Reproduce “ActiveLang: Active Open-Vocabulary 3D Mapping with Semantic-Uncertainty-Guided Exploration” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Liyan Chen, Hairong Yin, Huangying Zhan, Yi Xu, Raymond A. Yeh, Philippos Mordohai
ActiveLang maps rooms with language-queryable semantic labels while autonomously selecting views. It scores viewpoints for geometric coverage, semantic uncertainty and travel cost, while adapting a language representation as observations arrive. On ScanNet++, it achieves 20.13% 2D novel-view mIoU averaged over eight scenes, versus 9.22% for OnlineLangSplat. The comparison uses passive baselines and ground-truth poses, so does not isolate autonomous exploration.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.