initial dataset commit

.gitattributes

@@ -1,59 +1,29 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.lz4 filter=lfs diff=lfs merge=lfs -text
-*.mds filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
+# Auto detect text files and perform LF normalization
+* text=auto
+# Images
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.bmp filter=lfs diff=lfs merge=lfs -text
+# JSON files
+*.json text
+# Markdown files
+*.md text
+# Python files
+*.py text
+# Text files
+*.txt text
+# CSV files
+*.csv filter=lfs diff=lfs merge=lfs -text
+images/* filter=lfs diff=lfs merge=lfs -text
 *.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
+*.jsonl filter=lfs diff=lfs merge=lfs -text
+*.pdf filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
-# Audio files - uncompressed
-*.pcm filter=lfs diff=lfs merge=lfs -text
-*.sam filter=lfs diff=lfs merge=lfs -text
-*.raw filter=lfs diff=lfs merge=lfs -text
-# Audio files - compressed
-*.aac filter=lfs diff=lfs merge=lfs -text
-*.flac filter=lfs diff=lfs merge=lfs -text
-*.mp3 filter=lfs diff=lfs merge=lfs -text
-*.ogg filter=lfs diff=lfs merge=lfs -text
 *.wav filter=lfs diff=lfs merge=lfs -text
-# Image files - uncompressed
-*.bmp filter=lfs diff=lfs merge=lfs -text
-*.gif filter=lfs diff=lfs merge=lfs -text
-*.png filter=lfs diff=lfs merge=lfs -text
-*.tiff filter=lfs diff=lfs merge=lfs -text
-# Image files - compressed
-*.jpg filter=lfs diff=lfs merge=lfs -text
-*.jpeg filter=lfs diff=lfs merge=lfs -text
-*.webp filter=lfs diff=lfs merge=lfs -text
-# Video files - compressed
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,164 @@
+---
+language:
+- vi
+- en
+task_categories:
+- visual-question-answering
+- question-answering
+tags:
+- infographic
+- vietnamese
+- vqa
+- document-understanding
+size_categories:
+- 10K<n<100K
+---
+
+# ViInfographicVQA
+
+## Overview
+
+**ViInfographicVQA** is a Vietnamese **Visual Question Answering (VQA)** benchmark for **infographic understanding**.  
+It evaluates models’ ability to **read, reason, and synthesize information** from data-rich, layout-heavy visuals that mix **text, charts, maps, and design elements**.
+
+Two settings are provided:
+- **Single-image VQA** – questions answered from one infographic.
+- **Multi-image VQA** – questions requiring reasoning across multiple, semantically related infographics.
+
+---
+
+## 📊 Dataset Summary
+
+| Split                 | #Images | #QAs  | Description                               |
+|----------------------|--------:|------:|-------------------------------------------|
+| Single-image (train) | 1,787   | 12,521| VQA on individual infographics             |
+| Single-image (test)  | 193     | 1,374 | Held-out evaluation                        |
+| Multi-image (train)  | 5,861   | 5,878 | Cross-image reasoning (training)           |
+| Multi-image (test)   | 653     | 636   | Cross-image reasoning (test)               |
+| **Total**            | **6,747** | **20,409** | Across all splits                     |
+
+- **Language:** Vietnamese  
+- **Domains:** Economy, Healthcare, Education, Society & Culture, Disasters & Accidents, Sports & Arts, Weather, etc.
+
+
+## 🗂️ Repository Layout
+
+```
+
+ViInfographicVQA/
+├── images/                # all image files (referenced by filename)
+├── <parquet files>        # four splits stored as parquet shards on the Hub
+└── README.md
+
+````
+
+
+## 🚀 Quickstart
+
+```python
+from datasets import load_dataset
+
+# Load all splits (parquet)
+ds = load_dataset("VLAI-AIVN/ViInfographicVQA")
+
+single_train = ds["single_train"]
+multi_train  = ds["multi_train"]
+
+# Each sample:
+# - images_paths: list of filenames (relative to `images/`)
+# - image: preview Image() (the first file)
+ex = multi_train[0]
+print(ex["images_paths"])  # e.g. ["13321.jpg", "13028.jpg", "13458.jpg"]
+preview = ex["image"]      # PIL.Image preview (for quick visualization)
+````
+
+### Read **all images** for multi-image samples (no local download)
+
+Use Hub file URIs, then cast to `Image()`:
+
+```python
+from datasets import Image, Sequence, load_dataset
+
+ds = load_dataset("VLAI-AIVN/ViInfographicVQA")
+repo_base = "hf://datasets/VLAI-AIVN/ViInfographicVQA/images"
+
+def add_full_paths(example):
+    example["images_full"] = [f"{repo_base}/{fn}" for fn in example["images_paths"]]
+    return example
+
+multi = ds["multi_train"].map(add_full_paths, remove_columns=[])
+multi = multi.cast_column("images_full", Sequence(Image()))
+
+all_imgs = multi[0]["images_full"]   # list[PIL.Image] — all referenced images
+```
+
+### Streaming (large-scale training)
+
+```python
+from datasets import load_dataset, Image, Sequence
+
+ds = load_dataset("VLAI-AIVN/ViInfographicVQA", streaming=True)
+repo_base = "hf://datasets/VLAI-AIVN/ViInfographicVQA/images"
+
+def add_full_paths(example):
+    example["images_full"] = [f"{repo_base}/{fn}" for fn in example["images_paths"]]
+    return example
+
+multi_stream = ds["multi_train"].map(add_full_paths)
+multi_stream = multi_stream.cast_column("images_full", Sequence(Image()))
+
+ex = next(iter(multi_stream))
+imgs = ex["images_full"]  # list of PIL.Image (lazy/streamed)
+```
+
+### Local download (offline use)
+
+```python
+from huggingface_hub import snapshot_download
+from datasets import load_dataset
+
+# Download the entire dataset repo locally (parquet + images)
+local_dir = snapshot_download(repo_id="VLAI-AIVN/ViInfographicVQA", repo_type="dataset")
+
+# Load from disk
+ds = load_dataset(local_dir)
+
+# Reconstruct absolute paths to images on disk if needed:
+import os
+images_root = os.path.join(local_dir, "images")
+def to_abs(example):
+    example["images_abs"] = [os.path.join(images_root, fn) for fn in example["images_paths"]]
+    return example
+
+multi_local = ds["multi_train"].map(to_abs)
+print(multi_local[0]["images_abs"][:3])  # ['/.../images/13321.jpg', ...]
+```
+
+> **Speed tip:** set `HF_HUB_ENABLE_HF_TRANSFER=1` to accelerate uploads/downloads.
+
+
+## 🔍 Research Applications
+
+* Multimodal reasoning on charts, tables, and dense text
+* Cross-image synthesis and comparison
+* Low-resource VQA in Vietnamese
+* Evaluation of OCR, layout parsing, and numerical reasoning
+
+
+## 🧮 Evaluation
+
+We use **Average Normalized Levenshtein Similarity (ANLS)** for string-based answer evaluation, which tolerates minor textual variations while penalizing semantic errors.
+
+
+## 📚 Citation
+
+If you use this dataset, please cite:
+
+```bibtex
+@article{van2025viinfographicvqa,
+  title={ViInfographicVQA: A Benchmark for Single and Multi-image Visual Question Answering on Vietnamese Infographics},
+  author={Van-Dinh, Tue-Thu and Tran, Hoang-Duy and Duong, Truong-Binh and Pham, Mai-Hanh and Le-Nguyen, Binh-Nam and Nguyen, Quoc-Thai},
+  journal={Proceedings of the AAAI Conference on Artificial Intelligence},
+  year={2026}
+}
+```

