This Book Does Not Exist 2 is a redone version of This Book Does Not Exist. It utilises Generative Adversial Neural Networks to create new book covers.
Note: The video may not autoplay on some mobile devices.
A Generative Adversial Network or GAN for short is a type of AI that uses unsupervised learning to generate images.
Essentially, a GAN has two different AI both trying to outsmart the other, the generator and the discriminator.
The generator looks at input images, which in this case were book covers, and tries to create similar images.
The discriminator looks at these images and makes an assessment as to whether the images the generator made are real or fake.
If the discriminator classifies it as fake, the generator learns and tries again, making better results each time it tries.
Eventually, it gets to a point where the generator creates an image the discriminator classifies as real.
The GAN I used for this project was HyperGAN. Currently I am working on a version in StyleGAN.