Reproduce “OASIS: Online Anchor-centric Scene Inference System for Open-Vocabulary Retrieval and Navigation in Dynamic Environments” 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
Kaiyao Liao, Qingfeng Li, Chen Chen, Xiaoze Wu, Qing Sun, Jianwei Niu
OASIS organizes robot memory around stable semantic anchors and selected representative images, retaining visual context for retrieval. During navigation, it stores unmatched detections as transient crops, contextual frames, and camera poses. When a target is missing, CLIP ranks crops and a vision-language model checks likely views. In cross-anchor tests, success was 77.4% versus 56.6% without the buffer, within 1 meter of the queried object.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.