Front and Rear Classification of Vehicles in Indian Context using Deep Neural Networks

Since the inception of deep learning, we have seen various systems that detect and classify objects into categories, an application of which includes the classification of vehicles into cars, trucks, scooters, etc. To detect whether a vehicle under consideration is moving in the correct lane, we find various hindrances such as faded Markings, shadows, different lighting conditions, rural roads, etc. Hence, one of the important intermediary tasks is to detect the view of the other vehicles that come in the vicinity of the vehicle. This paper works toward the classification of four-wheeled vehicles in visibility as front or rear for Indian roads. We’ve trained and compared three Convolutional Neural Networks for our binary-class classification problem that classifies the view of the vehicle as front or rear based on the data acquired from the dashcam in our vehicle.