Self-attention for raw optical Satellite Time Series Classification

Dataset for crop type mapping in Bavaria. Contains raw (unpreprocessed) Sentinel 2 satellite time series acquired over the year 2018 associated with crop type labels of the field parcels. This dataset was used to compare several deep learning models on preprocessed satellite data to industry standards with raw satellite time series to assess the robustness of the tested deep learning models to noise in the data.巴伐利亚州作物类型制图数据集。该数据集包含2018年获取的未预处理（raw, unpreprocessed）哨兵2号（Sentinel 2）卫星时序遥感数据，以及对应田块的作物类型标注。本数据集被用于将基于预处理后卫星数据的多款深度学习模型，与采用原始卫星时序数据的行业基准模型进行对比，以评估所测试深度学习模型对数据噪声的鲁棒性。