{"slug": "revisiting-local-context-for-long-horizon-streaming-3d-reconstruction", "title": "Revisiting Local Context for Long-Horizon Streaming 3D Reconstruction", "summary": "ABot-Recon, a streaming 3D reconstruction model introduced in an arXiv paper (arXiv:2608.27529v1), achieves an ATE of 4.35 m and an RPE-R of 0.12° on the Oxford Spires benchmark, reducing both errors by approximately 40% relative to the best prior results. The model caches KV features from only the preceding 11 frames, keeping the learned temporal state strictly local to maintain long-horizon stability.", "body_md": "arXiv:2608.27529v1 Announce Type: new\nAbstract: Streaming 3D reconstruction from extremely long videos requires estimating camera motion and scene geometry online under bounded memory and computation. Early streaming models achieve causal, bounded-cost inference using finite context buffers or compact recurrent states, yet their estimates often deteriorate as sequences grow. Recent methods improve long-horizon stability by coupling short-range context with persistent or multi-level long-range memory. We pursue a different route: we keep the learned temporal state strictly local and formulate predictions whose targets remain independent of sequence length. We present ABot-Recon, a simple streaming model that caches KV features from only the preceding 11 frames. It predicts a point map in the current camera coordinate system together with an adjacent-frame relative pose. These predictions remain equivariant under changes of reference frame, and global poses and geometry are recovered through sequential composition. To reduce accumulated drift, a lightweight temporal refiner improves relative rotations using recent visual and motion context, while a composition-aware pose loss supervises multi-step pose composition. Extensive evaluations on challenging long-sequence benchmarks demonstrate the superior long-horizon performance of our local-context approach. On Oxford Spires, ABot-Recon achieves an ATE of 4.35 m and an RPE-R of $0.12^\\circ$, reducing both errors by approximately 40\\% relative to the best prior results.", "url": "https://wpnews.pro/news/revisiting-local-context-for-long-horizon-streaming-3d-reconstruction", "canonical_source": "https://arxiv.org/abs/2608.27529", "published_at": "2026-08-31 04:00:00+00:00", "updated_at": "2026-08-31 04:22:38.024346+00:00", "lang": "en", "topics": ["computer-vision", "machine-learning", "artificial-intelligence"], "entities": ["ABot-Recon", "Oxford Spires", "arXiv"], "alternates": {"html": "https://wpnews.pro/news/revisiting-local-context-for-long-horizon-streaming-3d-reconstruction", "markdown": "https://wpnews.pro/news/revisiting-local-context-for-long-horizon-streaming-3d-reconstruction.md", "text": "https://wpnews.pro/news/revisiting-local-context-for-long-horizon-streaming-3d-reconstruction.txt", "jsonld": "https://wpnews.pro/news/revisiting-local-context-for-long-horizon-streaming-3d-reconstruction.jsonld"}}