Engineering and Applied Sciences Journal

Cubo: Edge AI Device System for Real-Time Driver Distraction Detection Using Geometric Rules and Quantum-Optimized CNNs

Abstract

Dhruva Valluru

Driver distraction is a significant factor in global road deaths, causing over 1,000 fatalities every day. Existing detection technologies are expensive and mostly limited to high-end vehicles. This study proposes a widely deployable, nonintrusive system combining facial landmark tracking, object detection, and deep learning to monitor driver distraction. Using a multilayered model based on Media Pipe, YOLOv5, and Mobile Net CNN, trained on over 14,000 frames and optimized with QSGD, the system achieved 88.1% accuracy with strong temporal consistency. This embedded solution offers real-time, scalable distraction detection to improve road safety. 

PDF