Generative Adversarial Networks have become quite popular due to their wide variety of applications in the fields of Computer Vision, Digital Marketing, Creative artwork and so on. One key challenge with GANs is that they are very difficult to train.
This book is a comprehensive guide that highlights the common challenges of training GANs and also provides guidelines for developing GANs in such a way that they result in stable training and high-quality results. This book also explains the generative learning approach of training ML models and its key differences from the discriminative learning approach. After covering the different generative learning approaches, this book deeps dive more into the Generative Adversarial Network and their key variants.
This book takes a hands-on approach and implements multiple generative models such as Pixel CNN, VAE, GAN, DCGAN, CGAN, SGAN, InfoGAN, ACGAN, WGAN, LSGAN, WGAN-GP, Pix2Pix, CycleGAN, SRGAN, DiscoGAN, CartoonGAN, Context Encoder and so on. It also provides a detailed explanation of some advanced GAN variants such as BigGAN, PGGAN, StyleGAN and so on. This book will make you a GAN champion in no time.
What will you learn
Learn about the generative learning approach of training ML models
Understand key differences of the generative learning approach from the discriminative learning approach
Learn about various generative learning approaches and key technical aspects behind them
Understand and implement the Generative Adversarial Networks in details
Learn about some key challenges faced during GAN training and two common training failure modes
Build expertise in the best practices and guidelines for developing and training stable GANs
Implement multiple variants of GANs and verify their results on your own datasets
Learn about the adversarial examples, some key applications of GANs and common evaluation strategies