ldkien304/AoDai-VTON-Benchmark

Ao Dai VTON Benchmark

A face-obfuscated dataset for virtual try-on of Vietnamese white ao dai (áo dài trắng).

Dataset Summary

The Ao Dai VTON Benchmark contains paired images of high school students wearing white ao dai, along with garment references, human parsing masks, and agnostic (pose-only) representations. The dataset is designed for training and evaluating image-based virtual try-on models in a culturally specific context.

Privacy

All face regions have been obfuscated to protect personal identity. Do not attempt to recover, infer, or expose personal identities from the images.

Dataset Structure

train/
├── <school>/
│   ├── images/       # Person wearing ao dai
│   ├── masks/        # Human parsing masks
│   ├── agnostic/     # Pose representation (clothing removed)
│   └── garment/      # Reference garment image
test/
├── <school>/
│   ├── images/
│   ├── masks/
│   ├── agnostic/
│   └── garment/

Splits

Split	Images	Subjects
Train	4,027	~1,339
Test	472	~154

Usage

from datasets import load_dataset

dataset = load_dataset("ldkien304/AoDai-VTON-Benchmark")

Or access raw files directly from the repository.

License

This dataset is released under a Research Use Only Data Use Agreement. See LICENSE for full terms.

By accessing this dataset, you agree:

• Non-commercial research and evaluation only
• No face recognition, identity inference, or re-identification
• No redistribution without permission
• No attempt to recover personal identities

Citation

If you use this dataset in your research, please cite:

@misc{ldkien2026-ao-dai-vton-benchmark,
  title        = {Ao Dai VTON Benchmark},
  author       = {Le Duc Kien},
  year         = 2026,
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/ldkien304/AoDai-VTON-Benchmark}},
}

Contact

For access requests, please open a request on the Hugging Face dataset page.