Detecting Driver Sleepiness using Convolutional Neural Networks

Abstract

The development in computer vision has aided drivers in the form of automatic self-driving cars etc. The accidents are caused by driver's exhaustion and drowsiness about 20%. Its carriages a dangerous issue for which numerous methods were proposed. However, they are not appropriate for real-time implementation. The major encounters confronted by these approaches are forcefulness to handle dissimilarity in human face and lightning conditions. Our intention is to implement a smart operating system that can lower the rate of road accidents considerably. This method enables us to find driver's face features like eye closure percentage, eye-mouth aspect ratios, blink rate, yawning, head movement, etc. In this classification, the driver is uninterruptedly observed by using a webcam. The car driver’s facial features along with the eye movements are observed using a cascade classifier. Eye images are pull out and fed to Custom designed Convolutional Neural Network for categorizing whether both left and right eye are closed. Based on the sorting, the eye closure score is considered. Upon finding that the driver is being detected drowsy that a high alarm will be raised.