A repository for the inference of face restoration

A list of face restoration papers and codes, can see in this repository .

Year	Pub	Abbreviation	Release
2021	CVPR	GFPGAN	✔
2021	CVPR	GPEN	✔
2022	NeurIPS	CodeFormer	✔
2022	CVPR	RestoreFormer	✔
2022	ECCV	VQFR	✔
2023	TPAMI	RestoreFormer++	✔

Update

• 2025.03.07: Add VQFR
• 2025.03.05: Add RestoreFormer、RestoreFormer++ inference codes.
• 2024.11.03: Add GFPGAN、CodeFromer 、GPEN inference codes.

TODO

• Add torch2onnx script
• Add more face restoration method

inference results

---

Installation

1. Clone repo

   git clone https://github.com/yeruiqian/Face_Restoration.git
   cd Face_Restoration

2. Install dependent packages

   # Install basicsr - https://github.com/xinntao/BasicSR
   pip install basicsr
   
   # Install facexlib - https://github.com/xinntao/facexlib
   pip install facexlib
   
   # you also can refer to my own environment 'requirements_self.txt'
   pip install -r requirements.txt
   python setup.py develop
   
   # If you want to enhance the background (non-face) regions with Real-ESRGAN,
   # you also need to install the realesrgan package
   pip install realesrgan

Download weights

Download the weights and put them in experiments/pretrained_models

链接: https://pan.baidu.com/s/1Dt0YElo1aKyPiZE0By1mGA 提取码: jvbt 

Inference

Method including GPEN、GFPGAN1.4、CodeFormer、RestoreFormer、RestoreFormer++、VQFRv2：

python inference.py -i inputs/cropped_faces -o results/out -v [method] -s 1

Usage: python inference_gfpgan.py -i inputs/whole_imgs -o results -v 1.3 -s 2 [options]...

  -h                   show this help
  -i input             Input image or folder. Default: inputs/cropped_faces
  -o output            Output folder. Default: results
  -v version           Different model. Option: GFPGANv1.4,GPEN,CodeFormer,RestoreFormer,RestoreFormer++,VQFRv2：. Default: GFPGANv1.4
  -s upscale           The final upsampling scale of the image. Default: 2
  -bg_upsampler        background upsampler. Default: realesrgan
  -bg_tile             Tile size for background sampler, 0 for no tile during testing. Default: 400
  -suffix              Suffix of the restored faces
  -only_center_face    Only restore the center face
  -aligned             Input are aligned faces
  -ext                 Image extension. Options: auto | jpg | png, auto means using the same extension as inputs. Default: auto

Acknowledgement

This project is based on BasicSR. Codes are brought from GFPGAN、CodeFromer 、GPEN、RestoreFormer++、VQFR. Thanks for their awesome works.