Vision-Based Driver Drowsiness Detection: From Deep Learning Models to Real-Time Mobile Deployment

Abstract

A significant percentage of traffic accidents worldwide result from sleepy drivers. Although detection methods have been established, their utility is often problematic. Physiological signals (EEG, ECG) and vision-based behavioral cues (eye closure, yawning) have been studied extensively, and deep learning models such as CNNs have shown excellent accuracy in controlled settings. Significant gaps still exist, however, especially regarding robustness against varying lighting and occlusions, on-road validation, and computationally efficient non-intrusive systems for real-time mobile deployment. This paper synthesizes and critiques current vision-based approaches and proposes a lightweight CNN architecture (MobileNetV2) optimized for on-device inference with TensorFlow Lite, offering a scalable solution for road safety on common Android devices.

Introduction

The text discusses the growing problem of driver fatigue and its major role in road accidents worldwide. Unlike intoxication, fatigue develops gradually through symptoms such as micro-sleep, drooping eyelids, and slower reactions, often unnoticed by drivers. Studies show that drowsiness contributes to 20–30% of highway fatalities, causing massive economic and social losses. With increasing long-haul transport, rideshare services, and extended work hours, the need for automated driver drowsiness detection (DDD) systems has become critical.

The paper reviews two main DDD approaches: physiological methods (using EEG, ECG, etc.) and camera-based behavioral methods. Physiological systems are accurate but intrusive and impractical for everyday drivers, while camera-based systems are non-intrusive and use indicators like Eye Aspect Ratio (EAR), Mouth Aspect Ratio (MAR), and head movement. Traditional threshold-based methods often fail under varied lighting, facial features, or obstructions. Deep learning models improve detection accuracy, but many are too computationally heavy for mobile devices.

To address this gap, the paper proposes a lightweight, mobile-first drowsiness detection system using MobileNetV2 and TensorFlow Lite for real-time smartphone deployment. The system combines facial landmarks, CNN-based image analysis, and features such as EAR and MAR. It also explores quantization techniques (Float16 and INT8) to reduce model size and improve speed while preserving accuracy. The approach emphasizes privacy through on-device processing and no permanent storage of biometric data.

The literature review identifies key research gaps, including poor mobile deployment support, lack of real-world testing, sensitivity to environmental conditions, high false alarms, reliance on intrusive sensors, and overdependence on single-modality systems. The proposed methodology addresses these issues through data augmentation, real-world datasets, hybrid CNN+LSTM modeling, and optimized mobile inference. Overall, the paper aims to create an efficient, practical, and privacy-conscious real-time driver drowsiness detection system suitable for everyday use.

Conclusion

This paper presented a comprehensive review of vision-based Driver Drowsiness Detection (DDD) systems [1]–[20] and identified persistent limitations in usability and real-world deployability. The most accurate systems in the literature tend to rely on intrusive sensors [6], [15] or computationally heavy deep learning models [9], making them impractical for everyday use [10]. These gaps are addressed by the proposed methodology, which emphasizes a lightweight, non-intrusive CNN architecture (MobileNetV2) [2], [3] optimized for on-device inference using TensorFlow Lite with INT8 quantization [17]. The experimental results demonstrate that the proposed system achieves 93.76% accuracy, sub-40ms inference latency at 26 FPS on a mid-range Android device, and maintains above 90% accuracy under challenging real-world conditions including night driving, sunglasses, and partial face occlusion [12]—representing a minimal accuracy trade-off against dramatic efficiency gains. By leveraging ubiquitous mobile technology and embedding strong privacy protections through a Privacy-by-Design protocol [12], this approach offers a scalable and cost-effective means of enhancing road safety globally [2], [10]. Future work will focus on real-world on-road validation across diverse driver populations and further improvements to robustness under extreme occlusion conditions [12], [18].

