README.md

---
license: other
extra_gated_prompt: >-
  This dataset is available **exclusively for non-commercial academic research**.
  You must confirm that you already have access to the original DressCode dataset, and use an **institutional email** to request access.
  By requesting access, you agree NOT to use this dataset or its content for any commercial purposes, and NOT to redistribute any part of it publicly or privately.
extra_gated_fields:
  Name: text
  Country: country
  Affiliation: text
  Email (Education Email Only): text
  Have you been granted access to the original DressCode dataset?: checkbox
  I agree to use this dataset for non-commercial research purposes ONLY and not to share it publicly or privately with others: checkbox
viewer: false
---

Okay, I can help with that. Here is the updated README with the provided links, formatted to be left-aligned.

-----

# DressCode-MR: A Large-Scale Multi-Reference Virtual Try-On Dataset

Supported by [LavieAI](https://lavieai.com/) and [LoomlyAI](https://www.loomlyai.com/en)

[arXiv](https://arxiv.org/abs/2508.20586) | [Hugging Face](https://huggingface.co/zhengchong/FastFit-MR-1024) | [GitHub](https://github.com/Zheng-Chong/FastFit) | [Demo](https://fastfit.lavieai.com) | [License](https://github.com/Zheng-Chong/FastFit/tree/main)

**DressCode-MR** is a large-scale, multi-reference virtual try-on dataset constructed upon the original [DressCode](https://github.com/aimagelab/dress-code) dataset. It contains over **28,000 multi-reference virtual try-on samples** designed to facilitate and evaluate virtual try-on models capable of handling multiple fashion items—such as tops, bottoms, dresses, shoes, and bags—simultaneously.


<div align="center">
  <img src="dataset.png" alt="DressCode-MR Dataset" width="800">
</div>

## Dataset Details

  * **Multi-Reference Samples**: Each sample consists of a person's image paired with a set of compatible clothing and accessory items, enabling models to learn how to coordinate multiple fashion pieces within a single scene.
  * **Large Scale**: The dataset includes a total of **28,179** high-quality multi-reference samples, with **25,779** designated for training and **2,400** for testing.
  * **Source**: This dataset is built upon the [DressCode](https://github.com/aimagelab/dress-code) dataset.

## Access and License

This dataset is released under the exact same license as the original [DressCode](https://github.com/aimagelab/dress-code) dataset. Therefore, before you can request access to the **DressCode-MR** dataset, you must first complete the following steps:

1.  Apply for and be granted a license to use the [DressCode](https://github.com/aimagelab/dress-code) dataset.
2.  Use your educational/academic email address (e.g., one ending in `.edu`, `.ac`, etc.) to request access to the **DressCode-MR** dataset on Hugging Face. **Any requests from non-academic email addresses will be rejected.**

## Usage

After downloading the dataset, you can decompress the files using the following commands:

```bash
cat DressCode-MR.tar.gz.part_* > DressCode-MR.tar.gz
tar -zxvf DressCode-MR.tar.gz
```

## Citation

If you use the **DressCode-MR** dataset in your research, please cite our [FastFit](https://arxiv.org/abs/2508.20586) paper.

```bibtex
@misc{chong2025fastfitacceleratingmultireferencevirtual,
      title={FastFit: Accelerating Multi-Reference Virtual Try-On via Cacheable Diffusion Models}, 
      author={Zheng Chong and Yanwei Lei and Shiyue Zhang and Zhuandi He and Zhen Wang and Xujie Zhang and Xiao Dong and Yiling Wu and Dongmei Jiang and Xiaodan Liang},
      year={2025},
      eprint={2508.20586},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2508.20586}, 
}
```

## Acknowledgement

We thank the contributors to the [DressCode](https://github.com/aimagelab/dress-code) project, as their work provided the foundation for our **DressCode-MR** dataset.