Ship detection is crucial for defense and security, port management, environmental monitoring, insurance and risk assessment, as well as maritime search and rescue operations. Although the Automated Identification System (AIS) is widely used for ship tracking, it has limitations—such as incomplete data when AIS devices are turned off or malfunctioning—and not all vessels are equipped with AIS transponders. Satellite imagery-based detection addresses these gaps.

This pretrained model identifies and localizes ships in high-resolution optical satellite imagery, effectively handling both dense and sparse ship distributions. The detected polygons are aligned with the ships’ rotation angles, providing more accurate orientation information and improving the spatial representation of each vessel. Built on the Mask R-CNN architecture and implemented with the ArcGIS API for Python, the model has been trained on a proprietary ship detection dataset that includes a diverse range of ship types and sizes across the United States.