UNET (different configurations) - Simple UNET, DeepUNET, DeepLabV3+ with backbones
The dataset is divided into two parts: one is orthophotos with 10 cm per pixel resolution and the other is a building mask image created on the basis of cadastral data (rasterization of cadastral data). Dataset was created on the basis of data provided by Head Office of Geodesy and Cartography (GUGiK) (at https://mapy.geoportal.gov.pl/imap/Imgp_2.html?gpmap=gp0&locale=en)
Glinka, S.; Owerko, T.; Tomaszkiewicz, K. Using Open Vector-Based Spatial Data to Create Semantic Datasets for Building Segmentation for Raster Data. Remote Sens. 2022, 14, 2745. https://doi.org/10.3390/rs14122745