Ship Detection from Optical Remote Sensing Images

Ship detection plays a critical role in many areas such as maritime security, smuggling prevention, environmental monitoring, and maritime traffic management. In particular, ship detection using satellite imagery enables the effective monitoring of large maritime areas and makes significant contributions to both civilian and military operations. In this study, YOLOv5 and YOLOv8 models are used to detect different types of ships in the coastline and offshore using a dataset of optical satellite imagery. The results show that both YOLOv5 and YOLOv8 models achieve sufficient accuracy for ship detection. It is also observed that YOLOv5 is more affected by the selected hyperparameters compared to YOLOv8, which improves its accuracy. As a result, while the ’Yolov5l’ model provides the best performance for YOLOv5, the ’Yolov8m’ model provides the best performance in YOLOv8.