tacofoundation/cloudsen12

CloudSEN12+是一个专门用于云和云影检测的大型专家标记像素数据集，基于Sentinel-2卫星数据。它提供了三个不同的模式，包括基于Sentinel-2 Level-1C和Level-2A图像的数据，以及额外的元数据集，以增强对地形的理解。数据集包含四个类别：晴朗、厚云、薄云和云影。所有图像都已填充至512x512和2048x2048的大小，以确保补丁可被32整除。数据集遵循TACO规范，并在公共领域发布，允许任何人免费使用、修改和分发。

The CloudSEN12+ dataset is a large expert-labeled pixel dataset for cloud and cloud shadow detection using Sentinel-2 satellite data. It offers three distinct modes, including data based on Sentinel-2 Level-1C and Level-2A imagery, as well as an extra metadata collection for enhanced landscape understanding. The dataset includes four classes: clear, thick cloud, thin cloud, and cloud shadow. All images are padded to 512x512 and 2048x2048 sizes to ensure patch divisibility by 32. The dataset follows the TACO specification and is released into the public domain under the CC0 license, allowing free use, modification, and distribution by anyone.