A Novel Ship Detection Method in Remote Sensing Images via Effective and Efficient PP-YOLO

Ship detection in the remote sensing images could provide valuable information for military and civilian applications. The deep learning method has shown its power on object detection in images. Among the existing methods, You Only Look Once (YOLO) has been paid much attention in this domain. Especially, YOLO-v3 and YOLO-v4 are two typical branches that can realize relatively ideal detection results. However, the large image size and complex algorithm bring challenges to implement ship detection effectively and efficiently. This article studies a novel ship detection method, which utilizes PP-YOLO to achieve this objective. Two kinds of remote sensing images are utilized in the evaluation experiments, including visible light and synthetic aperture radar. Compared with YOLO-v3 and YOLO-v4, the proposed method has advantages of average accuracy and detection speed. To be specific, the average accuracy is increased by 10.8% and 32.7%, respectively. And the detection speed is improved by 26.6% and 75.6%, respectively.