DualMat: PBR Material Estimation via Coherent Dual-Path Diffusion

TL;DR

DualMat introduces a dual-path diffusion framework that accurately estimates PBR materials from single images under complex lighting, combining pretrained visual knowledge and specialized latent spaces for superior results.

Contribution

It proposes a novel dual-path diffusion approach with feature distillation and efficient inference, advancing single-image PBR material estimation.

Findings

Achieves up to 28% improvement in albedo estimation.

Reduces metallic-roughness prediction errors by 39%.

Extends to high-resolution and multi-view inputs for 3D integration.

Abstract

We present DualMat, a novel dual-path diffusion framework for estimating Physically Based Rendering (PBR) materials from single images under complex lighting conditions. Our approach operates in two distinct latent spaces: an albedo-optimized path leveraging pretrained visual knowledge through RGB latent space, and a material-specialized path operating in a compact latent space designed for precise metallic and roughness estimation. To ensure coherent predictions between the albedo-optimized and material-specialized paths, we introduce feature distillation during training. We employ rectified flow to enhance efficiency by reducing inference steps while maintaining quality. Our framework extends to high-resolution and multi-view inputs through patch-based estimation and cross-view attention, enabling seamless integration into image-to-3D pipelines. DualMat achieves state-of-the-art performance on both Objaverse and real-world data, significantly outperforming existing methods with up to 28% improvement in albedo estimation and 39% reduction in metallic-roughness prediction errors.