WorldRecon reconstructs large-scale 3D scenes from unposed multi-view images, cutting Chamfer distance 43%
ZhejiangUniversity · hf · 2026-10-14
Zhejiang University with Tencent Hunyuan and Fudan present WorldRecon, a generative multi-view framework for complete large-scale 3D scene meshes from unposed RGB images:
- Uses a coarse noisy global point cloud and estimated poses as a scaffold, partitioned into overlapping spatial blocks.
- Equips native 3D generator TRELLIS.2 with reconstruction-based conditioning to produce sparse structure, detailed geometry and PBR materials per block.
- Neighbour-aware latent inpainting and causal world generation propagate context across blocks for scene coherence.
Results cut Chamfer distance by 14.9%/43.1% and improve F-score by 17.3%/12.0% over the strongest baselines on synthetic and real indoor scenes, with strong sim-to-real generalization.
More from Research
- 654 Real-World Failures: First Large-Scale Study Maps How RAG Configuration Issues Break Systems — _reachsumit · 2026-10-14
- ReMaRk Cuts GRPO Responder Noise in LLM Recommenders, Reaching Parity With 25–53% of Training Steps — _reachsumit · 2026-10-14
- Disney+ details DRanker transformer ranker, driving 1% engagement and 0.5% DAU lift — _reachsumit · 2026-10-14
- Math Papers Pile On OpenAI's Quasi-Riemann Proof, One Declares 'No AI Used' — LucaAmb · 2026-10-14
- Netflix compares autoregressive beam search vs multi-token prediction for LLM recommender Semantic IDs — _reachsumit · 2026-10-14
- Alibaba proposes UniSGR-V2, anticipating continuous target-side features in generative recommendation — _reachsumit · 2026-10-14