ShipRSImageNet: A Large-Scale Fine-Grained Dataset for Ship Detection in High-Resolution Optical Remote Sensing Images
Zhengning Zhang,Lin Zhang,Yue Wang,Pengming Feng,Ran He
DOI: https://doi.org/10.1109/jstars.2021.3104230
IF: 4.715
2021-01-01
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Abstract:Ship detection in optical remote sensing images has potential applications in national maritime security, fishing, and defense. Many detectors, including computer vision and geoscience-based methods, have been proposed in the past decade. Recently, deep-learning-based algorithms have also achieved great success in the field of ship detection. However, most of the existing detectors face difficulties in complex environments, small ship detection, and fine-grained ship classification. One reason is that existing datasets have shortcomings in terms of the inadequate number of images, few ship categories, image diversity, and insufficient variations. This article publishes a public ship detection dataset, namely ShipRSImageNet, which contributes an accurately labeled dataset in different scenes with variant categories and image sources. The proposed ShipRSImageNet contains over 3435 images with 17 573 ship instances in 50 categories, elaborately annotated with both horizontal and orientated bounding boxes by experts. From our knowledge, up to now, the proposed ShipRSImageNet is the largest remote sensing dataset for ship detection. Moreover, several state-of-the-art detection algorithms are evaluated on our proposed ShipRSImageNet dataset to give a benchmark for deep-learning-based ship detection methods, which is valuable for assessing algorithm improvement.1
imaging science & photographic technology,remote sensing,engineering, electrical & electronic,geography, physical