DOI: -

Keywords: Traffic Accidents, Risky Driving, Driver Behavior, Monitoring System, Kecelakaan Lalu Lintas, Mengemudi Berisiko, Perilaku Pengemudi, Sistem Monitoring, YOLOv8

ABSTRACT

The significant growth in the number of vehicles is directly proportional to the increase in traffic accidents, especially in Indonesia. One of the leading causes of accidents is human factors, including risky driving behaviors such as drowsiness, drinking, using a phone, and not wearing a seatbelt. This study developed a Driver Focus Monitoring System based on YOLOv8 (SIMONKIDI) to detect such risky behaviors in real-time. The dataset used consists of 1,109 images, divided into 800 images for training data, 187 images for validation data, and 163 images for testing data. The training process was conducted over 100 epochs, detecting four classes of risky behavior. The Confusion Matrix values range from 0.800 to 0.960, with an accuracy score 0,865, Precision 0.916, Recall 0.920, and an F1 score 0.870. These findings indicate that the YOLOv8-based SIMONKIDI system demonstrates efficient and accurate performance in detecting risky driving behaviors. Therefore, this system is expected to contribute to reducing traffic accident rates and enhancing road safety.