Interpreting the Latent Space of GANs for Semantic Face Editing

TL;DR

This paper introduces InterFaceGAN, a framework for understanding and manipulating the semantic features in GAN latent spaces to achieve precise and controllable face editing, including attributes like gender, age, expression, and pose.

Contribution

The paper reveals that GANs learn a disentangled latent space for face attributes and proposes a method to manipulate these attributes through linear transformations.

Findings

Disentangled semantic representations are learned in GAN latent spaces.

Linear transformations can control facial attributes precisely.

The method enables realistic face editing and artifact correction.

Abstract

Despite the recent advance of Generative Adversarial Networks (GANs) in high-fidelity image synthesis, there lacks enough understanding of how GANs are able to map a latent code sampled from a random distribution to a photo-realistic image. Previous work assumes the latent space learned by GANs follows a distributed representation but observes the vector arithmetic phenomenon. In this work, we propose a novel framework, called InterFaceGAN, for semantic face editing by interpreting the latent semantics learned by GANs. In this framework, we conduct a detailed study on how different semantics are encoded in the latent space of GANs for face synthesis. We find that the latent code of well-trained generative models actually learns a disentangled representation after linear transformations. We explore the disentanglement between various semantics and manage to decouple some entangled semantics with subspace projection, leading to more precise control of facial attributes. Besides manipulating gender, age, expression, and the presence of eyeglasses, we can even vary the face pose as well as fix the artifacts accidentally generated by GAN models. The proposed method is further applied to achieve real image manipulation when combined with GAN inversion methods or some encoder-involved models. Extensive results suggest that learning to synthesize faces spontaneously brings a disentangled and controllable facial attribute representation.