Image super-resolution (restoration of rich details in a low resolution image)
Supports Keras with Tensorflow backend.
By default GPU use(recommended)
For CPU only please uncomment in main_dirpath.py 3 rd line
#os.environ['CUDA_VISIBLE_DEVICES'] = '-1'
Model weights http://epsilonsys.com/weights025-17-0.93.rar
Train data https://data.vision.ee.ethz.ch/cvl/DIV2K/
The model weights can be downloaded from http://epsilonsys.com/weights025-17-0.93.rar to /weights_Double/ folder :
python main_dirpath.py "imgpath", where imgpath is a full path to the images folder.
SSIM 0.904
Best SSIM (NTIRE 2017) 0.901
SSIM 0.879
Best SSIM (NTIRE 2017) 0.874
Official SSIM score from NTIRE 2017 :
http://www.vision.ee.ethz.ch/~timofter/publications/Agustsson-CVPRW-2017suppl.pdf