🔥 DiffRetouch: Using Diffusion to Retouch on the Shoulder of Experts (AAAI2025)

     

This is the official PyTorch codes for the paper:

DiffRetouch: Using Diffusion to Retouch on the Shoulder of Experts
Zhengpeng Duan^1,2, Jiawei Zhang², Zheng Lin⁴, Xin Jin¹, Xun-Dong Wang⁵, Dongqing Zou^2,3, Chunle Guo^1,6, Chongyi Li^1,6,†
1 VCIP, CS, Nankai University, ² SenseTime Research, ³ PBVR, ⁴ BNRist, Department of Computer Science and Technology, Tsinghua University, ⁵ Wuhan University of Technology, ⁶ NKIARI, Shenzhen Futian
^†Corresponding author.

⭐ If DiffRetouch is helpful to your images or projects, please help star this repo. Thank you! 👈

🔧 Dependencies and Installation

1. Clone repo

git clone https://github.com/adam-duan/DiffRetouch.git
cd DiffRetouch

2. Install packages

conda env create --file environment.yaml
conda activate diffretouch

🏄 Quick Inference

Step1: Download Checkpoints

Download the [checkpoints] and place them in the following directories: diffretouch_models/.

Step2: Prepare testing data

Download the [test_data.zip] and unzip it.

Step 3: Running testing command

bash test_adobe.sh/test_ppr.sh \
    promptdir      \      # Specify the expert style, e.g. '01-Experts-A' for Adobe5K and 'target_a' for PPR10K
    seed           \      # Set the random seed
    steps          \      # Set the number of diffusion steps, 20 by default

Step 4: Check the results

The processed results will be saved in the results/ directory.

🌱 Gradio Demo

python gradio_diffretouch.py \
    --ckpt_path   \       # Specify the checkpoint path, e.g. 'diffretouch_models/adobe.ckpt'

📜 License

This project is licensed under the Pi-Lab License 1.0 - see the LICENSE file for details.

📖 Citation

If you find our repo useful for your research, please consider citing our paper:

@inproceedings{duan2025diffretouch,
  title={DiffRetouch: Using Diffusion to Retouch on the Shoulder of Experts},
  author={Duan, Zheng-Peng and Zhang, Jiawei and Lin, Zheng and Jin, Xin and Wang, XunDong and Zou, Dongqing and Guo, Chun-Le and Li, Chongyi},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  year={2025}
}

📮 Contact

For technical questions, please contact adamduan0211[AT]gmail.com