MatLat: Material Latent Space for PBR Texture Generation

TL;DR

MatLat introduces a novel fine-tuning approach for pretrained latent image models to generate consistent, high-quality PBR textures on 3D meshes, addressing dataset scarcity and distribution shift issues.

Contribution

The paper presents a new material latent space, MatLat, with a fine-tuning method that preserves locality and improves cross-view consistency in PBR texture generation.

Findings

Enhanced texture fidelity demonstrated through ablation studies

Effective incorporation of new material channels with minimal distribution shift

State-of-the-art performance in PBR texture quality

Abstract

We propose a generative framework for producing high-quality PBR textures on a given 3D mesh. As large-scale PBR texture datasets are scarce, our approach focuses on effectively leveraging the embedding space and diffusion priors of pretrained latent image generative models while learning a material latent space, MatLat, through targeted fine-tuning. Unlike prior methods that freeze the embedding network and thus lead to distribution shifts when encoding additional PBR channels and hinder subsequent diffusion training, we fine-tune the pretrained VAE so that new material channels can be incorporated with minimal latent distribution deviation. We further show that correspondence-aware attention alone is insufficient for cross-view consistency unless the latent-to-image mapping preserves locality. To enforce this locality, we introduce a regularization in the VAE fine-tuning that crops latent patches, decodes them, and aligns the corresponding image regions to maintain strong pixel-latent spatial correspondence. Ablation studies and comparison with previous baselines demonstrate that our framework improves PBR texture fidelity and that each component is critical for achieving state-of-the-art performance.