Convolutional Generation of Textured 3D Meshes

TL;DR

This paper introduces a novel framework that generates textured 3D meshes from 2D images using differentiable rendering and 2D convolutional GANs, enabling improved 3D reasoning and control.

Contribution

It presents a new method for generating textured 3D meshes from single-view images by encoding meshes and textures as 2D representations aligned with 2D GANs.

Findings

Effective generation of textured 3D meshes from 2D supervision

Works well on Pascal3D+ Cars and CUB datasets

Supports unconditional and conditioned generation with labels and text

Abstract

While recent generative models for 2D images achieve impressive visual results, they clearly lack the ability to perform 3D reasoning. This heavily restricts the degree of control over generated objects as well as the possible applications of such models. In this work, we bridge this gap by leveraging recent advances in differentiable rendering. We design a framework that can generate triangle meshes and associated high-resolution texture maps, using only 2D supervision from single-view natural images. A key contribution of our work is the encoding of the mesh and texture as 2D representations, which are semantically aligned and can be easily modeled by a 2D convolutional GAN. We demonstrate the efficacy of our method on Pascal3D+ Cars and CUB, both in an unconditional setting and in settings where the model is conditioned on class labels, attributes, and text. Finally, we propose an evaluation methodology that assesses the mesh and texture quality separately.