An Improved YOLOv8 Model for Traffic Sign Detection and Classification

Intelligent transportation systems are aimed at improving the safety of roads and traffic management, making traffic sign detection a critical part of the system. In this work, we evaluated and compared two popular object detection models YOLOv8 and ResNet50 for the purpose of traffic sign recognition. The models were trained and assessed in various epochs to evaluate their performance. Our experiments show that the performance of YOLOv8 outperforms at epoch 47. The YOLOv8 model is characterized by an impressive accuracy of 98.32% as it presents precision, recall and F1-score values at 0.9345, 0.9428, and 0.9322, respectively. These metrics indicate that the model has a strong capability to detect and classify traffic signs based on high accuracy in field settings. On the contrary, the ResNet50 model performs worse with an accuracy of 84.32% and a loss level of 0.9932. It should be noted that this comparison is favourable as normally inter-rater reliability sits more while intra rate mostly varies upwards according to sources which have provided extensive information regarding diversity in the data. The lower accuracy and higher loss show that, in the traffic sign detection scenario, the YOLOv8 model is superior to ResNet50. Detailed evaluation of the performance models highlights that YOLOv8 possesses more effectiveness for accurate and reliable traffic sign detection, which could be implemented in intelligent transportation systems.