Driver Safety System: A Real-Time Sleep Detection and Lane Detection Model Using IoT and Deep Learning

Car driving security is an inevitable factor due to the growing number of vehicles and the increased occurrence of road accidents. The driver assistance systems play a vital role in preventing 90% of the traffic accidents. Driver drowsiness and the failure in maintaining lane discipline results in serious road accidents. In this study, we propose a model that can identify the initial signs of fatigue and notify the driver before a critical situation arises. This helps to prevent crashes caused by fatigue thereby advising the drivers either to take a break from driving or to be alert from sleeping. The system also identifies whether the driver is following lane discipline. The proposed model uses facial landmark detection by calculating eye aspect ratio (EAR) to identify drowsiness. Then a semantic image segmentation with UNET architecture is applied for lane discipline detection. The results are promising and demonstrate the feasibility of driver safety systems using google coral and deep learning.