{"slug": "g-2-sr-geometric-methods-for-fast-and-memory-efficient-gaussian-based-surface", "title": "G$^2$SR: Geometric Methods for Fast and Memory-Efficient Gaussian-based Surface Reconstruction", "summary": "Researchers propose G2SR, a geometric method for fast and memory-efficient Gaussian-based surface reconstruction from few RGB views. It uses a lightweight neural frontend for 2D splat detection and an analytic backend for 3D triangulation, achieving 69-89 reconstructions per second with 5-107x less GPU memory than end-to-end methods while matching geometric accuracy.", "body_md": "arXiv:2607.14470v1 Announce Type: new\nAbstract: Few-view surface reconstruction recovers the visible surfaces of a scene from a few posed RGB images, providing the 3D models that robots need to explore and interact online. On mobile platforms, the reconstruction must be fast and geometrically accurate while keeping a small memory footprint to ensure safe and efficient operation. 3D Gaussian Splatting (3DGS) offers a high-fidelity scene representation, but building it from a few views is ill-posed, as many distinct surfaces reproduce the same images, making traditional photometric methods prone to \"floater\" artifacts. End-to-end methods resolve the ambiguity by regressing splats with large, usually Transformer-based, networks that require heavy compute and memory while generalizing poorly to new scenes. We propose G2SR, which exploits a well-posed core of the task: given cross-view 2D splat correspondences, 3D splats follow analytically from multi-view geometry. G2SR employs a lightweight neural frontend to detect and track 2D Gaussian splats on the image plane and an analytic backend to triangulate each into a metric-scale 3D splat. On ScanNet, Replica, and DTU, G2SR matches or exceeds the geometric accuracy of state-of-the-art end-to-end methods while running at 69-89 reconstructions per second within 203 MB of GPU memory (5-107x less) for 2- and 3-view inputs at 384 x 512 resolution, offering a practical path to online Gaussian-based surface reconstruction.", "url": "https://wpnews.pro/news/g-2-sr-geometric-methods-for-fast-and-memory-efficient-gaussian-based-surface", "canonical_source": "https://arxiv.org/abs/2607.14470", "published_at": "2026-07-17 04:00:00+00:00", "updated_at": "2026-07-17 04:08:16.708152+00:00", "lang": "en", "topics": ["computer-vision", "machine-learning", "artificial-intelligence"], "entities": ["ScanNet", "Replica", "DTU"], "alternates": {"html": "https://wpnews.pro/news/g-2-sr-geometric-methods-for-fast-and-memory-efficient-gaussian-based-surface", "markdown": "https://wpnews.pro/news/g-2-sr-geometric-methods-for-fast-and-memory-efficient-gaussian-based-surface.md", "text": "https://wpnews.pro/news/g-2-sr-geometric-methods-for-fast-and-memory-efficient-gaussian-based-surface.txt", "jsonld": "https://wpnews.pro/news/g-2-sr-geometric-methods-for-fast-and-memory-efficient-gaussian-based-surface.jsonld"}}