Skip to content

lordtt13/GAN-s

Repository files navigation

GAN

Description:

All GAN models present here have been taken either from the book by Rowel Atienza "Advanced Deep Learning with Keras" or various web sources on Advanced GAN Models

Models Added:

  • gan - Builder Script for Higher Models
  • Vanilla GAN - Simple Convolutional GAN for mnist Data
  • Deep Convolutional GAN - Deep Convolutional GAN
  • Wasserstein GAN - Higher GAN which uses Wasserstein Loss Function
  • Conditional GAN - Conditional GAN's where you can give an additional input of label and get the desired result
  • Least Squares GAN - Higher GAN, more stable, Uses MSE loss
  • Info GAN - Disentagled GAN,used to differentiate and generate between features in the generated images
  • Stacked GAN - Hybridized Disentangled GAN,uses Enocders to build features that are used for generation
  • Auxiliary Conditional GAN - Auxiliary Conditional GAN, same as cgan with I/O different
  • Pix2Pix - Instance Based Cycle GAN, used for conditioning Noise(or other images) onto Specific Images
  • Cycle GAN - Non mapping GAN, which can be used to learn representations between two sets of images, (~ Style Transfer)

More Models will be added soon!!!