Algorithm for Distracted Driver Detection and Alert Using Deep Learning

Driver distraction is a significant source of road accidents and car crashes. A distracted driver poses a threat to not only himself and the ones in the car but also others in the road, namely nearby pedestrians, bicyclists, and other vehicles. Although distractions while driving may majorly seem to involve cell phone usage and texting, it also comprises of other events such as eating/drinking, communicating with co-passengers, adjusting hair/makeup, or fiddling around the radio or climate controls. Therefore, a system must be built that monitors the driver's activity and detects any distractions to alert him. Motivated by the advancements of Deep Learning, we propose a Convolutional Neural Network (CNN) model-based system to classify and identify distracted drivers to alert them, thus providing a potential solution to the problem. Our model classifies the driver activity into ten distinct classes, out of which nine are of the driver distracted by other events, and one is of "safe-driving". If the driver is found in any of the nine said classes, it means he is distracted, and our system will warn him so that any chance of an accident is eliminated.