MS-SSD: multi-scale single shot detector for ship detection in remote sensing images

Object detection is a fundamental problem in computer vision. Although impressive results have been achieved on large/medium-sized objects, the detection performance of small objects remains a challenging task. Automatic ship detection on remote sensing images is an important module in maritime surveillance system, and it is challenging due to the high variance in appearance and scale. In this work, we thoroughly discuss the issues of SSD on multi-scale objects and propose a multi-scale single-shot detector (MS-SSD) to improve the detection effect of small ship targets and enhance the model’s robustness to scale variance. It enjoys two benefits by introducing (1) more high-level context and (2) more appropriate supervision. Extensive experiments on the Airbus Ship Detection Challenge dataset demonstrate the effectiveness of the proposed method in ship detection from complex backgrounds in remote sensing images. We also achieve better detection performance on the COCO dataset, outperforming state-of-the-art approaches, especially for small targets.