This is the Pytorch implementation for "A general model compression method for image restoration network".
The SWConv is depicted as follows:
The SWConv-F is depicted as follows:
The SW-DnCNN, SW-RCAN and SW-DnCNN-3 are implemented in SWDnCNN.py, SWRCAN.py and SWDnCNN3.py, respectively. The original networks can be replaced by the corresponding ones for retraining and test.
If you find this project useful, we would be grateful if you cite this paper：
@article{model_compression,
author = {Jie Xiao, Zhi Jin*, Huanrong Zhang},
journal = {Signal Processing: Image Communication},
title = {A general model compression method for image restoration network},
year = {2021}
}
This repository is released under the MIT License as found in the LICENSE file. Code in this repo is for non-commercial use only.