Abnormal Behavior Detection for Driving Assistance in Real-Time Using Improved YOLO

Anxious, nervous, unstable, and aggressive driving are examples of abnormal driving behavior that endangers road safety. Any of these could result in dangerous circumstances while driving. Traffic accidents may be prevented and many lives could be saved if it were possible to robustly detect anomalous driving behavior. As a result, the system trained a reliable model using the improved YOLO network to increase the accuracy of abnormal behavior detection for driver in real time. The detection of telephoning, drinking, smoking, and drowsiness while driving is the basis for abnormal behavior detection. Five classes from a newly created dataset are used to train and evaluate this system. The main classes include Driver-Doze, Normal-Driving, Smoking, Drinking, and Telephoning.