Monitoring Maritime Traffic with Ship Detection via YOLOv4

In India a large part of goods transportation is carried out by sea, leading to an emerging requirement for remote maritime patrolling system, which also serves as an asset during wartime and peacetime for defence. In this research paper, we propose an automated maritime patrolling solution by making a Deep Learning Model Pipeline for Ship Detection from satellite images using existing State of the art Object Detection Algorithms like Faster-RCNN, SSD, YOLOv3, and YOLOv4. We compare results based on various Evaluation Metrics. Further we also release our own dataset which consists of around 300 satellite images of the top 13 busiest Sea-ports of India. After performing the validations, we found that the YOLO v4 displayed the best re-sults with a balanced mAP and FPS score to detect the ships in the satellite images.