In this chapter, we will explain one of the most interesting deep learning models, Generative Adversarial Networks (GANs). We will start by reviewing what GANs are and what they are used for. After briefly covering the evolution paths of GAN models, we will illustrate a variety of GAN architectures, along with image generation examples.
Imagine you are in a competition of mimicking an artwork (such as Vincent van Gogh's The Starry Night) that you don't know enough about initially. You are allowed to participate as many times as you wish. And every time you submit your entry, the judge gives you feedback on what the real artwork looks like and how close your replica is. In the first few trials, your work does not score high, owing to your very limited knowledge of the original piece. After a few trails, your submissions are getting closer...