BUILDING A MOBILE APPLICATION FOR TRAFFIC SIGN DETECTION USING YOLO

Abstract

In this thesis, we propose a method for detecting Vietnamese traffic signs using YOLO, specifically the YOLOv10 model, which stands out for its lightweight structure and high accuracy. About the dataset, it contains 12,774 images spanning five classes that equivalent to five groups of Vietnamese traffic signs: Cam, Chi_dan, Hieu_lenh, Nguy_hiem, Phu. The dataset was divided into 86% for training and 7% each for validation and test. The YOLOv10 model achieved a mAP of 0.976, outperforming other models and demonstrating its suitability for real-time mobile applications. The resulting mobile application offers three key features: Images available in the phone’s library, photo capture and live detection, providing users with an efficient tool for identifying Vietnamese traffic signs.