Reproduce “AdaFusion: Visual-LiDAR Fusion with Adaptive Weights for 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

Haowen Lai, Peng Yin, Sebastian Scherer

AdaFusion addresses place recognition by adapting the contributions of camera and LiDAR features for each observation. Separate visual and point-cloud branches use multi-scale attention to generate two weights, which scale and combine their features into a retrieval descriptor. On NCLT at descriptor dimension 128, it raises AR@1 from 92.61% to 94.06% versus direct concatenation, in tested settings.

Summary by alphaXiv.

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