Chapter 11 - Generative adversarial network

Abstract

Generative Adversarial Network (GAN) has shown remarkable results in many computer vision tasks such as image generation, text generation, text-to-image generation, image-to-image translation, and so on. This chapter discusses the structure and principle of GAN. Next, it introduces and explains Wasserstein distance and the Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP), which is an improved version of the GAN. Finally, the chapter presents an example program of building, training, and testing the WGAN-GP model on the large-scale CelebFaces Attributes (CelebA) Dataset.