Smart Monitoring System for Driver Alertness

Driver fatigue remains a major contributor to road accidents worldwide, resulting in significant losses in both human lives and economic terms. Traditional detection methods based on physiological signals or behavioral cues often fall short of reliability under diverse realworld conditions. In this work, we present a non-intrusive driver drowsiness detection system that integrates deep learning with advanced computer vision techniques. A ResNet50V2-based model, fine-tuned using the Driver Drowsiness Detection (DDD) dataset, is employed to monitor key facial features such as head orientation, eyelid closure, and yawning frequency for accurate driver state classification. The proposed system achieves an overall detection accuracy of 97% by leveraging data augmentation, transfer learning, and regularization strategies. With its high efficiency and robustness, the system is well-suited for real-time deployment in next-generation in-vehicle safety applications.