Region Based Convolutional Neural Network for Smart Driving Monitoring System

Road accident is a major problem in Bangladesh which claims many lives. The major reason behind these road accidents are due to the careless driving of the drivers. In this paper, an efficient driving monitoring system has been Proposed which subsumes Bangladeshi traffic sign, vehicle detection as well as driver alerting system. Twenty important road signs of Bangladesh have been used for such purpose. The traffic signs and vehicle images have been captured from roads during daytime and night time. A Convolutional Neural Network (CNN) is modeled by tuning its parameters to Figure out the most robust architecture to detect traffic signs and vehicles from images. Alongside the signs and vehicle detection, Aspect ratio based driver sleepiness detection is made to ensure road safety by avoiding such road accidents. This work is conducted using 4200 images which has gained an average recognition accuracy of 96.8%.