---
license: apache-2.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
dataset_info:
  features:
  - name: tgt_img
    dtype: image
  - name: cond_img_0
    dtype: image
  - name: cond_img_1
    dtype: image
  - name: prompt
    dtype: string
  - name: cond_prompt_0
    dtype: string
  - name: cond_prompt_1
    dtype: string
  splits:
  - name: train
    num_bytes: 27532187682.875
    num_examples: 29859
  download_size: 27509349685
  dataset_size: 27532187682.875
task_categories:
- text-to-image
tags:
- text-to-image
- customization
---

<img src='./MUSAR_Gen.png' width='100%' />

⭐️ Although MUSAR is trained solely on diptych data constructed from concatenated single-subject samples, we recognize that a high-quality multi-subject paired dataset is highly beneficial for the field of image customization. To accelerate progress in this field, we are releasing the high-quality multi-subject dataset generated by MUSAR: [MUSAR-Gen](https://huggingface.co/datasets/guozinan/MUSAR-Gen). It delivers FLUX-comparable image quality without exhibiting attribute entanglement issues. Hope it will be helpful to researchers working on related topics.

[Paper](https://huggingface.co/papers/2505.02823)

# dataset info
Construction details: The condition images are two subjects randomly selected from the [subjects200k](https://huggingface.co/datasets/Yuanshi/Subjects200K) dataset (excluding the 111,761 subjects used during the model training process). The prompt format is: "An undivided, seamless, and harmonious picture with two objects. in the xxx scene, Subject A and Subject B are placed together." By collecting the outputs of the MUSAR model, we obtained approximately 30,000 samples.

## Quick Start
- Load dataset
  ```python
  from datasets import load_dataset
  # Load dataset
  dataset = load_dataset('guozinan/MUSAR-Gen')

## Data Format
  | Key name             | Type    | Description                                                                                                                                                                                                |
  | -------------------- | ------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
  | `cond_img_0`         | `image` | Reference Image Information (first image).                                                                                                                                                                 |
  | `cond_img_1`         | `image` | Reference Image Information (second image).                                                                                                                                                                |
  | `tgt_img`            | `image` | Multi-subject customized result generated by the MUSAR model.                                                                                                                                              |
  | `cond_prompt_0`      | `str`   | Textual description of the corresponding subject in cond_img_0.                                                                                                                                            |
  | `cond_prompt_1`      | `str`   | Textual description of the corresponding subject in cond_img_1.                                                                                                                                            |
  | `prompt`             | `str`   | Textual description of the tgt_img content.                                                                                                                                                                |

## Citation
If you use MUSAR-Gen dataset, please cite our paper:
```
@article{guo2025musar,
  title={MUSAR: Exploring Multi-Subject Customization from Single-Subject Dataset via Attention Routing},
  author={Guo, Zinan and Zhang, Pengze and Wu, Yanze and Mou, Chong and Zhao, Songtao and He, Qian},
  journal={arXiv preprint arXiv:2505.02823},
  year={2025}
}
```