SDDD: Stacked Ensemble Model for Driver Drowsiness Detection

The leading cause of road accidents is drowsiness. According to statistics, drowsy drivers cause accidents. According to a National Sleep Foundation survey, 20% of drivers are exhausted and sleep deprived behind the wheel.Hence, drowsiness detection systems must be installed in cars without requiring high GPU-powered hardware. The main components of a drowsiness system must have an accelerated inference time with an advanced straightforward design. As a result, this study proposes a stacked ensemble model for detecting driver drowsiness, in which the architecture identifies driver drowsiness by comparing and contrasting various lightweight models MobileNet-V2, SqueezeNet, and ShuffleNet, which have shorter inference time and low computing complexity. In the NTHU-DDD dataset, experiments were carried out to determine the efficacy of the proposed stacked ensemble model. This research work aims to maximize the accuracy while reducing the computational complexity and processing time.