G$^2$SR：基于几何方法的快速且内存高效的高斯曲面重建
G$^2$SR: Geometric Methods for Fast and Memory-Efficient Gaussian-based Surface Reconstruction
浏览论文内容
中文总结 AI 辅助
研究少视图曲面重建问题，提出G2SR方法，利用跨视图2D平铺对应关系，通过轻量级神经前端和解析后端，实现快速且内存高效的高斯曲面重建，在多个数据集上几何精度高且内存占用小。
中文摘要 AI 辅助
少视图曲面重建从少量有姿态的RGB图像中恢复场景的可见表面，为机器人在线探索和交互提供3D模型。在移动平台上，重建必须快速、几何精确且内存占用小。3D高斯平铺（3DGS）提供高保真场景表示，但从少视图构建存在不适定性。端到端方法通过大型网络回归平铺解决模糊性，但计算和内存需求大且泛化性差。我们提出G2SR，利用任务的适定核心，通过跨视图2D平铺对应关系解析得出3D平铺。G2SR采用轻量级神经前端检测和跟踪图像平面上的2D高斯平铺，并使用解析后端将每个平铺三角测量为度量尺度的3D平铺。在ScanNet、Replica和DTU上，G2SR在几何精度上匹配或超过最先进的端到端方法，同时在384x512分辨率下，对于2视图和3视图输入，在203MB的GPU内存内每秒可进行69 - 89次重建（少5 - 107倍），为基于高斯的在线曲面重建提供了实用途径。
英文摘要
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.
发表机构
- Massachusetts Institute of Technology（麻省理工学院）
机构由 AI 辅助整理，请以论文原文为准。
