Traffic Sign Recognition for Autonomous Vehicle with YOLOs: A Comparative Study

In the rapidly evolving field of autonomous driving, proficient detection and recognition of traffic signs are fundamental for ensuring the safety and efficiency of selfdriving vehicles. This research aims to identify the most suitable algorithms for this critical task by evaluating the performance of three prominent algorithms: YOLOv5, YOLOv8, and YOLONAS. The evaluation is conducted across three diverse datasets: Kaggle Self-Driving Cars Image Datasets, Kaggle Road Sign Image Datasets, and CeyRo Datasets. YOLO-NAS emerges as the standout performer across these datasets, showcasing commendable results. Additionally, the prediction results highlight YOLO-NAS's capability to accurately detect small and distant traffic signs. These findings collectively underscore YOLO-NAS as the optimal choice for traffic sign detection and recognition, particularly in situations characterized by challenging conditions.