dslisleedh / x-restormer

Arxiv：A Comparative Study of Image Restoration Networks for General Backbone Network Design

X-Restormer [Paper Link]

A Comparative Study of Image Restoration Networks for General Backbone Network Design

Xiangyu Chen*, Zheyuan Li*, Yuandong Pu*,Yihao Liu, Jiantao Zhou, Yu Qiao and Chao Dong

BibTeX

@article{chen2023comparative,
  title={A Comparative Study of Image Restoration Networks for General Backbone Network Design}, 
  author={Chen, Xiangyu and Li, Zheyuan and Pu, Yuandong and Liu, Yihao and Zhou, Jiantao and Qiao, Yu and Dong, Chao},
  journal={arXiv preprint arXiv:2310.11881},
  year={2023}
}

Environment

• PyTorch>=1.13.0 (Recommend NOT using torch 1.8!)
• BasicSR==1.4.2

Installation

Install Pytorch first. Then,

pip install -r requirements.txt
python setup.py develop

How To Test

• Refer to ./options/test for the configuration file of the model to be tested, and prepare the testing data and pretrained model.
• The pretrained models are available at Google Drive or Baidu Netdisk (access code: im3q).
• Then run the following codes (taking sr_300k.pth as an example):

python xrestormer/test.py -opt options/test/001_xrestormer_sr.yml

The testing results will be saved in the ./results folder.

• Refer to ./options/test/001_xrestormer_sr.yml for inference without the ground truth image.

How To Train

• Refer to ./options/train for the configuration file of the model to train.
• Preparation of training data can refer to this page. ImageNet dataset can be downloaded at the official website.
• The training command is like

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python -m torch.distributed.launch --nproc_per_node=8 --master_port=1231 xrestormer/train.py -opt ./options/train/001_xrestormer_sr.yml --launcher pytorch

• Note that the default batch size per GPU is 4, which will cost about 60G memory for each GPU.

The training logs and weights will be saved in the ./experiments folder.

Results

The inference results on benchmark datasets are available at Google Drive or Baidu Netdisk (access code: g9dw).

Contact

If you have any question, please contact [email protected] or [email protected].