Deep Learning-Based Road Sign Detection & Recognition System for Autonomous Vehicles

The rapid evolution of AV technology demands reliable systems for the detection and recognition of road signs. Especially in unique traffic environments like Sri Lanka. This study addresses a gap in localized research by developing a deep learning-based road sign recognition system. The system is designed specifically for Sri Lankan traffic scenarios. A robust dataset of Sri Lankan Road signs was created through manual and online data collection. The data was annotated using Roboflow. The model, based on YOLOv8m, was trained and integrated into a moving vehicle for real-time detection and recognition. The experimental setup integrates the trained model into a moving vehicle for real-time detection and recognition. The results show very good accuracy and reliability. Thus, proving the potential of this system to improve road traffic safety and efficiency. This study contributes to autonomous navigation research in Sri Lanka. It also provides a foundation for future work on dataset expansion and hybrid models to enhance efficiency and robustness.