Blind Super-Resolution Kernel Estimation using an Internal-GAN

"KernelGAN"

Sefi Bell-Kligler, Assaf Shocher, Michal Irani

(Official implementation)

Paper: https://arxiv.org/abs/1909.06581

Project page: http://www.wisdom.weizmann.ac.il/~vision/kernelgan/

Accepted NeurIPS 2019 (oral)

Usage:

Quick usage on your data:

To run KernelGAN on all images in <input_image_path>:

python train.py --input-dir <input_image_path>

This will produce kernel estimations in the results folder

Extra configurations:

--X4 : Estimate the X4 kernel

--SR : Perform ZSSR using the estimated kernel

--real : Real-image configuration (effects only the ZSSR)

--output-dir : Output folder for the images (default is results)

Data:

Download the DIV2KRK dataset: dropbox

Reproduction code for your own Blind-SR dataset: github