ngnquan/ViFinTab

📄 Vietnamese Financial Reports – Table Detection Dataset

This dataset contains images and annotations of tables extracted from Vietnamese financial reports, focusing on the three most important financial statement types used in accounting and auditing. Each table in the dataset is annotated with bounding boxes in YOLO OBB format, and stored in a Parquet file along with the base64-encoded image.

The dataset is optimized for use with the HuggingFace Datasets library.

🏷 Class Labels

Class ID	Table Type	Vietnamese Name
0	Balance Sheets	Bảng cân đối kế toán
1	Cash Flow Statements	Báo cáo lưu chuyển tiền tệ
2	Income Statements	Báo cáo kết quả hoạt động kinh doanh

📦 Dataset Files

Each Parquet file contains a list of images and their corresponding annotations.

Parquet Schema

Column	Type	Description
image_path	string	Relative path of the original image
image_base64	string	Base64-encoded image bytes
objects	list(dict)	List of annotated objects per image

📄 License

GPL-3.0

✨ Citation

Please cite or reference this dataset if used in your work.

BibTeX

@inproceedings{tran2025vifintab,
  title={ViFinTab: A Domain-Specific Dataset and Baseline for Table Detection in Vietnamese Financial Statements},
  author={Tran, Hong Tai and Mai, Xuan Toan and Nguyen, Nam-Quan and Tran, Tuan-Anh},
  booktitle={International Conference on Multi-disciplinary Trends in Artificial Intelligence},
  pages={387--398},
  year={2025},
  organization={Springer}
}

---