R3PM-Net: Real-time, Robust, Real-world Point Matching Network

R3PM-Net, a lightweight point cloud registration network, achieves 7x faster inference than state-of-the-art (RegTR) while maintaining competitive accuracy. The paper introduces two new real-world datasets (Sioux-Cranfield and Sioux-Scans) for evaluating registration of imperfect photogrammetric and event-camera scans against CAD models. Prioritizes generalizability to industrial scenarios over clean synthetic benchmarks.