Evaluation of YOLO-Based Models for Traffic Sign Detection in Urban Ecuadorian Environments

Traffic sign detection in urban environments presents unique challenges due to varying illumination, weather conditions, and regional sign diversity, particularly in Latin American countries such as Ecuador. In this study, a comparative analysis of YOLO-based models—YOLOv11, YOLOv12, and YOLOv26—was conducted to identify the optimal solutions for autonomous vehicle perception. The models were trained under identical conditions, using pretrained weights and fine-tuning for localized traffic sign detection, then evaluated using precision, recall, mean average precision (mAP), and inference time metrics. YOLOv12 achieves the highest mAP, indicating superior detection accuracy. The study provides empirical evidence for model selection based on deployment priorities, balancing accuracy, and computational efficiency.