Reproduce “EmbodiedPlace: Learning Mixture-of-Features with Embodied Constraints for Visual Place Recognition” 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

Bingxi Liu, Hao Chen, Shiyi Guo, Yihong Wu, Jinqiang Cui, Hong Zhang

EmbodiedPlace uses associations among stored images to refine global visual descriptors for visual place recognition. After initial retrieval, embodied constraints select neighbors whose descriptors are combined through learned weights to rerank candidates, avoiding a dedicated local-feature branch. On Pitts-30k-test, it raises DINOv2-GeM’s Recall@1 from 91.8% to 92.7%; gains are reported on four tested datasets, not universally.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.