Edge AI - Driven Vehicle Recognition and Tracking for Optimized Traffic Monitoring
The project contains the codes for achieving vehicle detection, vehicle type classification, vehicle passage counting, vehicle trajectory tracking, and vehicle retrograde recognition on applicable edge AI devices (e.g. Nvidia Jetson Nano). The folder of YOLO Train Results contains the results/outputs of YOLO models from YOLO v5 to YOLO v9. The "yolov8_det_trt_cam_track_count.py" contains the code to activate the TensorRT boost engine and YOLO v8 series models on Jetson Nano for classifying, tracking, and counting vehicles. The "OpenCV_V8.py" contains the code using the OpenCV library only, and it can theoretically track and count the vehicles without using the YOLO model on most embedded system devices (e.g. Raspberry Pi). The "OpenCV_V7_Cuda.py" and "OpenCV_V7_Part_Cuda.py" can be implemented on CUDA cores, including edge AI devices. The "requirments.txt" is same as YOLO v8 official GitHub Repository.