查看文章

Road accidents and related casualties can be prevented by identifying dangerous driving styles and proactively taking action to make improvements. While different driver assistance systems and mobile app-based solutions partially cover some of the aspects like warning and profiling systems, we present a standalone comprehensive camera and sensor based IoT system that runs state of the art models on a low cost edge device and performs driving behavior analysis by keeping track of vehicles’ acceleration, deceleration, turning, lane departure, and tailgating events as well as driver’s drowsiness and attentiveness. The system is deployed on a NVIDIA Jetson Nano, running in two parallel pipelines with IMU smoothened sensor data updating at 4 Hz while the rest of the models run using TensorRT Engine at approximately 2 fps when using the lightweight classification model for secondary task detection and 1.2 …