Driver drowsiness poses a serious threat to road safety, necessitating the development of an effective drowsiness detection system to prevent accidents. Numerous research endeavors have sought to address challenges of detecting abnormal driver behavior by analyzing the driver’s frontal face and vehicle dynamics using computer vision methods. With the emergence of deep learning architectures, extensive research is conducted to assess and identify driver drowsiness using neural networks. This research presents a new approach using YoloV5 frameworks to detect driver drowsiness. We have developed a customized pre-trained model within YoloV5 specifically for extracting faces to pinpoint Regions of Interest. To overcome shortcomings in existing methods, we have integrated drowsiness detection for binary image classification, trained and tested using the MRL Eye Dataset. This model achieves training and validation accuracies of 95.3 and 98.1%. The framework is evaluated on custom dataset of 39 subjects under different lighting conditions, achieving accuracy of 92%.