WorldFloods

该数据集专为洪水检测定制，包含了119个真实的洪水事件和424张洪水地图。数据集分为三个类别：水体、陆地和云层。规模上，共有424张洪水地图，其中407张用于训练，6张用于验证，11张用于测试。该数据集的任务是对洪水进行语义分割检测。This dataset is specifically tailored for flood detection, encompassing 119 real flood events and 424 flood maps. It is divided into three categories: water body, land, and cloud. In total, the dataset includes 424 flood maps, among which 407 are allocated for training, 6 for validation, and 11 for testing. The task of this dataset is semantic segmentation detection for floods.

搜集汇总

背景与挑战

背景概述

WorldFloods是一个全球性的洪水卫星图像数据集，包含119个历史洪水事件的422个洪水范围地图及其配对的Sentinel-2多光谱图像，旨在为机器学习驱动的洪水检测和分割模型提供训练与测试数据。该数据集整合了来自多个权威机构的标注数据，覆盖广泛的地理区域和时间范围，支持星载实时洪水监测算法的开发与应用。