Unpaired Image to Image Translation with CycleGAN

Yann LeCun, director of AI research at Facebook and Professor at NYU described Generative Adversarial Networks, GANs as the most interesting idea in Machine Learning in the last 10 years. Since the invention of GANs in 2014 by Ian Goodfellow, we’ve seen a ton of variants of these interesting neural networks from several research groups like NVIDIA and Facebook but we are going to look at one from a research group at UC Berkeley called the Cycle Consistent Adversarial Network. Before we dive into a Cycle Consistent Adversarial network, CycleGAN for short, we are going to look at what a Generative Adversarial Network is. This article is intended to give insights into the working mechanism of a Generative Adversarial Network and one of its popular variants, the Cycle Consistent Adversarial Network. Most of the code used here was taken from the official TensorFlow documentation page. Full code for this article can be obtained from : https://www.tensorflow.org/beta/tutorials/generative/cyclegan

This is a companion discussion topic for the original entry at https://blog.paperspace.com/unpaired-image-to-image-translation-with-cyclegan/