Currently, there is significant progress in the interpretation of remote images of the earth’s surface using various recognition systems. To train these systems, large labeled datasets are required. The creation of such datasets is carried out by experts and is quite a time-consuming task. This paper proposes a method for creating specialized datasets for their use in monitoring the state of agricultural lands undergoing degradation. The method allows you to automate the process of creating such datasets The datasets are generated from freely available data obtained by the Sentinel satellites as part of the European Copernicus space program. The method is based on the processing of the results of the preliminary classification of images of the earth’s cover, previously produced by the Sentinel Hub service.