Driver drowsiness constitutes a critical safety concern on the roads, demanding effective preventive measures. This study addresses the challenge of driver drowsiness detection with a sharp focus on enhancing safety. This innovative approach employs Convolutional Neural Networks (CNNs) and transfer learning techniques, chosen for their exceptional performance in image analysis and feature extraction. CNNs, with their ability to capture intricate spatial patterns, offer notable advantages over other ML models, making them the optimal choice for robust eye feature recognition. By processing real-time image and video data from the driver’s environment, the methodology incorporates a robust model trained on diverse datasets, ensuring adaptability to various driving conditions. Through rigorous experimentation, the system achieves remarkable accuracy, surpassing conventional methods, and holds the potential to substantially reduce drowsiness-related accidents, ultimately advancing road safety.