Autonomous Object Identification from Dynamic Visual Streams Using YOLOv8

Abstract

This research presents a deep learning-based approach for real-time object detection and classification using a hybrid dataset composed of RoboFlow images and a custom Canadian Vehicle Dataset (CVD). The system employs YOLOv8, an advanced object detection algorithm known for high speed and accuracy. The CVD comprises 10,000 annotated images collected in varying weather conditions to address detection challenges in autonomous environments. The proposed method enhances model robustness by training on diverse scenarios including fog, rain, snow, and nighttime. Comparative analysis with baseline models demonstrates a significant improvement in detection performance. The resulting model proves to be an effective solution for intelligent transportation systems, enhancing safety and operational efficiency in both autonomous and human-driven vehicles.