Research map / 3D vision / Depth and stereo
Stable3R: Streaming 3D Reconstruction with Stable Geometric References
Jincheng Xiong, Jiarong Han, Hang Zhang, Mu Xu, Ming Qian
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes Stable3R, a streaming feedforward 3D reconstruction method that uses stable geometric references to reduce the irreversible bias from early frames in KV-cache-based causal architectures.
Problem
Streaming 3D reconstruction built from offline feedforward models via causal attention and KV-caching suffers instability because early-frame bias is irreversible.
Approach
Addresses instability in KV-cache causal streaming reconstruction by introducing stable geometric references, rather than relying solely on early frames; the abstract attributes the instability to the causal architecture strictly prioritizing early frames and a second coupled factor not captured in the abstract.
Closest papers by meaning
- GHOST: Geometry-Hierarchical Online Streaming Token Eviction for Efficient 3D Reconstruction (NeurIPS 2026 · similarity 0.90)
- FrameVGGT: Coherence-Preserving Memory for Bounded Streaming Geometry (NeurIPS 2026 · similarity 0.90)
- CLoSeR: Closing the Loop for Long-Context Streaming Reconstruction (NeurIPS 2026 · similarity 0.90)
- Point3R: Streaming 3D Reconstruction with Explicit Spatial Pointer Memory (NeurIPS 2025 · similarity 0.90)
- Rethinking the State Update Gate for Long-Sequence Recurrent 3D Reconstruction (NeurIPS 2026 · similarity 0.89)
- BAT3R: Robust Online 3D Reconstruction with Bayesian Adaptive State Updates (NeurIPS 2026 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
