GAN-Control: Explicitly Controllable GANs

TL;DR

This paper introduces GAN-Control, a framework that enables explicit, fine-grained control over generated images' attributes across multiple domains by leveraging contrastive learning for disentangled latent spaces.

Contribution

It presents a novel method for explicit attribute control in GANs that is extendable beyond human faces and does not rely on morphable 3D models.

Findings

Achieves state-of-the-art control over face attributes like age, pose, and expression.

Demonstrates control in domains of painted portraits and dog images.

Outperforms existing methods both qualitatively and quantitatively.

Abstract

We present a framework for training GANs with explicit control over generated images. We are able to control the generated image by settings exact attributes such as age, pose, expression, etc. Most approaches for editing GAN-generated images achieve partial control by leveraging the latent space disentanglement properties, obtained implicitly after standard GAN training. Such methods are able to change the relative intensity of certain attributes, but not explicitly set their values. Recently proposed methods, designed for explicit control over human faces, harness morphable 3D face models to allow fine-grained control capabilities in GANs. Unlike these methods, our control is not constrained to morphable 3D face model parameters and is extendable beyond the domain of human faces. Using contrastive learning, we obtain GANs with an explicitly disentangled latent space. This disentanglement is utilized to train control-encoders mapping human-interpretable inputs to suitable latent vectors, thus allowing explicit control. In the domain of human faces we demonstrate control over identity, age, pose, expression, hair color and illumination. We also demonstrate control capabilities of our framework in the domains of painted portraits and dog image generation. We demonstrate that our approach achieves state-of-the-art performance both qualitatively and quantitatively.