Learning a Spatial Activation Function for Efficient Image Restoration.
Please refer our paper for more details.
- python (tested with 3.5)
- PyTorch >= 0.2.0
Clone this repository into any place you want.
git clone https://github.com/kligvasser/xUnit
cd xUnit
The average PSNR in [dB] attained by several state of the art denoising algorithms on the BSD68:
Methods | BM3D | WNNM | EPLL | MLP | DnCNN-S | xDnCNN |
---|---|---|---|---|---|---|
# Parameters | - | - | - | - | 555K | 303K |
σ=25 | 28.56 | 28.82 | 28.68 | 28.95 | 29.22 | 29.21 |
σ=50 | 25.62 | 25.87 | 25.67 | 26.01 | 26.23 | 26.26 |
The average PSNR in [dB] attained in the task of 3× and 4× SR on BSD100 dataset:
Methods | SRCNN | xSRCNNc | xSRCNNf | SRResNet | xSRResNet |
---|---|---|---|---|---|
# Parameters | 57K | 44K | 32K | 1.546M | 1.155M |
3× | 28.41 | 28.54 | 28.53 | - | - |
4× | 26.90 | 27.04 | 27.06 | 27.58 | 27.61 |