Development of an Artificial Intelligence-Driven Marksmanship Assessment System for Defence Training Applications

Abstract

The development of effective marksmanship skills is essential for defence readiness and operational performance. Traditional shooting evaluation methods rely heavily on manual inspection, which is labor-intensive, timeconsuming, and vulnerable to human error. This study presents the development of an artificial intelligence-driven marksmanship assessment system designed to automate bullet impact detection, scoring, and performance analysis for defence training applications. The proposed system integrates deep learning-based object detection and computer vision techniques to identify bullet holes and compute scoring metrics based on shot placement, radial distance from the target center, and dispersion patterns. The YOLOv5 framework is employed for robust and real-time bullet hole detection, enabling automated evaluation of precision, accuracy, and shot consistency. Experimental validation demonstrates that the system achieves a detection accuracy of 95% and reduces assessment time from several minutes to approximately 5–8 seconds per target. The results confirm the reliability, efficiency, and scalability of the system under realistic training conditions. The proposed solution offers a cost-effective and data-driven approach for modernizing defence training infrastructure, providing objective performance feedback and supporting informed decision-making in shooter development.