References

[1] H. A. Kassem, M. Chowdhury and J. H. Abawajy, “Drivers’ Fatigue Level Prediction Using Facial and Head Behavior Information,” IEEE Access, 2021. DOI: 10.1109/ACCESS.2021.3108561. [2] Y. Jebraeily and Y. Sharafi, “Driver Drowsiness Detection Based on Convolutional Neural Network Architecture Optimization Using Genetic Algorithm,” IEEE Access, 2024. DOI: 10.1109/ACCESS.2024.3381999. [3] M. Venkateswarlu and V. R. R. Ch, “DrowsyDetectNet: Driver Drowsiness Detection Using Lightweight CNN With Limited Training Data,” IEEE Access, 2024. DOI: 10.1109/ACCESS.2024.3440585. [4] A. Altameem, A. Kumar, R. C. Poonia, S. Kumar, and A. K. J. Saudagar, “Early Identification and Detection of Driver Drowsiness by Hybrid Machine Learning,” IEEE Access, 2021. DOI: 10.1109/ACCESS.2021.3131601. [5] U. M. Ravishankar and D. P. Hema, “Driver Drowsiness Detection System,” International Journal of Research Publication and Reviews, vol. 6, no. 8, pp. 5522–5530, Aug. 2025. [6] N. Keshan, P. V. Parimi, and I. Bichindaritz, “Machine learning for stress detection from ECG signals in automobile drivers,” in 2015 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2015. [7] P. K. Singh, A. Gupta, M. Upadhyay, A. Jain, M. Khari, and P. S. Lamba, “Multimodal Driver Drowsiness Detection from Video Frames,” Journal of Mobile Multimedia, vol. 19, no. 2, pp. 567–586, 2022. DOI: 10.13052/jmm1550-4646.19210. [8] H. A. Madni, A. Raza, R. Sehar, N. Thalji, and L. Abualigah, “Novel Transfer Learning Approach for Driver Drowsiness Detection Using Eye Movement Behavior,” IEEE Access, 2024. DOI: 10.1109/ACCESS.2024.3392640. [9] A. Sheikh and I. Z. Khan, “Enhancing Road Safety: Real-Time Detection of Driver Distraction through Convolutional Neural Networks,” arXiv preprint arXiv:2405.17788, 2024. [10] O. O. Ajayi, A. M. Kurien, K. Djouani, and L. Dieng, “A Multimodal Systematic Review of Drivers’ Fatigue Detection Methodologies, Datasets, and Models,” IEEE Access, 2025. DOI: 10.1109/ACCESS.2025.3606900. [11] Y. Guo, K. Yang, and Y. Wu, “A Multi-Modality Attention Network for Driver Fatigue Detection Based on Frontal EEG, EDA and PPG Signals,” IEEE Journal of Biomedical and Health Informatics, vol. 29, no. 6, pp. 4009–4020, June 2025. DOI: 10.1109/JBHI.2025.3527964. [12] A. Jarndal, H. Tawfik, A. I. Siam, I. Alsyof, and A. Cheaitou, “A Real-Time Vision Transformers-Based System for Enhanced Driver Drowsiness Detection and Vehicle Safety,” IEEE Access, 2025. DOI: 10.1109/ACCESS.2024.3522111. [13] A. Kolus, “A Systematic Review on Driver Drowsiness Detection Using Eye Activity Measures,” IEEE Access, 2024. DOI: 10.1109/ACCESS.2024.3424654. [14] S. Yaqoob, G. Morabito, S. Cafiso, G. Pappalardo, and A. Ullah, “AI-Driven Driver Behavior Assessment Through Vehicle and Health Monitoring for Safe Driving—A Survey,” IEEE Access, 2024. DOI: 10.1109/ACCESS.2024.3383775. [15] J. Alguindigue, A. Singh, A. Narayan, and S. Samuel, “Biosignals Monitoring for Driver Drowsiness Detection Using Deep Neural Networks,” IEEE Access, 2024. DOI: 10.1109/ACCESS.2024.3423723. [16] M. Golz, D. Sommer, B. Geißler, and A. Muttray, “Comparison of EEG-Based Measures of Driver Sleepiness,” Biomedical Engineering / Biomedizinische Technik, 2016. DOI: 10.1515/bmt2014-4079. [17] S. V. Dixith, S. Jadhav, Y. Kim, and N. Jayakumar, “Deep Learning-Based Drowsiness Detection System for Driver’s Safety,” IEEE Access, 2025. DOI: 10.1109/ACCESS.2025.3604728. [18] P. S. Lamba, R. Jain, S. Khan, S. M. Alanazi, A. Jain, A. T. Zamani, A. Panwar, and A. M. A. Abdallah, “Detection of Driver Drowsiness Using Adaptive Eye Characteristic Ratio for Enhanced Road Safety,” IEEE Access, 2025. DOI: 10.1109/ACCESS.2025.3566981. [19] M. Jain, S. Jain, and A. Dhawan, “Driver Drowsiness Detection and Alert System,” International Journal of Scientific Research in Computer Science, Engineering and Information Technology, vol. 11, no. 2, pp. 3504–3511, 2025. DOI: 10.32628/CSEIT25112815. [20] S. Yaqoob, G. Morabito, S. Cafiso, G. Pappalardo, and A. Ullah, “AI-Driven Driver Behavior Assessment Through Vehicle and Health Monitoring for Safe Driving—A Survey,” IEEE Access, 2024. DOI: 10.1109/ACCESS.2024.3383775.

Copyright

Copyright © 2026 Hitesh Jitendra Jadhav, Santosh Shriram Karvar, Atharv Arun Patil, Gaurav Anil Waje, Gaurav Vijay Barde, Bajirao Subhash Shirole. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.