PyTorch implementation of DualGAN

• Loss values are plotted using Tensorboard in PyTorch.

• Image size: 256x256

• RMSProp optimizer is used. Learning rate = 0.00005, batch size = 1, # of epochs = 45:
• 6 resnet blocks used for Generator.

GAN losses ( : Discriminator A / : Discriminator B : Generator A / : Generator B : L1 loss A / : L1 loss B )	Generated images (Input / Generated / Reconstructed)

• Generated images using test data

  Sketch to Photo 1st column: Input / 2nd column: Generated / 3rd column: Reconstructed
  Photo to Sketch 1st column: Input / 2nd column: Generated / 3rd column: Reconstructed

1. https://github.com/moono/GANs/tree/master/DualGAN
2. https://github.com/duxingren14/DualGAN
3. https://github.com/wiseodd/generative-models