Reproduce “KISS-Matcher: Fast and Robust Point Cloud Registration Revisited” 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
Hyungtae Lim, Daebeom Kim, Gunhee Shin, Jingnan Shi, Ignacio Vizzo, Hyun Myung, Jaesik Park, Luca Carlone
KISS-Matcher introduces an open-source C++ library for global point cloud registration that achieves competitive accuracy with existing methods while operating at substantially higher speeds and demonstrating superior scalability for large-scale and map-level environments. The learning-free approach provides robust performance and generalization across diverse datasets and sensor configurations.
Key insights
- Holistic optimization of classical algorithms and data structures can lead to a global registration system that outperforms modern approaches in speed and scalability without sacrificing accuracy.
- Efficiently addressing computational bottlenecks in individual components, such as FPFH computation and graph-theoretic pruning, is critical for achieving real-time performance in large-scale scenarios.
- Learning-free, hand-crafted feature methods, when carefully designed, offer superior generalization capabilities across diverse datasets and sensor types compared to data-driven approaches.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.