Reproduce “Spade-SLAM: Sparse Correspondence Anchors for Dense Self-Supervised Visual SLAM” 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
Julian Philipp Schmitt, Weirong Chen, Ganlin Zhang, Felix Wimbauer, Daniel Cremers
Spade-SLAM uses dense image-to-image training and matcher-derived sparse feature matches as geometric anchors, stabilizing camera and depth estimates without ground-truth pose or depth labels. It combines photometric and sparse reprojection supervision. On TartanAir’s hard split, it reports average aligned ATE RMSE of 0.50 m, versus 1.95 m photometric-only and 0.34 m supervised DroidCalib, placing it close to but below that baseline.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.