/SRDDN-TIE

The code of "Image Super-Resolution via Dense Discriminative Network"

Primary LanguagePython

SRDDN-TIE

The implementation of the following paper in Pytorch

Ma J, Wang X, Jiang J. Image Super-Resolution via Dense Discriminative Network[J]. IEEE Transactions on Industrial Electronics, 2019.

Dependencies

Python 3.6  PyTorch >= 1.0.0  numpy  skimage  imageio  matplotlib  tqdm

Training data

1.Download DIV2K training data (800 training + 100 validtion images) from https://data.vision.ee.ethz.ch/cvl/DIV2K/
2.Specify '--dir_data' based on the HR and LR images path. in option.py, 
	  '--ext' is set as 'sep_reset', which first convert .png to .npy

For training

Cd to 'TrainCode/code
for example
CUDA_VISIBLE_DEVICES=0  python main.py --model SRDDN --save SRDDNX2 --scale 2  --n_resblocks 12 --n_feats 64  --reset --chop --save_results --print_model --patch_size 96

you can change the settings found in option.py

For testing

Download the pre-trained model(SRDDN/x2/x3/x4) in https://pan.baidu.com/s/15sMaYQ3ODUZfwW3jf1aHCA  
password：wau0  
Put it in 'TestCode/model' file.
Cd to 'TestCode/code'  
run python main.py --data_test MyImage --scale 4 --model SRDDN --n_resblocks 12 --n_feats 64 --pre_train ../model/model_x4.pt --test_only --save_results --chop --save 'SRDDNX4' --testpath ../LR/LRBI --testset Set5  
Then you can find the SR result in 'TestCode/SR'. 
Run 'Evaluate_PSNR_SSIM.m'in MATLAB to obtain PSNR/SSIM values for paper