This is as close as possible an implementation of the paper Image Super-Resolution Using very deep Residual Channel Attention Networks. Link to the paper - https://arxiv.org/abs/1807.02758.
I have only implemented perceptual loss and left out adversarial loss. Also, 2 layers have been shaved off due to hardware contraints. I also have added a simple 10 layered autoencoder approach.
Train the model