swal's Introduction

SWAL

Code for Paper "Selective Wavelet Attention Learning for Single Image Deraining"

Prerequisites

Python 3
PyTorch

Models

We provide the models trained on DDN, DID, Rain100H, Rain100L, and AGAN datasets in the following links:

Download them into the model folder before testing.

Dataset

Download the rain datasets.
Arrange the images and generate a list file, just like the rain12 set in the data folder.

You can also modify the data_loader code in your manner.

Run

Train SWAL on a single GPU:

 CUDA_VISIBLE_DEVICES=0 python main.py --ngf=16 --ndf=64  --output_height=320  --trainroot=YOURPATH --trainfiles='YOUR_FILELIST'  --save_iter=1 --batchSize=8 --nrow=8 --lr_d=1e-4 --lr_g=1e-4  --cuda  --nEpochs=500

Train SWAL on multiple GPUs:

 CUDA_VISIBLE_DEVICES=0,1,2,3 python main.py --ngf=16 --ndf=64  --output_height=320  --trainroot=YOURPATH --trainfiles='YOUR_FILELIST'  --save_iter=1 --batchSize=32 --nrow=8 --lr_d=1e-4 --lr_g=1e-4  --cuda  --nEpochs=500

Test SWAL:

 CUDA_VISIBLE_DEVICES=0 python test.py --ngf=16  --outf='test' --testroot='data/rain12_test' --testfiles='data/rain12_test.list' --pretrained='model/rain100l_best.pth'  --cuda

Adjust the parameters according to your own settings.

Citation

If you use our codes, please cite the following paper:

 @article{huang2021selective,
   title={Selective Wavelet Attention Learning for Single Image Deraining},
   author={Huang, Huaibo and Yu, Aijing and Chai, Zhenhua and He, Ran and Tan, Tieniu},
   journal={International Journal of Computer Vision},
   volume={129},
   number={4},
   pages={1282--1300},
   year={2021},
  }

The released codes are only allowed for non-commercial use.

swal's People

Contributors

Stargazers

Watchers

swal's Issues

I get so many model after I trained ,Is this normal?

Thanks a lot for your contributions. I train it by my own dataset.
after training I got a lot of model ,is this normal?
so I need to chose the best one to use?

If I want to train a model, what kind of data set should I prepare?

Thanks a lot for your contributions.
I want to know how do I prepare the data set. Is the pattern of raindrops or a composite of raindrops and pictures?
Put all the data in the data folder? ground truthu also?
please give a answer

Problems with model training

Thanks a lot for your contributions. We are very interested in your work, but after retraining the model (with everything adjusted to match the parameters in the paper), we found the new model output unacceptable and appeared to be outputting the wavelet components. It is worth mentioning that we did not find any problem in the output using the provided pre-trained model. This made us confused. So we replaced another dataset and this still happened. We are sure that it is not a problem with the input and output of the model. Now we are at a loss, can you help us?
Our training environment is torch1.7 cuda10.1 on tesla v100

Translated with www.DeepL.com/Translator (free version)

Recommend Projects