Add dataset card and compatibility files

.gitattributes

@@ -58,3 +58,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+original_annotations/test.jsonl filter=lfs diff=lfs merge=lfs -text
+original_annotations/train.jsonl filter=lfs diff=lfs merge=lfs -text
+original_annotations/val.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,227 +1,83 @@
 ---
-dataset_info:
-- config_name: image_examples
-  features:
-  - name: image
-    dtype: image
-  - name: img_fn
-    dtype: string
-  - name: metadata_fn
-    dtype: string
-  - name: movie
-    dtype: string
-  - name: objects
-    list: string
-  - name: width
-    dtype: int64
-  - name: height
-    dtype: int64
-  - name: image_width
-    dtype: int64
-  - name: image_height
-    dtype: int64
-  - name: boxes
-    list:
-      list: float32
-  - name: segms_json
-    dtype: string
-  - name: question_count
-    dtype: int64
-  - name: source_zip_crc_mismatch
-    dtype: bool
-  - name: image_metadata_dimension_mismatch
-    dtype: bool
-  - name: annotations
-    list:
-    - name: movie
-      dtype: string
-    - name: objects
-      list: string
-    - name: interesting_scores
-      list: int64
-    - name: answer_likelihood
-      dtype: string
-    - name: img_fn
-      dtype: string
-    - name: metadata_fn
-      dtype: string
-    - name: answer_orig
-      dtype: string
-    - name: question_orig
-      dtype: string
-    - name: rationale_orig
-      dtype: string
-    - name: question_tokens
-      list:
-      - name: kind
-        dtype: string
-      - name: text
-        dtype: string
-      - name: object_indices
-        list: int64
-    - name: answer_choice_tokens
-      list:
-        list:
-        - name: kind
-          dtype: string
-        - name: text
-          dtype: string
-        - name: object_indices
-          list: int64
-    - name: answer_label
-      dtype: int64
-    - name: answer_match_iter
-      list: int64
-    - name: answer_sources
-      list: int64
-    - name: rationale_choice_tokens
-      list:
-        list:
-        - name: kind
-          dtype: string
-        - name: text
-          dtype: string
-        - name: object_indices
-          list: int64
-    - name: rationale_sources
-      list: int64
-    - name: rationale_match_iter
-      list: int64
-    - name: rationale_label
-      dtype: int64
-    - name: source_zip_crc_mismatch
-      dtype: bool
-    - name: img_id
-      dtype: string
-    - name: question_number
-      dtype: int64
-    - name: annot_id
-      dtype: string
-    - name: match_fold
-      dtype: string
-    - name: match_index
-      dtype: int64
-    - name: question_text
-      dtype: string
-    - name: answer_choice_texts
-      list: string
-    - name: rationale_choice_texts
-      list: string
-  splits:
-  - name: train
-    num_bytes: 29200718216
-    num_examples: 80418
-  - name: validation
-    num_bytes: 3755841689
-    num_examples: 9929
-  - name: test
-    num_bytes: 3731295388
-    num_examples: 9557
-  download_size: 30739927109
-  dataset_size: 36687855293
-- config_name: questions
-  features:
-  - name: movie
-    dtype: string
-  - name: objects
-    list: string
-  - name: interesting_scores
-    list: int64
-  - name: answer_likelihood
-    dtype: string
-  - name: img_fn
-    dtype: string
-  - name: metadata_fn
-    dtype: string
-  - name: answer_orig
-    dtype: string
-  - name: question_orig
-    dtype: string
-  - name: rationale_orig
-    dtype: string
-  - name: question_tokens
-    list:
-    - name: kind
-      dtype: string
-    - name: text
-      dtype: string
-    - name: object_indices
-      list: int64
-  - name: answer_choice_tokens
-    list:
-      list:
-      - name: kind
-        dtype: string
-      - name: text
-        dtype: string
-      - name: object_indices
-        list: int64
-  - name: answer_label
-    dtype: int64
-  - name: answer_match_iter
-    list: int64
