A Neural Network Algorithm for Detecting Driver Distraction

This paper addresses the critical issue of distracted driving, a major contributor to road accidents globally. Utilizing advanced Convolutional Neural Networks (CNNs) enhanced with the Convolutional Block Attention Module (CBAM), which integrates deep learning with a sophisticated attention mechanism, this study effectively detects and classifies ten types of driver distractions. It highlights the application of a lightweight yet powerful neural network framework optimized for real-time vehicular environments. Extensive experiments have been conducted to demonstrate the effectiveness of the CBAM -enhanced model over traditional CNNs, particularly in its ability to focus on salient features for improved accuracy and generalization. Additionally, the model benefits from a pruning approach that significantly reduces computational demand without compromising detection accuracy. The results indicate substantial improvements in model efficiency and real-time applicability, paving the way for advancements in vehicular safety technologies.