Enhanced Radar Object Detection and Classification with Deep Learning on Range-Doppler (RD) Maps

Radar-based object detection is gaining attention for its robustness in various weather conditions, outperforming traditional optical sensors like cameras and LiDAR. It is valuable in robotics, autonomous vehicles, and surveillance for reliable object detection, classification, and tracking. Deep learning models further enhance the detection accuracy in this field. In this research, we propose an object detection method using only Range-Doppler (RD) maps, employing a slightly modified Faster R-CNN and a deep learning model, YOLO. Using the open-source RADDet dataset, our models achieve [email protected] scores of 0.46 and 0.37 for YOLO and Faster R-CNN, respectively.