junboboshi/ETCI-2021-Flood-Detection

ETCI 2021洪水检测数据集是一个全面的洪水检测分割数据集，专注于由ESA Sentinel-1卫星拍摄的SAR（合成孔径雷达）图像。该数据集提供了VV（垂直发射，垂直接收）和VH（垂直发射，水平接收）极化图像对，这些图像已经通过Hybrid Pluggable Processing Pipeline (hyp3)进行处理。此外，数据集还包括相应的二进制洪水和水体地面实况掩码。数据集由66,810个256×256像素的图块组成（每个极化33,405个），分别分布在训练集、验证集和测试集中：每个极化分别有33,405、10,400和12,348个图块。每个图块包含3个RGB通道，这些通道是通过将54个标记的GeoTIFF文件（由Sentinel-1 C波段合成孔径雷达图像数据使用Hybrid Pluggable Processing Pipeline hyp3生成）进行切片转换得到的。训练图块对应VV和VH极化的强度值，并具有以下属性。ETCI 2021数据集对洪水检测和分割任务非常有价值，并促进了该领域的研究和开发。The ETCI 2021 Flood Detection Dataset is a comprehensive dataset for flood detection and segmentation, focusing on Synthetic Aperture Radar (SAR) images acquired by the ESA Sentinel-1 satellite. It provides paired VV (vertical transmit, vertical receive) and VH (vertical transmit, horizontal receive) polarized image pairs, which have been processed via the Hybrid Pluggable Processing Pipeline (hyp3). Additionally, the dataset includes matching binary flood and water body ground truth masks. The dataset comprises 66,810 256×256 pixel image tiles, with 33,405 tiles for each polarization. These tiles are divided into training, validation, and test subsets: for each polarization, there are 33,405 training tiles, 10,400 validation tiles, and 12,348 test tiles respectively. Each tile contains 3 RGB channels, which are derived by slicing and converting 54 labeled GeoTIFF files. These GeoTIFF files are generated from Sentinel-1 C-band SAR image data using the Hybrid Pluggable Processing Pipeline hyp3. The training tiles correspond to the intensity values of VV and VH polarizations, and possess the following attributes. The ETCI 2021 dataset holds significant value for flood detection and segmentation tasks, and promotes research and development within this domain.