Sensor Fusion for Traffic Management#
A multi-modal reference implementation to accurately monitor traffic conditions by fusing camera and sensor inputs. While cameras capture high-resolution visual data, radar and lidar sensors precisely measure speed and distance, even under challenging conditions such as fog, rain, or darkness.
Two complementary implementations are available:
Post-Fusion: the original camera + radar and camera + lidar pipelines that fuse independently processed sensor tracks after detection.
Intermediate-Fusion: a BEVFusion-based implementation that fuses camera and lidar features before the detection head for higher 3D object detection accuracy.
Choose the implementation that matches your sensor configuration and deployment requirements.