Overview: Evaluating Our First GAN

The evaluation of GANs is a very active and exciting research field in generative models. Evaluation is the procedure through which we estimate the quality of our model and the samples produced with it. In this chapter, we will learn how to use different methods to evaluate the quality and variety of the GAN samples we produced in the chapter “Implementing Our First GAN.”

We will learn about the challenges involved in evaluating GAN samples. We will also understand and learn to implement metrics for image quality. Furthermore, we will learn about using the birthday paradox to evaluate sample variety.

Topics covered in this chapter

The following topics will be covered in this chapter:

• The evaluation of GANs

• Qualitative methods

• Quantitative methods

• GANs and the birthday paradox