Event
SG^2SLAM Monocular Outdoor Navigation SLAM Method Released
1 reports · 1 sources
Ziang Guo and other authors propose SG^2SLAM, a 3D Gaussian splatting SLAM method for monocular outdoor navigation. It replaces per-frame cross-dimensional correspondence with block-level metric priors, allowing tracking and mapping to share the same geometric source, and introduces a ray-conditioned 3DGS kernel that extends a viewpoint direction subspace for each primitive, jointly optimizing spatial geometry and pose. The authors state that on the KITTI and Waymo Open datasets, SG^2SLAM outperforms per-frame association baselines in monocular outdoor navigation tasks. The code will be made public after the paper is accepted.
Report timeline
- SG^2SLAM: Spatial Geometry-Constrained 3D Gaussian Splatting SLAM for Monocular Outdoor Navigation
The authors propose SG^2SLAM, a 3D Gaussian Splatting SLAM method for monocular outdoor navigation that replaces per-frame cross-dimensional correspondence with a chunk-level metric prior, so that tracking and mapping share the same geometric source. The method also introduces a ray-conditioned 3DGS kernel that extends each primitive with a viewing-direction subspace, and jointly optimizes spatial geometry and poses. On the KITTI and Waymo Open datasets, SG^2SLAM outperforms per-frame association baselines in the monocular outdoor navigation task. Code will be made publicly available upon acceptance.
arXiv cs.RO ·