Automotive Perception
Fusion of Radar, Lidar and Camera for automotive scene understanding
Radar sensors have a significant potential for several applications due to their low-cost and robustness to weather conditions. However, their use in 3D-detection work is challenging due to the sparsity of 3D information compared to Lidar.
In our most recent paper to be presented at WACV 2026, “IMKD: Intensity-Aware Multi-Level Knowledge Distillation for Camera-Radar Fusion” , we presented a Knowledge distillation method from lidar features that is able to maintain the specific advantages of radar features while enhancing them:
In our 2022 WACV paper “Fusion Point Pruning for Optimized 2D Object Detection with Radar-Camera Fusion”, we propose a method to optimize the neural network architecture for camera-radar fusion in 2D object detection:
In our ICPR 2020 publication “Ghost Target Detection in 3D Radar Data using Point Cloud based Deep Neural Network”, we train a neural network to classify radar point cloud 3D points as normal or multi-path (ghost) detections: