Contribution of the YOLO model to the UXO detection process

This paper explores the application of the YOLOv5 deep learning model for detecting unexploded ordnance (UXO) using UAV aerial imagery. UXO poses a global humanitarian and ecological threat, necessitating AI-driven solutions for real-time detection in complex environments. Integrating YOLOv5 with UAV systems, the study enhances detection accuracy, operational efficiency, and safety in high-risk areas. Using 12,600 annotated images, the model achieves a 97.77% mAP with optimized precision-recall balance. Advanced data augmentation and hyperparameter tuning further improve robustness. The proposed framework enables scalable, real-time UXO detection, significantly improving demining operations. Findings also support broader applications in disaster response, public safety, and military operations, demonstrating AI's critical role in global security and humanitarian efforts.