Image-Super-Resolution

Enhancing Image Resolution with SRGAN

To Start Training Run:

python start_training_cars_dataset.py

What is Run?

Generator Network (SR-ResNet) is first trained using MSE (Mean-Squared Error) Loss then trained as an GAN with perceptual loss (vgg features) and BCE (Binary Cross Entropy) Loss with a Discriminator Network

Memory Requirements are Heavy for Generator Network, Discriminator Network and VGG19 Network

Implemented from Paper: https://arxiv.org/pdf/1609.04802.pdf