高分多源遥感洪水语义分割样本数据集：GF-FloodNet

该数据集GF-FloodNet，由精确配准的国产GF-3 SAR遥感图像和国产GF-2多光谱遥感图像组成，主要用于洪水区域提取任务，空间分辨率为1.5m、2.5m和4m，包含13388张样本图像及对应的像素级语义分割标签，样本及标签均以GEOTIFF格式存储。GF-FloodNet样本数据集具有高空间分辨率、样本有充分多样性、提供像素级二值标签、携带地理信息、多源遥感优势等特点，能够为深度学习等人工智能算法在遥感洪水监测领域的实际应用提供良好的训练样本，具有较强的实用价值。

This dataset, GF-FloodNet, is composed of precisely registered domestic GF-3 SAR remote sensing images and domestic GF-2 multispectral remote sensing images. It is primarily utilized for flood area extraction tasks. With spatial resolutions of 1.5 m, 2.5 m and 4 m, the dataset encompasses 13,388 sample images and their corresponding pixel-level semantic segmentation labels, both stored in GEOTIFF format. The GF-FloodNet sample dataset boasts several advantages, including high spatial resolution, adequate sample diversity, pixel-level binary labels, embedded geographic information, and the strengths of multi-source remote sensing data. It can provide excellent training samples for the practical deployment of artificial intelligence algorithms such as deep learning in the field of remote sensing-based flood monitoring, holding strong practical application value.

搜集汇总

背景与挑战

背景概述

该数据集是一个用于洪水区域提取的高分辨率多源遥感语义分割样本集，包含GF-3 SAR和GF-2多光谱遥感图像，分辨率达1.5-4米，提供13388个样本图像及像素级标签，存储为TIFF格式。其特点包括高空间分辨率、丰富的类内多样性和多源数据优势，适用于深度学习算法在洪水监测领域的训练和应用，具有强实用价值。数据集覆盖2021-2022年多个国家的区域，由中国科学院空天信息创新研究院开发。