Unpaired Image to Image Translation Using CycleGAN - An Introduction

Explore the fundamentals of unpaired image-to-image translation using CycleGAN in this introductory video tutorial. Delve into the key details from the original paper, including the PatchGAN discriminator architecture, generator description, and training process. Learn about the instance normalization layer and its role in standardizing feature map outputs. Gain insights into the combined/composite model structure and understand the core concepts behind this innovative approach to image translation. Note that while this tutorial focuses on theoretical concepts, it provides a solid foundation for implementing CycleGAN in Keras, which is covered in a subsequent tutorial.

Unpaired image to image translation Using CycleGAN
Walkthrough - Key details from the original paper
Discriminator architecture (PatchGAN)
Generator description
Generator architecture
Generator training
Combined / composite model