Dual Prompting Image Restoration with Diffusion Transformers (CVPR2025)

     

Dehong Kong^1,2, Fan Li³, Zhixin Wang³, Jiaqi Xu⁴, Renjing Pei³, Wenbo Li³, WenQi Ren^1,2,5

¹School of Cyber Science and Technology, Shenzhen Campus of Sun Yat-sen University ²MoE Key Laboratory of Information Technology ³Huawei Noah’s Ark Lab ⁴The Chinese University of Hong Kong ⁵Guangdong Provincial Key Laboratory of Information Security Technology

⭐ Some code is coming soon! However, due to the company's open-source policy, checkpoint of DPIR is still on the way. If DPIR is helpful to your images or projects, please help star this repo. Thanks!

🚩Accepted by CVPR2025

🔎 Overview framework

📷 Real-World Results

⚙️ Dependencies and Installation

## git clone this repository
git clone https://github.com/kongdehong/DPIR.git
cd DPIR

# create an environment
conda create -n dpir python=3.8
conda activate dpir
pip install -r requirements.txt

📧 Contact

If you have any questions, please feel free to contact: kongdh@mail2.sysu.edu.cn

📓 License

This project is released under the Apache 2.0 license.

🎓Citations

If our code helps your research or work, please consider citing our paper. The following are BibTeX references:

@misc{kong2025dualpromptingimagerestoration,
  title={Dual Prompting Image Restoration with Diffusion Transformers}, 
  author={Dehong Kong and Fan Li and Zhixin Wang and Jiaqi Xu and Renjing Pei and Wenbo Li and WenQi Ren},
  year={2025},
  eprint={2504.17825},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2504.17825}, 
}

🌈 Acknowledgement

This project is based on stable diffusion3 and ControlNeXt. Thanks for their awesome work.