folder.py

@@ -0,0 +1,63 @@
+import os
+import argparse
+import logging
+from huggingface_hub import HfApi, login
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger("hf-uploader")
+
+def main():
+    p = argparse.ArgumentParser(description="Upload images/ folder and README.md to a HF dataset repo")
+    p.add_argument("--repo_id", required=True, help="e.g. VLAI-AIVN/ViInfographicVQA")
+    p.add_argument("--hf_token", default=None, help="Optional; or use `huggingface-cli login`")
+    p.add_argument("--images_dir", default="images", help="Local images folder to upload")
+    p.add_argument("--readme", default="README.md", help="Local README.md to upload")
+    p.add_argument("--branch", default=None, help="Optional target revision/branch (e.g., parquet-v1)")
+    p.add_argument("--commit_message", default="Upload images folder and README", help="Commit message")
+    p.add_argument("--skip_images", action="store_true", help="Skip uploading images folder")
+    p.add_argument("--skip_readme", action="store_true", help="Skip uploading README.md")
+    args = p.parse_args()
+
+    if args.hf_token:
+        login(token=args.hf_token)
+
+    api = HfApi()
+
+    # Upload images folder
+    if not args.skip_images:
+        if not os.path.isdir(args.images_dir):
+            raise FileNotFoundError(f"images_dir not found: {args.images_dir}")
+        logger.info(f"Uploading folder '{args.images_dir}' -> {args.repo_id}/images (repo_type=dataset)")
+        api.upload_folder(
+            folder_path=args.images_dir,
+            repo_id=args.repo_id,
+            repo_type="dataset",
+            path_in_repo="images",
+            commit_message=args.commit_message,
+            revision=args.branch,
+            allow_patterns=None,   # or e.g. ["*.jpg","*.png"]
+            ignore_patterns=None,
+        )
+        logger.info("✅ Images uploaded")
+
+    # Upload README.md
+    if not args.skip_readme:
+        if not os.path.isfile(args.readme):
+            raise FileNotFoundError(f"README not found: {args.readme}")
+        logger.info(f"Uploading README.md -> {args.repo_id}/README.md (repo_type=dataset)")
+        api.upload_file(
+            path_or_fileobj=args.readme,
+            path_in_repo="README.md",
+            repo_id=args.repo_id,
+            repo_type="dataset",
+            commit_message=args.commit_message,
+            revision=args.branch,
+        )
+        logger.info("✅ README uploaded")
+
+    logger.info("🎉 Done.")
+
+if __name__ == "__main__":
+    # Speed tip for large uploads:
+    # export HF_HUB_ENABLE_HF_TRANSFER=1
+    main()