A Video Streaming Vehicle Detection Algorithm Based On Yolov4

- Safe driving depends on being able to sense information about the environment around the car, and one of the most important technologies to solve this problem is computer vision. The Series YOLO and SSD, Retina Net method are two one-stage target identification algorithms that excel in both accuracy and speed. Vehicle detection is still some distance from real-time, but the newest algorithm in the YOLO series, YOLOv4, has improved the rapidity and precision of vehicle target recognition compared to prior versions. This research suggests a faster vehicle target recognition technique based on YOLOv4 to address the issue with the slow detection speed. The YOLOv4 technique is theoretically introduced in this paper, followed by a suggestion for an algorithmic method to quicken the detection speed, and finally, actual road tests. According to testing results, the method presented in this research can increase detection speed without compromising accuracy, which can act as a foundation for decision-making for secure vehicle operation.