| Summary: | 1.35 million annual deaths or disabilities were caused by traffic accidents as registered by the global road safety report in 2018. The most prevalent traffic safety problem is distracted driving, according to 87.5% of 4,178 respondents to a survey done by the AAA Foundation for Traffic Safety in 2021. Distracted driving, particularly phone use and drowsiness, is recognized as a significant safety issue and the public is highly aware of this matter, emphasizing the need for a solution that can mitigate this problem. This paper discusses the technical challenges involve the extensive use of mobile devices, incorporation of in-car technologies, and a societal tendency towards multitasking, collectively contributing to a persistent challenge where drivers frequently divert their attention from the road, amplifying the risk of accidents and fatalities. Here, several videos of distracted drivers were initially collected as training, validation and testing data for the proposed Deep Learning (DL) algorithm. After acquiring more than 90% accuracy in the testing phase, the inference model is deployed to an embedded system, incorporating Jetson Nano which is connected to a Digital Video Recorder (DVR). Any video captured from the DVR is fed into the model to detect drivers’ behaviour. Through 5G connectivity, crucial data of the drivers were transmitted to the Amazon Web Service (AWS) cloud infrastructure. A user-friendly web app dashboard plays a central role in reducing distracted driving by providing real-time feedback and alerts to drivers in proactive accident prevention, while benefiting law enforcement or policymakers for further understanding on why the drivers were distracted. The integration of DL to classify driver behaviour, along with the monitoring web app, holds the potential to mitigate distracted driving and in advancing road safety in future mobility. © School of Engineering, Taylor’s University.
|
|---|