Keeping Drivers Alert: A Solution for Monitoring Driver Attention in Assisted-Driving Vehicles

Robert Jutresa, Peter Peer, Ziga Emersic, Jihun Kim

Abstract: The advent of the new industrial revolution has led to a surge in the number of self-driving and assisted-driving vehicles on the roads. Although this development enhances the overall quality of life, it also poses new challenges and risks. One such difficulty is the need for driver attention in the event of system errors. To overcome this issue, a driver monitoring system is necessary to ensure that the driver remains focused and can take control if necessary, while also sending alerts in case of driver distraction. To address this challenge, we propose a method that employs a convolutional neural network based on the ResNet architecture. This system can recognize selected types of driver distractions in captured infrared images. We experimented with different depths of the ResNet architecture to determine the most optimal results. The best model achieved an accuracy rate of 94% in detecting various forms of driver distraction.

External IDs:dblp:conf/mipro/JutresaPE023