Drivers Drowsiness and Mental Health Detection using Deep Learning

Abstract

One of the most important body part is the face which holds a lot of information. Any person's mental state is revealed through their facial expression.The purpose of the study is to develop a system that can ensure safe driving with great accuracy. The main objectives of this study is to detect whether the driver is asleep or not, avoid road accidents, eliminate reckless driving, alert drivers about their mental and emotional situation. For this research, I have collected data from two different datasets. One is FER-2013 and another one is drowsiness dataset for open eyes and closed eyes. The images of the emotion dataset contains only the face which are enough cropped and the drowsiness dataset contains only the eyes. I have used angry, fear, happy, neutral, sad and yawning for emotion classifications. I have used 4 deep learning models in this research. The 4 four models are Xception, InceptionV3, ResNet50 and VGG19. These neural network models are used for feature extraction and classification tasks. The model that gives the higher accuracy than other models is Xception. In both tasks, The Xception outperformed the competition. For eye detection, it obtained 98.97% accuracy, and for face emotion detection, 99.26% accuracy. It showed excellent accuracy and metrics when it came to classifying emotions. The model performs quite well, with an average precision, recall, and F1-score of about 0.99. Overall, Xception performed exceptionally well across a number of emotion classifications and attained 99% accuracy on the eye dataset. The weighted and macro averages both confirmed the effectiveness of the system. This study suggests an improved pretrained model based approach for detecting driver’s inattention, which will ensure safe driving with great accuracy and improve the drivers driving efficiency. In future, I will work on adding night vision capabilities, the system will be able to identify and recognize objects better and adapt to different driving circumstances more easily. Furthermore, it may be proposed that driving behaviors like speeding, safe driving, and braking suggest a more advanced system. The combination of these factors can result in an improved and more advanced system that can identify and mark drowsy drivers, improve drivers' concentration, and reduce the number of traffic accidents.