Single image super resolution network based on SRResNet architecture ("Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network") but with reduced number of channels and depthwise convolution in decoder. It enhances the resolution of the input image by a factor of 4.
Low resolution:
Linear interpolation:
Super resolution:
Metric | Value |
---|---|
PSNR | 28.61 dB |
GFlops | 39.713 |
MParams | 0.363 |
Source framework | Pytorch* |
For reference, PSNR for bicubic upsampling on test dataset is 26.35 dB.
Link to performance table
name: "input" , shape: [1x3x200x200] - An input image in the format [BxCxHxW], where:
Expected color order is BGR.
[*] Other names and brands may be claimed as the property of others. [*] Other names and brands may be claimed as the property of others.