image archives are not yet available on this Space.<br>
-                Total size: ~11 GB (Album 1: ~5.1 GB, Album 2: ~3.4 GB, Album 3: ~2.4 GB).<br>
-                Please contact the authors for access or check back later.
-            </p>
-        </div>
-        """)
-
-    # ---- Data Format Reference ----
-    gr.Markdown("""
-    ### Data Format
-
-    **Test Query JSON** (`albumN_test.json`):
-    ```json
-    [
-      {
-        "query_cn": "摆满的书桌",
-        "query_en": "cluttered desk"
-      }
-    ]
-    ```
-
-    **Submission Format** for the Leaderboard:
-    ```json
-    [
-      {
-        "album_id": "1",
-        "query_en": "cluttered desk",
-        "pred": ["IMG_1234.jpg", "IMG_5678.jpg", ...]
-      }
-    ]
-    ```
-    """)
-
-    # ---- Footer ----
-    gr.Markdown(f"""
-    <div style="text-align:center; margin-top:48px; padding:24px; color:#888; font-size:0.9em;">
-        Ready to submit? Go to the
-        <a href="{LEADERBOARD_URL}" target="_blank" style="color:#F37021; font-weight:600; text-decoration:none;">PhotoBench-Full Leaderboard ↗</a>
-    </div>
-    """)
-
-
-demo.launch()

requirements.txt

@@ -1 +0,0 @@
-gradio>=4.0

scripts/package_images.py

@@ -1,44 +0,0 @@
-"""
-Package raw album images into zip archives for upload to the download Space.
-
-Usage:
-    python scripts/package_images.py /path/to/raw_albums /output/dir
-"""
-import argparse
-import zipfile
-from pathlib import Path
-
-
-def package_album(album_dir: Path, output_dir: Path):
-    album_name = album_dir.name
-    zip_path = output_dir / f"{album_name}.zip"
-
-    image_files = sorted([f for f in album_dir.iterdir() if f.is_file() and f.suffix.lower() in (".jpg", ".jpeg", ".png")])
-
-    print(f"Packing {album_name}: {len(image_files)} images -> {zip_path}")
-
-    with zipfile.ZipFile(zip_path, "w", zipfile.ZIP_DEFLATED) as zf:
-        for img in image_files:
-            zf.write(img, arcname=img.name)
-
-    size_mb = zip_path.stat().st_size / (1024 * 1024)
-    print(f"  Done: {size_mb:.1f} MB")
-
-
-def main():
-    parser = argparse.ArgumentParser(description="Package PhotoBench raw albums into zip files.")
-    parser.add_argument("raw_albums", type=Path, help="Path to raw_albums directory")
-    parser.add_argument("output_dir", type=Path, help="Output directory for zip files")
-    args = parser.parse_args()
-
-    args.output_dir.mkdir(parents=True, exist_ok=True)
-
-    for album_dir in sorted(args.raw_albums.glob("album*")):
-        if album_dir.is_dir():
-            package_album(album_dir, args.output_dir)
-
-    print("\nAll albums packaged.")
-
-
-if __name__ == "__main__":
-    main()