Preserving Full Degradation Details for Blind Image Super-Resolution

Pytorch implementation of paper Preserving Full Degradation Details for Blind Image Super-Resolution.

Installation

Install python 3.8.0, torch 2.0.0, CUDA 11.7 and other essential packages (Note that using other versions of packages may affect performance).
Clone this repo

git clone https://github.com/Chernobyllight/ReDSR
cd ReDSR

test setting1: model trained only on isotropic Gaussian blur kernels

test setting2: model trained on anisotropic Gaussian blur kernels and additive Gaussian noise

Get into evaluation codes folder './TEST/TEST_setting1'
```
cd ./TEST/TEST_setting1
```
We provide pretrained models in './checkpoints/'. Before testing, specify the model checkpoint in 'test.yml'.
```
DANET_CHECKPOINT: './checkpoints/setting1/setting1_ckp.tar'
```
Or change the path with your own absolute path
Run 'test.py'
```
python test.py
```

Get into training codes folder './TRAIN/TRAIN_setting1'
```
cd ./TRAIN/TRAIN_setting1
```
Specify the dataset path in 'train.yml'
```
data:
	clean_dir: './datasets/div2k'
```
Or change the path with your own dataset absolute path
Run 'train.py'
```
python train.py
```