Smart Driver Monitoring System Using Yolov10

Abstract

Driver fatigue and inattention are major contributors to road accidents across the world. Continuous monitoring of driver behavior is essential to reduce accident risks and improve road safety. Traditional driver monitoring systems often rely on landmark-based facial analysis techniques, which are sensitive to lighting conditions, facial occlusions, and camera positioning. This paper presents a Smart Driver Monitoring System using YOLOv10, designed to detect unsafe driver behaviors such as drowsiness through real-time video analysis. The proposed system treats driver state detection as an object detection problem rather than relying on handcrafted facial measurements. YOLOv10 is employed for its high inference speed and accuracy, enabling continuous monitoring with minimal latency. A temporal decision mechanism is incorporated to reduce false alerts caused by natural facial movements. When unsafe behavior persists beyond a defined threshold, the system generates immediate audio and visual alerts. The proposed approach demonstrates improved robustness, scalability, and real-time performance, making it suitable for intelligent transportation systems and in-vehicle safety applications.