Real-time Vehicle Detection using YOLOv10-nano Algorithm for Monitoring Traffic Analysis

Abstract — Vehicle detection becomes crucial in the automation of traffic analysis, enabling assignment for instance vehicle categorization, speed estimation, and traffic flow monitoring. While vision-based systems are essential for recognizing vehicle types and sizes, improving detection performance and efficiency remains a significant challenge. This work aiming to evaluate how effective the YOLOv10n for real-time vehicle detection in resource-limited context. YOLOv10n, the most efficient version of the YOLO iteration to date, offers remarkable advancements in feature extraction and computational efficiency. Using the Vehicle-COCO dataset, which reflects real-world surveillance traffic conditions, the proposed framework attained score 0.637 at mAP@50 of and 0.442 at mAP50:95, demonstrating its capability to accurately detect various vehicle types. Furthermore, the model runs at 23.08 frames per second on a standard CPU, underscoring its suitability for edge deployment on low-cost devices. The findings confirm that YOLOv10n combines high efficiency with competitive accuracy, making it a viable solution for intelligent transportation systems. This paper also outlines potential directions for future research, including accuracy improvement, parameter tuning, and further optimization of computational resources. Key words — Deep Learning, Vehicle Detection, YOLOV10n, Efficient Model