G$^2$SR: Geometric Methods for Fast and Memory-Efficient Gaussian-based Surface Reconstruction
Quick Answer
G2SR introduces a lightweight neural approach for Gaussian-based surface reconstruction, achieving 69-89 reconstructions per second with 203 MB GPU memory.
Quick Take
G2SR introduces a lightweight neural approach for Gaussian-based surface reconstruction, achieving 69-89 reconstructions per second with 203 MB GPU memory. It matches or exceeds state-of-the-art methods in geometric accuracy on datasets like ScanNet and Replica, while addressing the challenges of few-view scenarios without heavy computational demands.
Key Points
- G2SR operates with 203 MB GPU memory, significantly less than traditional methods.
- Achieves 69-89 reconstructions per second, suitable for real-time applications.
- Utilizes 2D splat correspondences to derive 3D splats via multi-view geometry.
- Outperforms existing methods in geometric accuracy on ScanNet and DTU datasets.
- Addresses 'floater' artifacts common in few-view surface reconstruction.
Paper Resources
📖 Reader Mode
~2 min readAbstract: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.
| Comments: | 8 pages, 3 figures |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO) |
| Cite as: | arXiv:2607.14470 [cs.CV] |
| (or arXiv:2607.14470v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2607.14470 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Dasong Gao [view email]
[v1]
Thu, 16 Jul 2026 01:33:37 UTC (2,544 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing Modalities
ProMoE-FL introduces a Prototype-conditioned Mixture-of-Experts framework for multimodal federated learning, effectively addressing missing modalities. It outperforms existing methods on four chest X-ray datasets, demonstrating superior feature synthesis capabilities in both homogeneous and heterogeneous settings.
