Event
TopoGS-SLAM proposes stationary-sampling Gaussian SLAM method
1 reports · 1 sources
Wei-Tse Cheng, Yen-Jen Chiou, and Yuan-Fu Yang propose TopoGS-SLAM, which uses coverage-triggered Gaussian initialization and bounded topology optimization to decouple refinement from capacity, turning mapping into a bounded-dimensionality problem. The authors claim that on standard indoor benchmarks, the method's localization and reconstruction accuracy is not inferior to densification-based pipelines, while significantly improving memory scalability and throughput. On this basis, the authors argue that in finite-area static indoor scenes with sufficient multi-view coverage, Gaussian SLAM does not require iterative topology mutation.
Report timeline
- TopoGS-SLAM: Reconstructing Gaussian SLAM densification strategy from a static sampling perspective
The authors propose TopoGS-SLAM, which decouples refinement from capacity using coverage-triggered Gaussian initialization and bounded-topology optimization, turning mapping into a bounded-dimensional problem. On standard indoor benchmarks, its localization and reconstruction accuracy is not inferior to densification-based pipelines, while memory scalability and throughput improve significantly. The authors therefore conclude that in bounded-area static indoor scenes with sufficient multi-view coverage, Gaussian SLAM does not require iterative topology mutation.
arXiv cs.RO ·