Skip to content
SDGs Center UDINUS

Evidence record · UDINUS · SDGs

Performance Comparison of YOLOv8 and YOLOv11 Enhanced by Grey Wolf Optimizer for Motorcycle Helmet Detection in Traffic Surveillance

Rektorat UDINUS 2025 Published 2026-10-30

Evidence record · UDINUS · SDGs

Unit
Rektorat UDINUS
Period
2025
Goals
Record
bfa1c4fa-b7de-4320-bbf7-f379088312db
Published
2026-10-30

Universitas Dian Nuswantoro conducted a study comparing YOLOv8n and YOLOv11n object detection models for real-time motorcycle helmet detection in traffic surveillance. The activity involved UDINUS researchers Ubaidillah, Muhammad Naufal; Nugraha, Adhitya; Luthfiarta, Ardytha; Hidayat, Erwin Yudi; Buana, Chandra Lukita; and Marendratama, Radhitya. The study addressed helmet usage as a matter relevant to reducing fatalities and injuries in motorcycle accidents, while recognizing that enforcement remains a challenge in many regions.

UDINUS used the Grey Wolf Optimizer (GWO) to tune model hyperparameters with the objective of enhancing accuracy and efficiency. The optimization covered learning rate, batch size, image size, and momentum. The models were trained using a helmet detection dataset acquired from Roboflow. Their performance was evaluated with standard object detection metrics, namely Recall, Precision, mAP@0.5:0.95, and mAP@0.5. The study therefore examined both detection performance and inference time within a real-time detection context.

The experimental results indicated that GWO tuning improved model performance across most metrics. YOLOv8n combined with GWO achieved the highest reported mAP@0.5 of 0.900 and mAP@0.5:0.95 of 0.624. YOLOv11n combined with GWO reached a slightly higher mAP@0.5 of 0.901, while its mAP@0.5:0.95 was 0.614. For inference time, optimized YOLOv8n maintained a balance with 3.8 ms per image. The findings present metaheuristic optimization, particularly GWO, as an approach for improving deep learning-based object detection in safety-critical applications such as intelligent transportation systems.

This UDINUS activity connects helmet detection research with health and safety considerations and with traffic surveillance in intelligent transportation systems. The work provides evidence on model comparison, hyperparameter optimization, detection performance, and inference time while remaining focused on the methods and findings reported in the study.

Reference https://www.scopus.com/pages/publications/105031566669?origin=resultslist