This is a set of PyTorch implementation of latest research paper: Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network and Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network.
The implementation is roughly verified its correctness on small dataset. Using larger dataset such as ImageNet DET data, the quality of the product will be better, but according to the original paper the boost by using large dataset is not that big.
TODO:
- Much intensive check for the reproducibility with large trainingset will be done some time soon.
- Trainined model will be uploaded once training on my side has done.
Currently the outputs look differently from the orginal paper so it has to be inspected more.