"HRSCD-CNA"
收藏DataCite Commons2026-04-30 更新2026-05-03 收录
下载链接：
https://ieee-dataport.org/documents/hrscd-cna
下载链接
链接失效反馈官方服务：
资源简介：
"HRSCD\u2011CNA is a large\u2011scale benchmark dataset for semantic change detection of farmland conversion (cultivated land to non\u2011agricultural areas, primarily artificial surfaces). Built upon the public HRSCD dataset, it provides 28,226 pairs of dual\u2011temporal high\u2011resolution aerial RGB images (512\u00d7512 pixels, 0.5 m), together with pixel\u2011level binary farmland masks for each time step and change masks identifying farmland\u2011to\u2011non\u2011agricultural transitions. The dataset is divided into 279 geographic scenes and partitioned at scene level into training (167 scenes, ~16,900 patches), validation (42 scenes, ~4,200 patches), and test (70 scenes, ~7,100 patches) sets, following a 60%\/15%\/25% random split. All labels are 8\u2011bit single\u2011channel PNGs (255 = positive class, 0 = background), and the original HRSCD \u201cNo information\u201d pixels are mapped to 0. HRSCD\u2011CNA is released under a Creative Commons BY\u2011NC\u2011SA 4.0 license and is intended to serve as a challenging and reproducible baseline for semantic change detection and multitask learning in remote sensing."
提供机构：
IEEE DataPort
创建时间：
2026-04-30
