Drowsiness Detection System using Machine Learning

Abstract

In today's busy world, people are unable to get complete rest and proper sleep and hence, they end up dozing off while driving which can be dangerous. Also, with the increase of automation in the automotive industry most of the actions previously performed by the driver are being automated. This provides comfort to the driver, but in certain situations the need of human judgement is necessary which cannot be completely replaced by artificial intelligence (AI). Thus, the driver needs to be always alert and ready.The Drowsiness Detection System is a device which will help reduce accidents by detecting if the driver is drowsy and alerting them by various means. The Real-Time Drowsiness Detection framework will be constructed using TensorFlow, Python3 and OpenCV. The TensorFlow model will be trained using datasets to detect signs of oncoming sleep like yawning, closing of eyes for a longer period-of-time, reduction in driver's movements, etc. Various sensors will be used to detect careless changing of lanes and oversteering.If this data indicates that the driver is not alert an alarm will be sounded within the vehicle and the seat of the driver will begin to vibrate. If the driver remains unresponsive, further actions like gradual slowing of the car can be taken.