Efficient Geometry-aware 3D Generative Adversarial Networks

TL;DR

This paper presents a novel 3D GAN architecture that enhances computational efficiency and image quality, enabling real-time, multi-view-consistent image and 3D shape synthesis from single-view photographs.

Contribution

It introduces a hybrid explicit-implicit network design that leverages 2D CNN generators like StyleGAN2 for efficient, high-quality 3D-aware image synthesis.

Findings

Achieves real-time, multi-view-consistent image generation

Produces high-quality 3D geometry from 2D images

Demonstrates state-of-the-art results on FFHQ and AFHQ Cats datasets

Abstract

Unsupervised generation of high-quality multi-view-consistent images and 3D shapes using only collections of single-view 2D photographs has been a long-standing challenge. Existing 3D GANs are either compute-intensive or make approximations that are not 3D-consistent; the former limits quality and resolution of the generated images and the latter adversely affects multi-view consistency and shape quality. In this work, we improve the computational efficiency and image quality of 3D GANs without overly relying on these approximations. We introduce an expressive hybrid explicit-implicit network architecture that, together with other design choices, synthesizes not only high-resolution multi-view-consistent images in real time but also produces high-quality 3D geometry. By decoupling feature generation and neural rendering, our framework is able to leverage state-of-the-art 2D CNN generators, such as StyleGAN2, and inherit their efficiency and expressiveness. We demonstrate state-of-the-art 3D-aware synthesis with FFHQ and AFHQ Cats, among other experiments.