Development and Application of Ship Detection and Classification Datasets: A review

Ship detection and classification pose significant challenges in remote sensing. The potent feature extraction capabilities of deep learning algorithms render them pivotal for these tasks. However, the effectiveness of deep learning algorithms hinges on extensive datasets. Consequently, researchers have developed numerous datasets for ship detection and classification. Recent research in this field has predominantly concentrated on crafting increasingly intricate deep neural network architectures and refining training strategies, with limited attention given to ship detection and classification datasets. In this paper, we analyze the evolution, applications, and future directions of ship detection and classification datasets. First, we review the historical development of ship detection and classification datasets and introduce currently available datasets. Second, we summarize and analyze the issues faced by ship datasets. Then, we discuss the current application status of ship datasets to explore solutions to these problems. Finally, we look forward to the future development direction of ship detection and classification datasets and provide some suggestions for people to construct new datasets. To enhance the practical application and dissemination of ship detection and classification datasets, we propose the utilization of a broader range of remote sensing data sources to achieve robust generalization performance. Moreover, there is an urgent need for the construction of large benchmark datasets for ship detection and classification. We anticipate that this paper contributes to an understanding of the distinctions between ship datasets and other target datasets in the remote sensing community and guides the future development of ship datasets.