ATLANTIS

ATLANTIS数据集由南卡罗来纳大学创建，专注于水体图像的语义分割，包含5195张水体及相关物体的图像，具有高质量的像素级手动标注，涵盖56个类别。该数据集旨在解决水资源管理和洪水应急中的重要问题，通过提供大规模的训练和测试数据，支持计算机视觉领域的研究。ATLANTIS数据集的创建过程涉及使用Flickr API收集图像，并通过两阶段层次化程序筛选和标注，确保数据集的质量和多样性。该数据集适用于计算机视觉和水利工程领域的研究人员，特别是在水体识别和分割方面的应用。

The ATLANTIS dataset, created by the University of South Carolina, focuses on semantic segmentation of water body images. It contains 5,195 images of water bodies and related objects, with high-quality pixel-level manual annotations covering 56 categories. This dataset aims to address critical issues in water resource management and flood emergency response, and supports computer vision research by providing large-scale training and testing data. The development of the ATLANTIS dataset involved collecting images via the Flickr API, followed by a two-stage hierarchical procedure for filtering and annotation to ensure its quality and diversity. It is suitable for researchers in the fields of computer vision and hydraulic engineering, particularly for applications related to water body recognition and segmentation.