Recently, Generative Adversarial Models (GANs) have made significant progress in image synthesis. In this talk, I will present our recent paper “Encoding in Style: a StyleGAN Encoder for Image-to-Image Translation”. The paper suggests a new method for image-to-image translation, focusing on the domain of human faces. We propose a method that utilizes StyleGAN, the state-of-the-art unconditional GAN, in order to perform various tasks such as generating a realistic face image from a sketch, face frontalization, super resolution, etc. Furthermore, our framework can be used to invert a real image into the latent space of StyleGAN. By doing so, it is possible to employ the great latent space of StyleGAN in order to edit real images.

Project page: https://eladrich.github.io/pixel2style2pixel/