Recognizing traffic signs is essential in guaranteeing traffic safety and reducing the risk of traffic accidents. This study proposes a deep learning-based approach that attempts various YOLO architecture versions to perform common traffic sign recognition in Vietnam. First, data collection is conducted by collecting images taken on roads in Can Tho City and Vinh Long province and then combining them with ZaloAI dataset of Vietnamese traffic signs in 2020. Next, a data augmentation process is deployed to form an enhancement dataset. Then, two versions of YOLO, the YOLOv5 model and the YOLOv8 model, are applied to the enhancement dataset for recognizing traffic signs and comparing the effectiveness of the two approaches. The experimental results show that although the YOLOv5 model takes more training time and has fewer parameters than the YOLOv8 model, the former can perform better in traffic sign recognition tasks.