Reproduce “UltraTex: Aligning Physical Knowledge for High-Fidelity PBR Texture Generation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Jingzhi Bao, Xiaohe Ma, Hong Li, Yuanming Hu, Xiaoguang Han

UltraTex estimates reusable physically based material maps for a 3D mesh from a reference image. It generates albedo and surface-detail cues across views, then predicts metallic and roughness using semantic tokens and features aligned to controlled-light material renderings. On 152 held-out Objaverse-XL objects at 1024², its metallic and roughness MSEs were 46.77 and 16.57; stylized or transparent assets remain ambiguous.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.