ABSTRACT
Detecting Incorrect driving behavior detection is becoming increasingly popular at the moment, since it is critical in guaranteeing the safety of the both passengers and drivers in vehicles, and it is a crucial step in realizing automatic driving at this point. Deep learning techniques have made significant progress recently, which can help with this challenging identification problem, this include the notable capacity for generalization of advanced deep learning models and the substantial amounts of video footage needed for the comprehensive training of these deep learning models that are data-driven. In order to achieve the goal of video-based aberrant driving behavior identification, fusions of deep learning approaches are highlighted, and three innovative deep learning based fusion models are presented, inspired by the recently suggested and popular densely connected convolutional network (DenseNet).These three Deep Learning models named as WGD,WGRD and WGRD. This study on incorrect driving behavior identification suggests that their superiority, is justified by thorough comparisons with other well-known deep learning models.
