Summary

In this chapter, we first looked at what GANs are. They are a new kind of generative model that helps us to generate new images. 

We also touched upon other kinds of generative models, such as Variational Auto-encoders and PixelRNN, to get an overview of different kinds of generative models. We also talked about different kinds of GANs to discuss the progress that had been made in this space since the first paper on GANs was published in 2014.

Then, we learned about DiscoGANs, a new type of GAN that can help us to learn about cross- domain relationships. Specifically, in this chapter, our focus was on building a model to generate handbag images from shoes and vice versa.

Finally, we learned about the architecture of DiscoGANs and how they differ from usual GANs.

In the next chapter, we will learn how to implement capsule networks on the Fashion MNIST dataset.