Radiant Earth Foundation, a U.S.-based nonprofit organization that creates machine learning resources for sustainable development goals, has created the world’s first global dataset of satellite images to train land cover classification models. The dataset contains 8,941 satellite images of seven types of land cover, including water, natural bare ground, artificial bare ground, woody vegetation, cultivated vegetation, semi-natural vegetation, and permanent snow and ice, as well as human annotations specifying the type of land cover identified in the image. Researchers can use the dataset to train models to locate changes in the global landscape, such as urbanization or deforestation.
↧