WorldRecon: Generative Multi-View Mesh Reconstruction for Large-Scale 3D World

Key points

• Reconstructing complete, high-quality meshes of large-scale scenes from unposed multi-view images requires faithful geometric grounding while plausibly completing surfaces that are unobserved or only sparsely captured.1
• WorldRecon uses a coarse, noisy global point cloud and camera poses estimated from the input images by a reconstruction model as a global scaffold for whole-scene generation and partitions it into overlapping spatial blocks.1
• To robustly leverage these imperfect geometric priors, we equip the native 3D generative model TRELLIS.2 with a dedicated reconstruction-based conditioning architecture that generates each block from its local point-cloud crop and block-relevant image observations, producing sparse structure, detailed geometry, and PBR materials.1
• To maintain scene-level coherence, we introduce neighbour-aware latent inpainting and a causal world-generation procedure that propagates structural, geometric, and material context across overlapping blocks.1

Sources (2)

• [1]WorldRecon: Generative Multi-View Mesh Reconstruction for Large-Scale 3D World
  Hugging Face Daily Papers · Oct 13, 12:00 AM

  We present WorldRecon, a generative multi-view reconstruction framework for recovering complete, large-scale 3D scene meshes from unposed RGB images.

  Reconstructing complete, high-quality meshes of large-scale scenes from unposed multi-view images requires faithful geometric grounding while plausibly completing surfaces that are unobserved or only sparsely captured.

• [2]WorldRecon: Generative Multi-View Mesh Reconstruction for Large-Scale 3D World
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 13, 05:57 PM · same content

Extractive summary: sentences quoted from the sources.