-  - name: answer_sources
-    list: int64
-  - name: rationale_choice_tokens
-    list:
-      list:
-      - name: kind
-        dtype: string
-      - name: text
-        dtype: string
-      - name: object_indices
-        list: int64
-  - name: rationale_sources
-    list: int64
-  - name: rationale_match_iter
-    list: int64
-  - name: rationale_label
-    dtype: int64
-  - name: source_zip_crc_mismatch
-    dtype: bool
-  - name: img_id
-    dtype: string
-  - name: question_number
-    dtype: int64
-  - name: annot_id
-    dtype: string
-  - name: match_fold
-    dtype: string
-  - name: match_index
-    dtype: int64
-  - name: question_text
-    dtype: string
-  - name: answer_choice_texts
-    list: string
-  - name: rationale_choice_texts
-    list: string
-  splits:
-  - name: train
-    num_bytes: 772530124
-    num_examples: 212923
-  - name: validation
-    num_bytes: 96810191
-    num_examples: 26534
-  - name: test
-    num_bytes: 83927907
-    num_examples: 25263
-  download_size: 760672290
-  dataset_size: 953268222
+license: other
+pretty_name: "VCR v1.0"
+task_categories:
+- visual-question-answering
+- image-to-text
+tags:
+- image
+- visual-commonsense-reasoning
+- vcr
+- r2c
+- arxiv:1811.10830
 configs:
 - config_name: image_examples
   data_files:
   - split: train
-    path: image_examples/train-*
+    path: image_examples/train-*.parquet
   - split: validation
-    path: image_examples/validation-*
+    path: image_examples/validation-*.parquet
   - split: test
-    path: image_examples/test-*
+    path: image_examples/test-*.parquet
+  default: true
 - config_name: questions
   data_files:
   - split: train
-    path: questions/train-*
+    path: questions/train-*.parquet
   - split: validation
-    path: questions/validation-*
+    path: questions/validation-*.parquet
   - split: test
-    path: questions/test-*
+    path: questions/test-*.parquet
 ---
+
+# VCR v1.0
+
+VCR is the Visual Commonsense Reasoning dataset from "From Recognition to Cognition: Visual Commonsense Reasoning" (CVPR 2019).
+
+This Hugging Face version has two loadable configs:
+
+- `image_examples`: the default viewer-friendly config, one row per unique image, with grouped annotations.
+- `questions`: one row per original VCR question/answer/rationale example.
+
+The original annotation JSONL files are also included under `original_annotations/` for legacy compatibility.
+
+## Source archive note
+
+The hosted `vcr1images.zip` object has one JPEG member whose zip CRC does not match, although the member's full byte stream can be recovered and decoded successfully:
+
+- `vcr1images/lsmdc_1010_TITANIC/1010_TITANIC_02.33.49.825-02.33.53.701@0.jpg`
+
+Rows whose image comes from that zip member have `source_zip_crc_mismatch=true`.
+
+The `width` and `height` columns preserve the original metadata JSON dimensions used by boxes/segmentations and by legacy `r2c` code. `image_width` and `image_height` store the decoded JPEG dimensions. Rows where those differ have `image_metadata_dimension_mismatch=true`.
+
+## Legacy r2c layout
+
+The original `rowanz/r2c` code expects `train.jsonl`, `val.jsonl`, `test.jsonl`, and a `vcr1images/` directory. Use:
+
+```bash
+python scripts/export_r2c_layout.py <repo_id> /path/to/r2c/data
+```
+
+## License and terms
+
+Use of VCR is governed by the original VCR/AI2 dataset license and website terms:
+
+- https://visualcommonsense.com/license/
+- https://visualcommonsense.com/terms/
+
+## Citation
+
+```bibtex
+@inproceedings{zellers2019vcr,
+  author = {Zellers, Rowan and Bisk, Yonatan and Farhadi, Ali and Choi, Yejin},
+  title = {From Recognition to Cognition: Visual Commonsense Reasoning},
+  booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
+  month = {June},
+  year = {2019}
+}
+```
+
+## Content warning
+
+VCR is annotated from movie images and may contain nudity, violence, offensive content, and biased or problematic depictions.

