LangGS-SLAM: Real-Time Language-Feature Gaussian Splatting SLAM
Seongbo Ha, Sibaek Lee, Kyungsu Kang, Joonyeol Choi, Seungjun Tak, Hyeonwoo Yu · 2026
Read LangGS-SLAM: Real-Time Language-Feature Gaussian Splatting SLAM by Seongbo Ha, Sibaek Lee, Kyungsu Kang, Joonyeol Choi, Seungjun Tak, Hyeonwoo Yu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-06991.txt.
Description
Seongbo Ha and five co-authors build a real-time SLAM system that maps a scene in both shape and language. The system rebuilds a dense field of language-aligned features from RGB-D input while tracking and mapping at low latency, bridging 3D seeing and language-based thinking. A top-K rendering pipeline renders high-size feature maps fast without warping their meaning.
To handle the mismatch between meaning and shape data and to curb memory use, a multi-criteria map manager prunes extra or inconsistent Gaussians while keeping scene wholeness. A hybrid tuner then refines shape and meaning jointly at different rates matched to each field's character. The system beats shape-only rivals on shape closeness, matches offline methods on meaning closeness, and runs at 15 frames per second.
Catalog record
| CREATOR | Seongbo Ha, Sibaek Lee, Kyungsu Kang, Joonyeol Choi, Seungjun Tak, Hyeonwoo Yu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-06991