2026-10-13T22:24:29Z https://www.scidb.cn/oai

10.57760/sciencedb.29478 2025-10-09T17:46:45Z

2025-10-09 A multi-temporal image sample set for the composition classification of Taihu Wetland based on Sentinel-2 doi:10.57760/sciencedb.29478 en The sample images were created by stacking the following layers in sequence: the median RGBN composite from June-September, median NDVI and NDWI from March-May, and median NDVI and NDWI from October-December. These layers were derived from 2024 Sentinel-2A/B imagery with cloud coverage less than 20%. After 4× upsampling, samples were collected as 32×32 pixel patches. The dataset encompasses ten ecological classes: Forest, Grassland, Bareland, Impervious Surface, Water, Cropland, Algal Bloom (AB), Emergent (EAV), Submerged (SAV), and Floating Aquatic Vegetation (FAV).The training and validation sets were divided in a 7:3 ratio. The training samples were augmented through 90°, 180°, and 270° rotations, as well as horizontal and vertical flipping. Sentinel-2; taihu wetland; Multi-temporal image sample set Yuxuan Hu PUBLIC https://creativecommons.org/licenses/by-sa/4.0/ dataset http://www.doi.org/10.1016/j.rsase.2026.102012 Science Data Bank