original_annotations/test.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f59afa1f1fb663d301d3e772dd6d7d9b6b6ab2fb0485345aef9ee07ca439c864
+size 34097024

original_annotations/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d911d5c55b1bbf1ac5a6e3f73a9a6e2c2467559463c50c13e74292e0da4ccfaa
+size 392317321

original_annotations/val.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:328478b2f429fd899b0217ad4bcac18e2c017829c43dfcff7cf01d66857a8a01
+size 48969062

scripts/export_r2c_layout.py

@@ -0,0 +1,91 @@
+#!/usr/bin/env python3
+"""Export the Hugging Face VCR dataset back to the original r2c layout."""
+
+from __future__ import annotations
+
+import argparse
+import json
+import shutil
+from pathlib import Path
+
+from datasets import load_dataset, load_from_disk
+from huggingface_hub import hf_hub_download
+from tqdm import tqdm
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("repo_id", help="Hugging Face dataset repo id, e.g. rowanz/vcr")
+    parser.add_argument("output_dir", type=Path, help="Directory to create in r2c-compatible layout")
+    parser.add_argument("--token", default=None, help="HF token for private/gated repos")
+    parser.add_argument("--limit", type=int, default=None, help="Optional per-split row limit for smoke tests")
+    parser.add_argument("--local-stage-dir", type=Path, default=None, help="Use a local vcr_hf.py stage directory instead of the Hub")
+    args = parser.parse_args()
+
+    args.output_dir.mkdir(parents=True, exist_ok=True)
+    images_dir = args.output_dir / "vcr1images"
+    images_dir.mkdir(parents=True, exist_ok=True)
+
+    for split in ("train", "val", "test"):
+        if args.local_stage_dir is None:
+            src = hf_hub_download(
+                repo_id=args.repo_id,
+                repo_type="dataset",
+                filename=f"original_annotations/{split}.jsonl",
+                token=args.token,
+            )
+        else:
+            src = args.local_stage_dir / "original_annotations" / f"{split}.jsonl"
+        dst = args.output_dir / f"{split}.jsonl"
+        if args.limit is None:
+            shutil.copyfile(src, dst)
+        else:
+            with open(src, "r", encoding="utf-8") as f_in, open(dst, "w", encoding="utf-8") as f_out:
+                for i, line in enumerate(f_in):
+                    if i >= args.limit:
+                        break
+                    f_out.write(line)
+
+    if args.local_stage_dir is None:
+        split_expr = {s: f"{s}[:{args.limit}]" if args.limit is not None else s for s in ("train", "validation", "test")}
+        ds = load_dataset(args.repo_id, "image_examples", split=split_expr, token=args.token)
+        if not isinstance(ds, dict):
+            ds = {"train": ds}
+    else:
+        local_ds = load_from_disk(args.local_stage_dir / "image_examples.dataset")
+        ds = {}
+        for split_name in ("train", "validation", "test"):
+            ds[split_name] = local_ds[split_name].select(range(min(args.limit, len(local_ds[split_name])))) if args.limit else local_ds[split_name]
+
+    seen: set[str] = set()
+    for split_name, split_ds in ds.items():
+        for row in tqdm(split_ds, desc=f"export {split_name} images"):
+            img_fn = row["img_fn"]
+            metadata_fn = row["metadata_fn"]
+
+            if img_fn not in seen:
+                image = row["image"]
+                dst_img = images_dir / img_fn
+                dst_img.parent.mkdir(parents=True, exist_ok=True)
+                image.save(dst_img)
+                seen.add(img_fn)
+
+            if metadata_fn not in seen:
+                metadata = {
+                    "boxes": row["boxes"],
+                    "segms": json.loads(row["segms_json"]),
+                    "names": row["objects"],
+                    "width": row["width"],
+                    "height": row["height"],
+                }
+                dst_meta = images_dir / metadata_fn
+                dst_meta.parent.mkdir(parents=True, exist_ok=True)
+                with open(dst_meta, "w", encoding="utf-8") as f:
+                    json.dump(metadata, f, separators=(",", ":"))
+                seen.add(metadata_fn)
+
+    print(f"Wrote r2c-compatible VCR layout to {args.output_dir}")
+
+
+if __name__ == "__main__":
+    main()