Learned Non-Maximum Suppression for 3D Object Detection
Researchers have developed new methods for post-processing in 3D object detection using LiDAR data. These techniques replace traditional non-maximum suppression (NMS) with learned filtering modules, D2D-Rescore and GossipNet3D, which leverage attention and message passing to refine detection results. The new approaches improve key metrics like mAP and NDS, especially for challenging classes, while maintaining low computational overhead. AI
IMPACT Improves reliability and accuracy in 3D object detection systems, potentially benefiting autonomous driving and robotics.