FUSU-Fine_grained_Urban_Semantic_Understanding
收藏资源简介:
FUSU数据集覆盖了中国南北五个城市区域,总面积847平方公里,包含17种土地利用和土地覆盖类别,支持分割、变化检测和领域适应任务。数据集由两部分组成:1)高分辨率卫星RGB图像,空间分辨率为0.2-0.5米,时间分辨率为2年;2)Sentinel-2和Sentinel-1的月度重访图像,空间分辨率为10米,时间分辨率为1个月。数据集包含62,752个图像块,每个图像块包含25个Sentinel图像、2个Google Earth图像和2个对应的标注。标注信息基于中国土地利用分类标准(GB/T21010-2017)一级分类系统,包含17个类别。
The FUSU Dataset covers five urban regions across northern and southern China, with a total area of 847 square kilometers. It includes 17 land use and land cover (LULC) categories, and supports tasks such as semantic segmentation, change detection and domain adaptation. The dataset consists of two parts: 1) High-resolution satellite RGB images, with a spatial resolution of 0.2–0.5 meters and a temporal resolution of 2 years; 2) Monthly revisit images from Sentinel-1 and Sentinel-2, which have a spatial resolution of 10 meters and a temporal resolution of 1 month. The dataset contains 62,752 image patches, each of which includes 25 Sentinel images, 2 Google Earth images and 2 corresponding annotations. The annotations are based on the first-level classification system of China's Land Use Classification Standard (GB/T 21010-2017), covering 17 categories.




