YOLO-Based Ship Detection in Remote Sensing: A Practical Perspective

Abstract

Ship detection in remote sensing imagery is essential for maritime surveillance, port operations, and coastal safety. This study presents a practical evaluation of YOLO-based models tailored for real-time deployment in civilian maritime environments. The experiments utilize a curated version of the ShipRSImageNet_V1 dataset, from which all military vessels were removed to focus exclusively on civilian ships. This refinement ensures the model's applicability to non-military monitoring tasks and enhances its relevance for port and commercial vessel management. To support scalable deployment, the proposed framework integrates unmanned aerial vehicles (UAVs) for data acquisition with a collaborative edge-cloud inference strategy. Lightweight models operate onboard UAVs, while centralized GPU resources handle complex computations, enabling efficient and responsive maritime monitoring. This architecture aligns with emerging trends in low-altitude economy and supports rapid decision-making in dynamic coastal scenarios. Comparative results demonstrate that YOLOv8n, YOLOv11n, and YOLOv12n achieve a favorable balance between detection accuracy and computational efficiency. Future research will focus on improving model robustness, expanding domain-specific datasets, and incorporating multi-modal data to enhance performance in diverse operational conditions.