Efficient 3D Articulated Human Generation with Layered Surface Volumes

TL;DR

This paper introduces layered surface volumes (LSVs), a novel 3D representation for articulated digital humans that combines the efficiency of template meshes with high-fidelity detail capture, enabling high-quality 3D human generation via GANs from 2D images.

Contribution

The paper proposes layered surface volumes (LSVs), a new 3D representation that improves detail and efficiency in GAN-based 3D human generation from single-view images.

Findings

LSVs naturally capture fine surface details like hair and accessories.

LSV-GAN produces high-quality, view-consistent 3D humans from 2D datasets.

The approach outperforms traditional template and volumetric methods in fidelity and speed.

Abstract

Access to high-quality and diverse 3D articulated digital human assets is crucial in various applications, ranging from virtual reality to social platforms. Generative approaches, such as 3D generative adversarial networks (GANs), are rapidly replacing laborious manual content creation tools. However, existing 3D GAN frameworks typically rely on scene representations that leverage either template meshes, which are fast but offer limited quality, or volumes, which offer high capacity but are slow to render, thereby limiting the 3D fidelity in GAN settings. In this work, we introduce layered surface volumes (LSVs) as a new 3D object representation for articulated digital humans. LSVs represent a human body using multiple textured mesh layers around a conventional template. These layers are rendered using alpha compositing with fast differentiable rasterization, and they can be interpreted as a volumetric representation that allocates its capacity to a manifold of finite thickness around the template. Unlike conventional single-layer templates that struggle with representing fine off-surface details like hair or accessories, our surface volumes naturally capture such details. LSVs can be articulated, and they exhibit exceptional efficiency in GAN settings, where a 2D generator learns to synthesize the RGBA textures for the individual layers. Trained on unstructured, single-view 2D image datasets, our LSV-GAN generates high-quality and view-consistent 3D articulated digital humans without the need for view-inconsistent 2D upsampling networks.