ActiveLang: Active Open-Vocabulary 3D Mapping with Semantic-Uncertainty-Guided Exploration
ActiveLang: Active Open-Vocabulary 3D Mapping with Semantic-Uncertainty-Guided Exploration
ActiveLang is an autonomous system for active open-vocabulary 3D mapping that uses semantic-uncertainty-guided exploration to reconstruct scene geometry, appearance, and semantics with minimal memory overhead. It efficiently selects informative viewpoints for mapping, reducing computational cost and observations needed. Experiments on Replica and ScanNet++ show significant improvements in open-vocabulary segmentation over both online and offline baselines.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source:arXiv Robotics — research abstracts · arxiv.org