Discriminator Synthesis: On reusing the other half of Generative Adversarial Networks

TL;DR

This paper introduces Discriminator Dreaming, a novel method that leverages the trained discriminator in GANs to modify or generate images, expanding its utility beyond training stabilization.

Contribution

It presents a new approach to reuse the discriminator in GANs for image editing and synthesis, which was previously underexplored.

Findings

Discriminator features can be used for image editing.

Discriminator can generate images from scratch.

The method enhances the versatility of GANs.

Abstract

Generative Adversarial Networks have long since revolutionized the world of computer vision and, tied to it, the world of art. Arduous efforts have gone into fully utilizing and stabilizing training so that outputs of the Generator network have the highest possible fidelity, but little has gone into using the Discriminator after training is complete. In this work, we propose to use the latter and show a way to use the features it has learned from the training dataset to both alter an image and generate one from scratch. We name this method Discriminator Dreaming, and the full code can be found at https://github.com/PDillis/stylegan3-fun.