Driver Drowsiness Detection Using Vision Transformer

The issue of driver drowsiness poses a considerable safety risk, which can result in accidents and loss of life. To confront this issue, reliable and user-friendly systems that can detect early signs of drowsiness are needed. Vision-based drowsiness detection (DDD) systems have emerged as a promising approach to achieve this goal. DDD systems combine image processing methods with artificial intelligence to monitor the driver’s eyes and facial expressions and identify signs of drowsiness. In this work, we present a DDD approach using vision transformers (ViT). The attention mechanism in our proposed method focuses on the specific patches of the input image which contribute to the drowsiness phenomenon. Empirical results on the UTA-RLDD dataset show that our proposed model achieves a high level of test accuracy of 98.10 and an average prediction time of 17 ms per frame. Hence, our approach offers a solution for detecting early signs of drowsiness, providing drivers with enough time to take corrective action and reduce accident risks.