Timorese Academic Journal of Science and Technology
ISSN : 2617 - 4944 (Print) ISSN : 2617 - 4952 (Online)
A YOLOv5 Model-Based Application for Traffic Sign Conditions Assessment
Author(s):
Clarinha de Jesus da Silva,
Vosco Pereira,
Frederico S. Cabral,
Carlito Pinto,
Hidekazu Fukai,
Full Paper Access: Donwload
Abstract: Clear and effective traffic signs are essential for the efficient operation of the road network, for the enforcement of traffic regulations, and for road safety. In Timor-Leste, due to poor trend, after an accomplishment of road construction, traffic sign management is not given proper assessment and management. This research introduces a novel application leveraging the YOLOv5s object detection framework for the assessment of traffic sign conditions which are classified into five categories namely very poor, poor, acceptable, good, and very good. The comparison experiment by using other versions of the YOLOv5 model is conducted against the proposed methodology. The experiment result shows model Yolov5s highlighting its superior performance in evaluating traffic signs conditions, encompassing factors such as visibility degradation, physical damage, and obstructions achieving 0.98 of Precision, 0.98 of Recall, and 0.989 of mAP.
Keywords: Traffic Sign, Object Detection, Deep Learning, YOLOv5.