Data supplement: Mapping small landscape features in agricultural lands at the national level using CNN- based semantic segmentation

Data supplement: Mapping small landscape features in agricultural lands using CNN-based semantic segmentation (https://doi.org/10.1080/10106049.2026.2691353) The dataset contains: model.zip: pretrained model weights used in the small landscape feature segmentation samples.zip: samples consisted of 500x500px images and mask pairs used in model training and validation model_output_field_islands_polygons.zip: final, post-processed model outputs of small landscape feature polygons across Estonia in geopackage format The terminology of Small Landscape Feature that we used in our study refers to patchy natural or semi-natural vegetation features (locally called field island) that may include trees, shrubs, rocks, and small ponds, with areas ranging from 0.01 to 0.5 ha. This definition follows the criteria set by local regulation and used by the Common Agricultural Policy (CAP)-paying agency in Estonia. Known limitation of the model: Commission error of hedgerows (false positives) Omission error of the field islands lacking a tree canopy If the samples or model outputs are used in a scientific publication, we kindly request that you contact us to discuss the possibility of co-authorship. Otherwise, please cite this dataset using the provided citation.