Abstract

Driver drowsiness is a significant contributor to road accidents, often leading to impaired focus, delayed reaction times, and poor decision-making. To address this issue, this study introduces SnoozeNet, a lightweight and efficient real-time driver drowsiness detection system that combines Convolutional Neural Networks (CNNs), MediaPipe facial landmark tracking, and Temporal Convolutional Networks (TCNs). The model extracts spatial features from eye and mouth regions to detect blink rate, eye closure, and yawning, while MediaPipe provides head pose estimations to assess posture and nodding behavior. These features are fused and processed by a TCN to model behavioral transitions over time. The system was trained on diverse public datasets and evaluated against LSTM-based baselines, showing improved accuracy, training efficiency, and responsiveness. Results confirm that the lightweight CNN-MediaPipe-TCN pipeline effectively detects drowsiness-related facial cues across varied lighting conditions and facial structures, offering a robust and deployable solution for real-world driver-monitoring applications. Comprehensive validation showed that the pipeline achieved strong performance with an overall accuracy of 94.6%, F1-score of 0.930, and AUROC of 0.984, while delivering real-time classification in a browser-based application at approximately 15 FPS.

References

• Chen, L., Xin, G., Liu, Y., & Huang, J. 2021. Driver fatigue detection based on facial key points and LSTM. Security and Communication Networks, 2021(1), 5383573.
• Deshpande, A. V., Kalambarkar, S., Kale, A., Joshi, R., Kadam, S., Kadam, V., & Kale, N. 2024. A hybrid approach to drowsiness detection using MediaPipe and YOLOv5. International Journal of Innovative Research in Technology (IJIRT), 11(6), 3556–3561.
• Van Houdt, G., Mosavi, A., Napoles, G., & Shahriari-Rad, M. 2020. A review on the long short-term memory model. AI, 1(1), 110–122.
• Rundo, F., Spampinato, C., & Rundo, M. 2023. Car-driver drowsiness assessment through 1D temporal convolutional networks. arXiv preprint arXiv:2308.02415.
• Ismail, A. R., Sodoyer, D., & Elbahhar Boukour, F. 2023. Drowsiness detection in humans based on ECG analysis using temporal convolutional network. In Proceedings of the 2023 International Conference on Automation, Control, and Electrical Engineering (CACEE). IEEE.
• Weng, C. H., Lai, Y. H., & Lai, S. H. 2016. NTHU Drowsy Driver Detection Dataset.
• Abtahi, S., Omidyeganeh, M., Shirmohammadi, S., & Hariri, B. 2014. YawDD: A Yawning Detection Dataset. IEEE DataPort.
• Bindu, J., Jose, B. M., & Anjali, V. 2020. Driver Drowsiness Dataset (D3S). GitHub.
• Ortega, J. D., Kose, N., Cañas, P., Chao, M.-A., Unnervik, A., Nieto, M., Otaegui, O., Salgado, L. 2020. DMD: A Large-Scale Multi-modal Driver Monitoring Dataset for Attention and Alertness Analysis. In: A. Bartoli & A. Fusiello (eds), Computer Vision — ECCV 2020 Workshops. Springer International Publishing.
• Mammadli, S. 2024. Open and Closed Eyes Dataset.