Paper Title
A NOVEL APPROACH TO DETECT DRIVER DROWSINESS USING DEEP CNN
Abstract
Abstract - Road accidents have only been increasing with time, and a significant contributor is negligent driving due to tiredness. A driver may unintentionally fall asleep, resulting in tragic accidents and a loss of lives. To bring down the count of such accidents, Driver Drowsiness Detection systems have come into existence to alert drivers when they doze off. Various Machine Learning and Deep Learning models are used in these systems. After drawing comparative analysis between various Deep Learning models used in drowsiness detection systems such as CNN, Deep CNN, and LSTM, we have focused our research on the CNN and Deep CNN models. The trained model considers the eye state and the frequency of blinking. It also considers the frequency of yawning by the driver in a given period. Depending on the eye state and yawn state, the model classifies the driver as drowsy or not. Though CNN can detect drowsiness in drivers, the Deep CNN model overcomes its limitations by providing a more efficient solution with an accuracy of 99.53%.Using the proposed model, we can enable a hardware system in automobiles to alert drivers if they feel drowsy.
Keywords - CNN, Deep CNN, LSTM, Face Detection, Drowsiness