Downoad the dataset from here, it consist of about 202k images. After downloading, move that folder in dataset foler.
Once the dataset is downloaded and moved, run:
python3 SRGAN.py
It take 5000 examples for training, one can change it to any value under 202k
Depending upon GPU, one can vary batch size. Since I had single GPU. I trained it on batch size of 16
For testing the network, just run :
python3 Test.py
It take last five images from the dataset,to produce the SR images. If anyone want to test more images one can chage it to another value
Left one is Downsampled image, Middle is the SR image produced by Generaor, and right is th HR image or the truth truth
