Reproduce “TEXGen: a Generative Diffusion Model for Mesh Textures” 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
Xin Yu, Ze Yuan, Yuan-Chen Guo, Ying-Tian Liu, JianHui Liu, Yangguang Li, Yan-Pei Cao, Ding Liang and 1 more
Researchers developed TEXGen, a generative diffusion model that produces high-resolution textures for 3D meshes by training directly in UV texture space. The model's hybrid 2D-3D architecture generates more detailed and realistic textures with improved global consistency and faster inference compared to existing methods, supporting versatile text and image-guided synthesis.
Key insights
- Training a diffusion model directly in UV texture space allows for efficient, feed-forward texture generation without per-object optimization.
- A hybrid 2D-3D architecture, combining 2D UV processing with 3D point cloud operations, effectively balances capturing fine local texture details with maintaining global consistency across the 3D mesh.
- Leveraging text and image conditioning alongside classifier-free guidance enables versatile and controllable texture synthesis for diverse 3D models.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.