Part 2 explores a selection of advanced topics in GANs. Building on the foundational concepts from part 1, you will deepen your theoretical understanding of GANs and expand your practical toolkit of GAN implementations:
- Chapter 5 covers many of the theoretical and practical hurdles to training GANs and how to overcome them.
- Chapter 6 presents a groundbreaking training methodology called Progressive GAN that has enabled GANs to synthesize images with unprecedented resolution.
- Chapter 7 covers the use of GANs in semi-supervised learning (methods of training classifiers with only a small fraction of labeled examples), an area of immense practical importance.
- Chapter 8 introduces the Conditional GAN, a technique that enables targeted data generation by using labels (or other conditioning information) while training the Generator and Discriminator.
- Chapter 9 explores the CycleGAN, a general-purpose technique for image-to-image translation—turning one image (such as a photo of an apple) into another (such as a photo of an orange).