monster-monash/S2Agri-10pc-34

S2Agri-10pc-34是一个土地覆盖分类数据集，包含了覆盖法国一个12,100平方公里区域的单块哨兵2号卫星数据（T31TFM）。数据集使用10个光谱波段，以10米分辨率提供。数据集包含从2017年1月到10月的24个时间序列。该区域有广泛的作物类型和地形条件。数据集经过重新组织，适用于像素级处理，共有59,268,823个像素。提供了两种土地覆盖分类标签集合，分别有17个和34个类别。数据集的两个版本分别是S2Agri-17和S2Agri-34，分别使用17个和34个类别标签。此外，还提供了每个版本10%大小的数据集，即S2Agri-10pc-17和S2Agri-10pc-34。数据集根据原始地块进行划分，以确保地块内所有像素分配到相同的折，且类别标签在各个折之间均匀分布。

The S2Agri-10pc-34 is a land cover classification dataset consisting of Sentinel-2 satellite data covering a 12,100 km² area in France. The dataset uses 10 spectral bands and is provided at a 10m resolution. It contains time series of 24 observations from January to October 2017. The area includes a wide range of crop types and terrain conditions. The dataset has been reorganized for pixel-based processing, resulting in a dataset of 59,268,823 pixels. Two sets of land cover classification labels are provided, with 17 and 34 classes respectively. There are two versions of the dataset, S2Agri-17 and S2Agri-34, using 17 and 34 class labels respectively, as well as smaller 10% subsets of each, named S2Agri-10pc-17 and S2Agri-10pc-34. The dataset is split based on the original parcels to ensure that all pixels in a parcel are allocated to the same fold, with stratification by class labels for even representation across folds.