WHU Building Dataset

该数据集，即武汉理工大学建筑物数据集，专门采用“卫星数据集I（全球城市）”进行实验，包含了204张建筑物卫星图像，每张图像的分辨率为512x512像素。该数据集展现了在不同大气条件、多光谱融合算法以及季节变化下的多样性，这对于评估建筑物分割算法的鲁棒性至关重要。此外，该数据集重点关注在变化的大气条件下进行建筑物分割。规模上，它包含了204张512x512大小的图像。该数据集的任务是对遥感图像中的建筑物进行语义分割。

This dataset, namely the Wuhan University of Technology Building Dataset, specifically adopts "Satellite Dataset I (Global Cities)" for experiments. It contains 204 building satellite images, each with a resolution of 512×512 pixels. This dataset exhibits diversity across varying atmospheric conditions, multispectral fusion algorithms, and seasonal changes, which is crucial for evaluating the robustness of building segmentation algorithms. In addition, this dataset focuses on building segmentation under changing atmospheric conditions. In terms of scale, it consists of 204 images sized 512×512 pixels. The task of this dataset is semantic segmentation of buildings in remote sensing images.

搜集汇总

背景与挑战

背景概述

WHU Building Dataset是一个包含高分辨率航空影像和多源卫星影像的建筑物数据集，涵盖全球多个城市和东亚地区，适用于建筑物提取和变化检测研究。数据集提供了丰富的标注数据和裁剪样本，支持不同格式和分辨率的需求，适合用于深度学习算法的训练和测试。