Super-Resolution

In this repository, I implemented SRGAN for MNIST dataset.

There are four modes :

• MSE
• GAN
• WGAN
• visual (plot the generated HR images given trained model)

More works

• See images folder for examples of generated images.
• See SR-CIFAR for network on CIFAR10
• See SR-FULL for full version of SRGAN

To be done :

• plot outputs by different networks together
• full-sized networks
• VGG loss
• benchmark on different sets

Other implementation

• torch
• another torch
• tensorflow
• paper reading list
• another paper reading list