LandCoverNet is a global annual land cover classification training dataset with labels for the multi-spectral satellite imagery from Sentinel-1, Sentinel-2, and Landsat-8 missions in 2018. Radiant Earth generated the dataset using 300 geographically diverse tiles of ESA's Sentinel-2 mission, covering the regions of Africa, Asia, Australia and Oceania, Europe, North America, and South America.
There are six individual training datasets. They comprise a total of 8,941 image chips of 256 x 256 pixels, labeled globally, resulting in ~586 million pixels for the entire training datasets. Each pixel is classified into one of seven land cover classes based on its annual time series. These classes are water, natural bare ground, artificial bare ground, woody vegetation, cultivated vegetation, (semi) natural vegetation, and permanent snow/ice.
The human annotation process generates a consensus score that is associated with each labeled pixel, indicating the uncertainty of the classification process. The LandCoverNet training datasets are an essential resource for researchers working on land cover classification and analysis using remote sensing data.