A Style-Based Generator Architecture for Generative Adversarial Networks (GAN)
StyleGAN is a type of generative adversarial network which uses a specialised generator architecture. The architecture borrows from the style transfer literature, and makes use of adaptive instance normalization. This allows it to create artificial images that look like authentic photographs.
A common application of this technology is the generation of artificial pictures of people's faces, using an existing dataset of famous faces as a reference.