MeshGAN: Non-linear 3D Morphable Models of Faces

TL;DR

MeshGAN introduces a novel GAN architecture that directly generates high-quality 3D face meshes, overcoming previous limitations of volumetric methods and enabling realistic 3D face synthesis with diverse identities and expressions.

Contribution

This paper presents the first intrinsic GAN architecture for 3D meshes, directly operating on mesh data for realistic 3D face generation.

Findings

MeshGAN can generate high-fidelity 3D faces.

The method produces diverse identities and expressions.

Quantitative and qualitative results validate effectiveness.

Abstract

Generative Adversarial Networks (GANs) are currently the method of choice for generating visual data. Certain GAN architectures and training methods have demonstrated exceptional performance in generating realistic synthetic images (in particular, of human faces). However, for 3D object, GANs still fall short of the success they have had with images. One of the reasons is due to the fact that so far GANs have been applied as 3D convolutional architectures to discrete volumetric representations of 3D objects. In this paper, we propose the first intrinsic GANs architecture operating directly on 3D meshes (named as MeshGAN). Both quantitative and qualitative results are provided to show that MeshGAN can be used to generate high-fidelity 3D face with rich identities and expressions.