Leveraging StyleGAN on custom images

In Chapter 11, we learned about neural style transfer. We generated an image by blending the style of one image with the content of another image. However, what if we want to create a younger version of a person in a picture or add certain attributes to an image, such as glasses? StyleGAN can do this. Let’s learn how in the following sections.

The evolution of StyleGAN

Let’s first look at a few developments prior to the invention of StyleGAN. As we know, generating fake faces (as we saw in the previous chapter) involves the usage of GANs. The biggest problem that research faced was that the images that could be generated were small (typically 64 x 64). Any effort to generate larger images caused the generators or discriminators to fall into local minima, which would stop training and generate gibberish. One of the major leaps in generating high-quality images appeared in a research paper that proposed Progressive GAN (ProGAN...