Related Experiment Video
Updated: Jan 11, 2026
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
5.2K
Real-time driver activity detection using advanced deep learning models
Md Al Emran1, Md Ariful Islam1, Md Obaydullahn Khan1
1Pabna University of Science and Technology, Pabna, Bangladesh.
Cognitive Neurodynamics
|November 17, 2025
Summary
A new AI model accurately detects driver behaviors like distraction and drowsiness to enhance road safety. This advanced system shows high accuracy in real-time applications, aiming to reduce traffic accidents globally.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Machine Learning
Background:
- Traffic accidents are a major global safety concern, often caused by driver inattention, sleepiness, and distraction.
- Computer vision and AI offer promising solutions for real-time driver monitoring systems to mitigate these risks.
Purpose of the Study:
- To develop and evaluate a novel deep learning architecture for multi-class driver activity categorization.
- To compare the proposed model's performance against established deep learning models for driver behavior analysis.
Main Methods:
- Assessed four deep learning models: MobileNetV2, DenseNet201, NASNetMobile, and VGG19.
- Proposed a unique Hybrid CNN-Transformer architecture enhanced with Efficient Channel Attention (ECA).
- The framework categorizes seven key driving behaviors: Closed Eye, Open Eye, Dangerous Driving, Distracted Driving, Drinking, Yawning, and Safe Driving.
Main Results:
- Baseline models DenseNet201 and MobileNetV2 achieved high validation accuracies (99.40% and 99.31%, respectively).
- The proposed Hybrid CNN-Transformer with ECA achieved a near-perfect validation accuracy of 99.72%.
- The model demonstrated flawless generalization, achieving 100% accuracy on an independent test set, with minimal misclassifications.
Conclusions:
- The Hybrid CNN-Transformer with ECA effectively merges CNN local feature extraction, attention-based refinement, and Transformer global context modeling for robust and efficient driver monitoring.
- The developed system shows significant potential for real-time intelligent transportation applications, contributing to reduced traffic accidents and improved road safety.
