pytorch-MRGAN-master
pytorch-MRGAN-master copied to clipboard
Code of the moire removal method "Unsupervised Moire Patterns Removal for Recaptured Screen Images"
pytorch-MRGAN-master
Code of the moire removal method "Unsupervised Moire Pattern Removal for Recaptured Screen Images"
Results

Framework

Prerequisites
- Linux or macOS
- Python 3
- CPU or NVIDIA GPU + CUDA CuDNN
Getting Started
Installation
- Clone this repo:
git clone https://github.com/JerryLeolfl/pytorch-MRGAN-master.git
cd pytorch-MRGAN-master
train/test
- Get the whole dataset(key:8wm4).
- To view training results and loss plots, run
python -m visdom.server
and click the URL http://localhost:8097. - Train a model:
python train.py --dataroot ./path/to/trainset --name moire_cyclegan --model cycle_gan
To see more intermediate results, check out ./checkpoints/moire_cyclegan/web/index.html
.
- Test the model:
python test.py --dataroot ./path/to/testset --name moire_cyclegan --model cycle_gan
- The test results will be saved to a html file here:
./results/moire_cyclegan/latest_test/index.html
.
Citation
If you use this code for your research, please cite our papers.
@article{YUE2021,
title = {Unsupervised Moiré Pattern Removal for Recaptured Screen Images},
journal = {Neurocomputing},
year = {2021},
author = {Huanjing Yue and Yijia Cheng and Fanglong Liu and Jingyu Yang}
}
Acknowledgments
Our code is inspired by pytorch-CycleGAN-and-pix2pix.