SRGAN is an image super-resolution method, our code is inspired the paper (https://arxiv.org/pdf/1609.04802.pdf).

oneflow==0.1.10

Download VOC2012 dataset here (about 1.8G), unzip and put it in "./data". Preprocess images to generate *.npy .
python of_data_utils

If you do not want to preprocess images, you can directly download *.npy here, unzip and put it in "./data".

To use the perceptual loss, you can download perceptual model(vgg16bn) here, unzip and put it in "./models". python of_srgan.py

If you just want to run test, you can download pretrained model here, unzip and put it in "./models".

You can test an image or images with the same or different shape. python of_srgan.py --test