Object Detection for Autonomous Vehicles in Urban Areas Using Deep Learning

This study details the development of a reliable automated system for detecting urban traffic using a custom dataset, designed specifically for this research. We developed a novel architecture based on the YOLOv8 framework, featuring a vehicle-mounted camera on a rotatable platform to continuously stream video data via a 4G network to a control center. The system processes this data using advanced deep learning techniques, achieving an impressive 85.5% accuracy rate in real-time identification of pedestrians, vehicles, and bicycles. This accuracy surpasses current state-of-the-art methods and demonstrates exceptional speed and reliability, crucial for the safety and efficiency of autonomous vehicles. By significantly reducing misclassification, our method not only enhances the safety of urban navigation but also sets a new benchmark in the application of deep learning technologies. This work bridges the technological divide between developed and developing nations, marking a substantial advancement in the field of autonomous vehicle technologies.