HierToulouse

Dataset Description: The dataset contains high-resolution remote sensing images of Toulouse, France, along with corresponding land cover labels organized at two hierarchical levels. Each image is annotated with land cover labels, classified into broad categories at the first hierarchical level. The second level provides more specific subcategories within these broad classes, enabling a finer granularity of land cover mapping. At the first level, the land cover is categorized into four primary classes: Artificial Surface Natural Surface Woody Vegetation Non-Woody Vegetation At the second, more detailed level, the land cover is further subdivided into ten secondary classes: Construction Area Non-Construction Area Mineral Material Area Water Surface Broad-Leaved Forest Coniferous Forest Mixed Forest Shrubbery Other Woodlands Herbaceous Plants Citation: If you use this dataset in your research, please cite the following paper: Liu, L., Tong, Z., Cai, Z., Wu, H., Zhang, R., Le Bris, A., & Olteanu-Raimond, A. M. (2024). HierU-Net: A hierarchical semantic segmentation method for land cover mapping. IEEE Transactions on Geoscience and Remote Sensing, 62, 1-14.