A real time driving emotion detection based on yolov7 neural network

According to the Ministry of Transportation and Communications (MOTC) report, 3,085 fatalities occurred on Taiwan's roads in 2022, marking a more than 4% increase in accidents, injuries, and deaths compared to the previous year. The international concern regarding the risks posed by fatigued driving to both pedestrians and drivers has prompted a pressing need for research in real-time fatigue detection among car drivers. This research adopted the You Only Look Once v7 (YOLOv7) method to detect different emotions in car drivers. The study utilized data from facial expressions of students at the National Chin-Yi University of Technology (NCUT), comprising a total of 3,015 images categorized as sad, happy, neutral, angry, disgusted, fearful, surprised, and sleepy. The preliminary experimental results demonstrated that YOLOv7 maintained excellent generalization ability even with slight differences in various emotions. The mean average precision (mAP) achieved exceeds 87.7%.