Lightweight Pyramid Networks for Image Deraining (TNNLS'2019)
@article{fu2019lightweight,
title={Lightweight pyramid networks for image deraining},
author={Fu, Xueyang and Liang, Borong and Huang, Yue and Ding, Xinghao and Paisley, John},
journal={IEEE transactions on neural networks and learning systems},
volume={31},
number={6},
pages={1794--1807},
year={2019},
publisher={IEEE}
}
Quantitative Result
The metrics are PSNR/SSIM. Both are evaluated on RGB channels.
| Method | Rain200L | Rain200H | Rain800 | Rain1200 | Rain1400 |
|---|---|---|---|---|---|
| lpnet_c16d5r5 | 32.12/0.955 | 21.96/0.785 | 22.81/0.820 | 27.74/0.862 | 28.30/0.884 |
Pretrained models can be downloaded from here
Network Complexity
| Input shape | Flops | Params |
|---|---|---|
| (3, 256, 256) | 1.77GFlops | 7.55k |