Implementing an Alarm Based Driver Drowsiness Detection System for Traffic Safety Using Neural Network

Driver drowsiness is a very common issue that can lead to severe accidents on the road. According to studies, drowsy driving is a factor in about 20% of road accidents, with some estimates suggesting that the number could be even higher. The consequences of drowsy driving accidents can be devastating, including loss of life, serious injuries, and significant property damage. Drowsy driving is especially dangerous because it impairs a driver's reaction time, decision-making ability, and overall awareness of their surroundings. In this study, we have developed a Driver Drowsiness Detection (DDD) method based on the Convolutional Neural Network (CNN) model. Different facial features such as eye-opening and closing state and eye blinking have been extracted from both the left and right eyes and were fed to the CNN model for drowsiness classification. The developed CNN model was trained and tested on the publicly available "Drowsiness Detection Dataset" dataset of Kaggle which consisted of 2000 images of opened and closed eyes. To evaluate the model performance, different evaluation metrics such as accuracy, precision, recall value, and f1 score were calculated. The model performance was also compared against the state-of-the-art drowsiness detection methods. The proposed model obtained an accuracy of 99.67%, and a recall, precision, and f1-score of 100% which far exceeded the performance of the relevant drowsiness detection models. Overall, the performance of the proposed model reflects its efficacy in recognizing the drowsiness state of drivers in real-time conditions and assuring road safety.