Reproduce “PadLoc: Training-Free Floor-Plan Localization by Sparse Geometric Matching” 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
Marian Bookhahn, Frank Neumann
PadLoc replaces dense pose grids with sparse matches and a multi-hypothesis tracker. It narrows heading candidates from wall directions, matches short depth windows to a floor plan, and combines candidate poses over time. With shared UnLoc depth at Gibson T=50, frozen chunk success is 1.000 versus 0.923 forward and 0.949 refined. On Bonn, pooled SR@1 m favors PadLoc, though motion inputs differ.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.