Reproduce “SelaVPR++: Towards Seamless Adaptation of Foundation Models for Efficient 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
Feng Lu, Tong Jin, Xiangyuan Lan, Lijun Zhang, Yunpeng Liu, Yaowei Wang, Chun Yuan
SelaVPR++ matches camera images to geo-tagged places despite changes in viewpoint and lighting. It trains a side adapter over frozen-backbone features, then uses compact binary global descriptors to retrieve candidates and floating-point global descriptors to rerank them without local-feature matching. On MSLS-challenge, it reports 93.7% R@5 versus SALAD-CM’s 91.2%; configurations differ.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.