（1-复现使用手册）Restormer: Efficient Transformer for High-Resolution Image Restoration CVPR2022

目录

一、Installation

1.配置基本环境

2.下载数据集

2.1命令行方式

2.1.1安装go指令

2.1.2 配置文件修改变量

2.1.3 下载数据集

2.2 网站直接下载（推荐）

二、Demo

三、Training and Evaluation

四、Results

Image Deraining ​

Single-Image Motion Deblurring

Defocus Deblurring

Gaussian Image Denoising 

Real Image Denoising​

五、Related Works

---

paper:[2111.09881] Restormer: Efficient Transformer for High-Resolution Image Restoration (arxiv.org)

Git：GitHub - swz30/Restormer: [CVPR 2022--Oral] Restormer: Efficient Transformer for High-Resolution Image Restoration. SOTA for motion deblurring, image deraining, denoising (Gaussian/real data), and defocus deblurring.

一、Installation

This repository is built in PyTorch 1.8.1 and tested on Ubuntu 16.04 environment (Python3.7, CUDA10.2, cuDNN7.6). Follow these intructions

1.配置基本环境

git clone https://github.com/swz30/Restormer.git
cd Restormer

conda create -n pytorch181 python=3.7
conda activate pytorch181

conda install pytorch=1.8 torchvision cudatoolkit=10.2 -c pytorch
pip install matplotlib scikit-learn scikit-image opencv-python yacs joblib natsort h5py tqdm
pip install einops gdown addict future lmdb numpy pyyaml requests scipy tb-nightly yapf lpips

python setup.py develop --no_cuda_ext

2.下载数据集

2.1命令行方式

To be able to download datasets automatically you would need go and gdrive installed.

2.1.1安装go指令

curl -O https://storage.googleapis.com/golang/go1.11.1.linux-amd64.tar.gz
mkdir -p ~/installed
tar -C ~/installed -xzf go1.11.1.linux-amd64.tar.gz
mkdir -p ~/go

第一个指令：go1.11.1.linux-amd64.tar.gz 文件会被下载到当前工作目录下

后面三个指令就是解压并安装 Go

记得修改对应的路径哈~，比如我的就会这么改，不然就会安到默认路径（比如我的是home/ ）

mkdir -p /data1/zhangjiening/Restormer/installed 

tar -C /data1/zhangjiening/Restormer/installed  -xzf go1.11.1.linux-amd64.tar.gz

mkdir -p /data1/zhangjiening/Restormer/go

2.1.2 配置文件修改变量

vim ~/.bashrc 打开配置文件，在最后添加这两行，Go 的可执行文件目录添加到系统 PATH 变量中了

export GOPATH=$HOME/go
export PATH=$PATH:$HOME/go/bin:$HOME/installed/go/bin

还是修改路径的问题：

export GOPATH=/data1/zhangjiening/Restormer/go
export PATH=$PATH:/data1/zhangjiening/Restormer/go/bin:/data1/zhangjiening/Restormer/installed/go/bin

 

2.1.3 下载数据集

go get github.com/prasmussen/gdrive

2.2 网站直接下载（推荐）

GitHub - prasmussen/gdrive: Google Drive CLI Client

不想上面这么麻烦的话，直接下载呗

直接把文件扔到   ~/go/bin   路径下

二、Demo

python demo.py --task Task_Name --input_dir path_to_images --result_dir save_images_here

python demo.py --task Deraining --input_dir path_to_images --result_dir save_images_here
python demo.py --task Motion Deblurring --input_dir path_to_images --result_dir save_images_here
python demo.py --task Defocus Deblurring --input_dir path_to_images --result_dir save_images_here
python demo.py --task Denoising --input_dir path_to_images --result_dir save_images_here

Example usage to perform Defocus Deblurring on a directory of images:

python demo.py --task Single_Image_Defocus_Deblurring --input_dir './demo/degraded/' --result_dir './demo/restored/'

Example usage to perform Defocus Deblurring on an image directly:

python demo.py --task Single_Image_Defocus_Deblurring --input_dir './demo/degraded/portrait.jpg' --result_dir './demo/restored/'

三、Training and Evaluation

Task	Training Instructions	Testing Instructions	Restormer's Visual Results
Deraining	Link	Link	Download
Motion Deblurring	Link	Link	Download
Defocus Deblurring	Link	Link	Download
Gaussian Denoising	Link	Link	Download
Real Denoising	Link	Link	Download

四、Results

SOTA for motion deblurring, image deraining, denoising (Gaussian/real data), and defocus deblurring.

Image Deraining 

Single-Image Motion Deblurring

Defocus Deblurring

Gaussian Image Denoising 

Grayscale	Color

Real Image Denoising

五、Related Works

• Learning Enriched Features for Fast Image Restoration and Enhancement, TPAMI 2022. Paper | Code
• Multi-Stage Progressive Image Restoration, CVPR 2021. Paper | Code
• Learning Enriched Features for Real Image Restoration and Enhancement, ECCV 2020. Paper | Code
• CycleISP: Real Image Restoration via Improved Data Synthesis, CVPR 2020. Paper | Code

 

---

后续会更新Restormer精读笔记和复现笔记~相关link随时同步在这里