Distracted driving-related traffic accidents persist to represent a serious threat to global traffic safety. The advanced object detection algorithms YOLOv8 and MobileNet SSD are used in this paper’s driver behaviour detection system to determine a variety of distracted driving behaviours, including texting, eating, and engaging with other passengers. The State Farm Distracted Driver Dataset, which included more than 20,000 labelled photos from ten different driver behaviour categories, was used to train and assess the system. The statistical accuracy of the dataset was guaranteed by preprocessing procedures consisting of class correction and annotation refining. In contrast with earlier models, the suggested system performs better, exhibiting exceptional accuracy, speed, and adaptability to real-world situations. Our methodology relies on evaluating these models’ performances in terms of detecting accuracy, classification ability, and computational efficiency. The analysis provides insightful information for applications in vehicle safety systems by highlighting each model’s pros and cons. The results of this comparison study aid in the continuous advancement of reliable driver monitoring systems leaning towards lowering traffic accidents and encourage safer driving habits. This study lays the groundwork for incorporating driver monitoring into advanced driver assistance systems (ADAS) and demonstrates the potential of YOLOv8 for safety-critical applications. For increased effectiveness and scalability, future research will investigate larger datasets, multi-modal sensor integration, and the deployment of on edge devices.