无论是自己、家人或是朋友、客户的照片，免不了有些是黑白的、被污损的、模糊的，总想着修复一下。作为一个程序员 或者 程序员的家属，当然都有责任满足他们的需求、实现他们的想法。除了这个，学习了本文的成果，或许你还可以用来赚点小钱。
Python Anaconda 的安装、配置等等请参阅：
https://blog.csdn.net/beijinghorn/article/details/134334021
https://blog.csdn.net/beijinghorn/article/details/134334021
7 DifFace
7.1 论文Paper
DifFace: Blind Face Restoration with Diffused Error Contraction
https://arxiv.org/abs/2212.06512
Authors:
Zongsheng Yue, https://zsyoaoa.github.io/
Chen Change Loy https://www.mmlab-ntu.com/person/ccloy/
⭐ If DifFace is helpful to your images or projects, please help star this repo. Thanks! 🤗
7.2 Update
2022.12.19: Add Colab demo google colab logo.
2022.12.17: Add the Hugging Face.
2022.12.13: Create this repo.
7.3 Applications
7.3.1 旧照片修复 Old Photo Enhancement
7.3.2 人脸重建 Face Restoration
7.4 依赖项Requirements
A suitable conda environment named DifFace can be created and activated with:
conda env create -f environment.yaml
conda activate DifFace
7.5 运行指令 Inference
7.5.1 Face image restoration (cropped and aligned)
python inference_difface.py --aligned --in_path [image folder/image path] --out_path [result folder] --gpu_id [gpu index]
7.5.2 Whole image enhancement
python inference_difface.py --in_path [image folder/image path] --out_path [result folder] --gpu_id [gpu index]
7.6 Training
7.6.1 Prepare data
Download the FFHQ dataset, and resize them into size 512x512.
https://github.com/NVlabs/ffhq-dataset
python datapipe/prepare/face/big2small_face.py --face_dir [Face folder(1024x1024)] --save_dir [Saving folder] --pch_size 512
Extract the image path into 'datapipe/files_txt/ffhq512.txt'
python datapipe/prepare/face/split_train_val.py --face_dir [Face folder(512x512)] --save_dir [Saving folder]
Making the testing dataset
python datapipe/prepare/face/make_testing_data.py --files_txt datapipe/files_txt/ffhq512.txt --save_dir [Saving folder]
7.6.2 Train diffusion model
CUDA_VISIBLE_DEVICES=0,1,2,3 torchrun --standalone --nproc_per_node=4 --nnodes=1 main_diffusion.py --cfg_path configs/training/diffsuion_ffhq512.yaml --save_dir [Logging Folder]
7.6.3 Train diffused estimator (SwinIR)
CUDA_VISIBLE_DEVICES=0,1,2,3 torchrun --standalone --nproc_per_node=4 --nnodes=1 main_sr.py --cfg_path configs/training/swinir_ffhq512.yaml --save_dir [Logging Folder]
7.7 权利 License
This project is licensed under NTU S-Lab License 1.0. Redistribution and use should follow this license.
https://github.com/sczhou/CodeFormer/blob/master/LICENSE
7.8 相关知识 Acknowledgement
This project is based on Improved Diffusion Model. Some codes are brought from BasicSR, YOLOv5-face, and FaceXLib. We also adopt Real-ESRGAN to support background image enhancement. Thanks for their awesome works.
该项目基于改进的扩散模型。一些代码来自BasicSR、YOLOv5 face和FaceXLib。我们还采用Real ESRGAN来支持背景图像增强。感谢他们出色的作品。
https://github.com/openai/improved-diffusion
https://github.com/XPixelGroup/BasicSR
https://github.com/deepcam-cn/yolov5-face
https://github.com/xinntao/facexlib
https://github.com/xinntao/Real-ESRGAN
7.9 联系作者 Contact
If you have any questions, please feel free to contact me via zsyzam@gmail.com.
7.10 点评
论文。
