Exploration of Deep Generative Adversarial Networks in the Context of Image Style Transfer

Abstract :

An investigation into the design and training of two accessible GAN architectures:

  1. Deep Convolutional GAN original paper

  2. CycleGAN original paper

In this notebook I tackle a straightforward style transfer problem between Windows and Apple-style emojis, attempting to translate one style to the other in great detail, subject to limited processing power.

This work was inspired by this AI-awareness project.

Download the Colaboratory notebook here.