VEHICLE DETECTION AND COUNTING UNDER MIXED TRAFFIC CONDITIONS IN VIETNAM USING YOLOV4

Recent video-based vehicle detection and counting algorithm are one of the main components to determine the traffic state. With improvements in computer vision and machine learning approaches for object detection, advanced algorithms based on artificial neural networks, such as YOLO (You Only Look Once) with high precision, are commonly used to replace classical approaches. In this paper, we provide a method using YOLOv4 for the vehicle detection and counting of mixed traffic flow in the context of Vietnam's transport. We have tested the network for five vehicle types, including motorcycles, bicycles, cars, trucks, and buses. The test results show that our algorithm achieves the vehicles' detection accuracy and counting on the test set than others (e.g., Background Subtraction and Haar Cascade), indicating that the proposed method has higher detection performance.