Lightweight driver drowsiness detection using Haar Cascade and MobileNetV2

Abdul Shukor, Muhammad Syahir and Zainol Abidin, Husna (2026) Lightweight driver drowsiness detection using Haar Cascade and MobileNetV2. Journal of Electrical and Electronic Systems Research (JEESR), 29 (1): 19. pp. 153-159. ISSN e-ISSN : 3030-640X
Identification Number (DOI): 10.24191/jeesr.v29i1.019
Abstract

Driver drowsiness affects the driver’s attention while driving which could harm the driver himself and other road users due to lack of alertness and slower reaction times. The risk of having an accident due to the driver drowsiness could be reduced through practical non-intrusive monitoring systems. This paper demonstrates an effective, lightweight proof-of-concept framework for real-time driver monitoring under laboratory webcam conditions. Full on-road vehicular validation remains subject to future field trials under naturalistic driving environments. The detection framework combines Haar Cascade and MobileNetV2 Convolutional Neural Network (CNN) where Haar Cascade is utilized for face detection and extraction of the facial Region of Interest (ROI), while MobileNetV2 classifies the driver’s eye states and yawning behaviour. The system was implemented in Python using OpenCV and TensorFlow and integrated with a standard webcam. The performance of the proposed lightweight architecture, MobileNetV2 was compared with VGG16 using the same eye state and yawning datasets. The results show that MobileNetV2 outperforms VGG16 where the accuracy for eye state classification is 98.17% and 88.07% respectively. In terms of yawning, the accuracy for MobileNetV2 is 78.60%, while VGG16 is 66.51%. The findings demonstrate that the integration of Haar Cascade and MobileNetV2 provides an efficient, lightweight, and practical approach for real-time driver drowsiness detection, with promising applications in intelligent transportation systems and vehicle safety technologies.

Item Details
Edit Item
Edit Item
Downloads & Files
[thumbnail of 149168.pdf]
Text
149168.pdf
Download (1MB)
Indexing & Metrics