2019-2021年三都澳近海岸海面目标可见光遥感图像数据集

本数据集通过谷歌地球下载原始影像后，经过裁剪、筛选、标注、格式转换等一系列操作流程制作而成，包含1761张遥感图像和对应的三类深度学习标签，可为近海岸海面目标（包括船舶、鱼排网箱养殖区、筏式养殖区）图像的语义分割、水平目标检测、旋转目标检测等深度学习研究领域提供基础数据。随着本数据集的不断更新发展，今后将建设成为国内标准的近海目标图像识别研究数据资源，为相关研究人员提供统一的训练与测试数据。通过深度学习方法建立相关的计算机视觉识别模型，推动近海目标图像识别研究的发展。

This dataset is developed by downloading raw imagery via Google Earth, followed by a series of processing steps including cropping, filtering, annotation, and format conversion. It contains 1761 remote sensing images and three corresponding types of deep learning labels, providing foundational data for deep learning research fields such as semantic segmentation, horizontal object detection, and rotated object detection of nearshore sea surface targets (including ships, fish cage culture zones, and raft culture areas). With continuous updates and improvements, this dataset will be developed into a national-standard data resource for offshore target image recognition research, providing unified training and test data for relevant researchers. The establishment of corresponding computer vision recognition models via deep learning methods will promote the advancement of offshore target image recognition research.