3Diffusion: A Python repository from Xiaobin-Lu

3D Priors-Guided Diffusion for Blind Face Restoration

Xiaobin Lu^1,*, Xiaobin Hu^2,*, Jun Luo³, Ben Zhu⁴, Yaping Ruan⁴, Wenqi Ren¹

¹Shenzhen Campus of Sun Yat-sen University
²Tencent Youtu Lab
³University of Chinese Academy of Sciences
⁴Tencent Cloud Architecture Platform

Should you find our method beneficial, your support in starring this repository would be greatly appreciated. Thank you!

📖Table Of Contents

Installation
Pretrained Models
Inference
Visual Results On The Synthetic And Real-world Images

⭐Installation

# clone this repo
git clone https://github.com/838143396/3Diffusion.git
cd 3Diffusion

# create environment
conda create -n faceModel python=3.10
conda activate faceModel
pip install -r requirements.txt

# install pytorch3d(https://github.com/facebookresearch/pytorch3d/blob/main/INSTALL.md)
conda install -c fvcore -c iopath -c conda-forge fvcore iopath
conda install pytorch3d -c pytorch3d

🧬Pretrained Models

If you need the pretrained model, please download from baidu drive (password:5iig) , and put it into weights folder.

⚔️Inference

We provide some examples for inference, check inference.py for more arguments.

Blind Face Restoration

# for aligned face inputs
python inference.py  --aligned --in_path ./input  --out_path ./output

👀Visual Results On The Synthetic And Real-world Images

Citation

Please cite us if our work is useful for your research.

@article{lu20243d,
  title={3D Priors-Guided Diffusion for Blind Face Restoration},
  author={Lu, Xiaobin and Hu, Xiaobin and Luo, Jun and Zhu, Ben and Ruan, Yaping and Ren, Wenqi},
  journal={arXiv preprint arXiv:2409.00991},
  year={2024}
}

License

This project is released under the Apache 2.0 license.

Acknowledgement

This project is based on StableDiffusion and Difface. Thanks for their awesome work.

Contact

If you have any questions, please feel free to contact with me at luxb@mail2.sysu.edu.cn.

Xiaobin-Lu/3Diffusion