Tools
Tools
Gupta, Shreyansh (2023) Real Time Driver Drowsiness Detection Based on Deep Learning. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (6MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (1MB) | Preview |
Abstract
Drowsiness refers to a state of feeling sleepy, sluggish, or fatigued. It is one of the biggest challenge faced by drivers all across the globe, contributing to a notable percentage of road accidents and fatalities. This research endeavors to develop and assess machine learning models tailored for eye detection and drowsiness prediction, targeting real-time applications. Motivated by the pivotal role of precise eye state detection in safety systems and human-computer interaction, Convolutional Neural Network (CNN) models, particularly leveraging the InceptionV3 architecture through transfer learning, were crafted to discern open and closed eyes. The model has achieved promising training accuracies ranging from 91.67% to 95.04%, and also on validation and test datasets, registering accuracies between 83.74% to 92.60%. Notably, varying convergence patterns led to early stopping at different epochs, highlighting challenges in achieving high generalization for practical deployment. These findings underscore the necessity for further research to enhance model reliability, generalization, and applicability in real-time scenarios, aiming to contribute to the development of robust safety systems reliant on accurate drowsiness detection.
Actions (login required)
 |
View Item |