Automated Drowsiness Detection for Driver Safety: A Deep Learning-based Approach

Road accidents are the most common cause of death and severe injuries today. Approximately 1.25 million people died worldwide from road accidents last year, and around 1.5 lakh people died in India [13]. The reason for many of these accidents is driver fatigue. In this paper, we have proposed a drowsiness detection system using different deep-learning algorithms. We used specific computer vision techniques to extract the facial features and later analyzed them to detect drowsiness precisely. We have compared the output of all two deep learning algorithms, i.e., CNN (Convolutional Neural Network) and MLP (Multilayer Perceptron), based on the metrics values, including accuracy, sensitivity, roc curve, F1 score, and specificity. The most efficient model in our research was MLP, with an accuracy of 86%. The suggested automated drowsiness detection system appears to be a promising and efficient solution to promote driving safety. The proposed model highlights the power of deep learning methods to tackle practical issues in detecting drowsiness in drivers.