Semantic SLAM with Hybrid 3D Gaussian Maps for Object-Level Indoor Understanding
International Journal of Information and Communication Technology Trends · 25.05.2026 · 6 (1) · 228-238
DOI: 10.66525/ijictt87 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Information and Communication Technology Trends, online erschienen am 25.05.2026. Verfasst von Alex Celeste und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Simultaneous Localization and Mapping constitutes the foundational cornerstone for autonomous systems, enabling mobile agents to navigate and interact with unknown environments. While traditional approaches focus predominantly on geometric reconstruction and camera pose estimation, contemporary applications require a deeper, semantic understanding of the surroundings. This paper introduces a novel framework for Semantic Simultaneous Localization and Mapping utilizing Hybrid 3D Gaussian Maps, specifically engineered for high-fidelity object-level indoor understanding. By integrating 3D Gaussian Splatting with real-time semantic segmentation networks, our proposed system jointly optimizes spatial geometry, photometric properties, and semantic labels within a unified representation. Unlike implicit neural representations that often suffer from slow rendering speeds and catastrophic forgetting, the explicit nature of 3D Gaussians allows for rapid rasterization, efficient local updates, and precise object-level manipulations. The hybrid map explicitly models discrete indoor objects as clusters of semantically homogenous Gaussians while capturing complex structural details and view-dependent lighting effects. Through comprehensive evaluations on standard indoor datasets, the proposed system demonstrates superior performance in both trajectory estimation accuracy and scene reconstruction quality compared to existing state-of-the-art baselines. Furthermore, the object-level understanding capabilities enable advanced downstream tasks, including scene editing and semantic navigation, bridging the gap between low-level metric mapping and high-level spatial reasoning.
Zitieren
APA
Celeste, A., Seo, T., Duane, C., & Kwak, T. (2026). Semantic SLAM with Hybrid 3D Gaussian Maps for Object-Level Indoor Understanding. International Journal of Information and Communication Technology Trends, 6(1), 228-238. https://doi.org/10.66525/ijictt87
Vancouver
Celeste A, Seo T, Duane C, Kwak T. Semantic SLAM with Hybrid 3D Gaussian Maps for Object-Level Indoor Understanding. International Journal of Information and Communication Technology Trends. 2026;6(1):228-238. doi:10.66525/ijictt87
BibTeX
@article{Celeste2026,
author = {Celeste, Alex and Seo, Taeyang and Duane, Carey and Kwak, Taemin},
title = {Semantic SLAM with Hybrid 3D Gaussian Maps for Object-Level Indoor Understanding},
journal = {International Journal of Information and Communication Technology Trends},
year = {2026},
volume = {6},
number = {1},
pages = {228--238},
doi = {10.66525/ijictt87}
}
EUVIMED