Poincar3: multi-view self-distillation yields emergent 3D geometry, code on GitHub
rsasaki0109 · x · 2026-11-07
Chalmers、IP Paris、Linköping、波尔多大学等机构的研究者开源了 Poincar3，一种通过多视角自蒸馏（multi-view self-distillation）做 3D 自监督学习的方法，代码采用 MIT 协议发布在 GitHub（目前已获 50 星）。作者称仅凭 Poincar3 的注意力图就能可靠地建立跨视角匹配并生成轨迹，匹配能力完全从自监督中涌现，无需显式监督。项目动机引彭加勒的观点——没有运动的生物无法理解三维空间，因此强调利用多视角数据。
More from Research
- DAS Dex 2.0 data glove rebuilt around magnetic encoders for precise finger-joint tracking — micoolcho · 2026-11-07
- Computer-Using World Model Lets GUI Agents Simulate Actions Before Executing Them — burkov · 2026-11-07
- LLM inference flips between compute-bound and memory-bound within a single request — techNmak · 2026-11-07
- TLAPS-Bench: agentic AI formal verification uncovers read consistency bug in SlateDB's WAL — tianyin_xu · 2026-11-07
- Hot take: 'continual learning' is just 'learning' — vincent_koc · 2026-11-07
- Stanford's AI-discovered peptide BRP cuts food intake by up to 50% in animals — emmanuelvivier · 2026-11-07