Enhancing Driver Safety Using Deep Learning Models

This proposed work emphasizes the critical role of eye movement detection, particularly open and closed eye states, in identifying drowsiness, providing a clear focus on the project's main technological approach and its importance for road safety. The proposed model ensures constant checks on the driver’s safety and provides necessary observations on their behavioral activities. To ensure effective drowsiness detection, different methodologies such as Convolutional Neural Networks (CNN), customized Neural Networks and Recurrent Neural Networks (RNN) were employed for image processing. The CNN model achieved the best results. Two different data sources from Kaggle provided various training images for the model to operate in different scenarios. The proposed model aims to achieve accurate predictions across various case scenarios and to develop the most efficient model for real-time detection and alert systems, ensuring driver safety. Among the models our system gives the best result using CNN reporting an accuracy of 98.53%, and precision, recall and F-score of 98.71%, 98.04% and 98.37% respectively.