World Tracing: New method generates 3D geometry beyond visible pixels

Researchers have developed a new generative representation called World Tracing, which aims to improve image-to-3D generation by aligning predicted 3D geometry with visible pixels while also completing occluded surfaces. This method uses a diffusion transformer model, WT-DiT, that treats multiple geometry layers as denoising tokens. Trained with pixel-space flow matching, World Tracing demonstrates superior performance in both visible-surface reconstruction and complete geometry generation across various benchmarks, outperforming existing depth estimators and image-to-3D generators. The approach also facilitates applications like text-driven 3D scene editing and novel-view video synthesis. AI

IMPACT This new method could enhance 3D content creation by enabling more accurate and complete geometry generation from 2D images, impacting fields like virtual reality and game development.

RANK_REASON The cluster contains a research paper detailing a new method for 3D geometry generation from images. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• World Tracing
• WT-DiT

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →