Detecting motorist drowsiness is crucial for ensuring road safety. Traditional methods for detecting tiredness rely on manual feature extraction and rule-based systems, which frequently lack precision and generalizability. This paper presents an improved fatigue detection system for automobiles that leverages the strength of deep learning methods. Deep convolutional neural network (CNN) architecture is utilised to automatically learn and extract features from driver-related data, such as facial expressions, eye movements, and physiological signals. The network is trained on a large dataset of labelled examples comprising multiple fatigue levels. The CNN optimises its internal parameters through backpropagation to accurately detect and classify fatigue levels in real-time. The proposed system accomplishes greater precision and reliability than conventional methods. Experimental results demonstrate that the approach based on deep learning outperforms existing systems in terms of detection precision, adaptability to a variety of parameters, and real-time performance. Implementing deep learning in drowsiness detection systems has the potential to significantly improve road safety by providing fatigued drivers with opportune warnings and alerts, thereby mitigating the risks associated with lethargic driving. Future research can concentrate on incorporating additional contextual information and multi-modal data sources to further enhance the performance and reliability of the system.

Deep Learning, Drowsiness Detection, Artificial Intelligence, Object Detection, Yolo