The Bavarian Building Dataset (BBD) is an analysis-ready dataset providing an openly available 40cm image dataset over Bavaria (CC-BY4.0, Digitales Ortophoto 40cm - DOP40, https://geodatenonline.bayern.de/) combined with building footprint information as GIS data (Shapefiles) and co-registered imagery. This data has been taken from official building footprints as published openly (CC-BY 4.0, Hausumringe, https://geodaten.bayern.de/). In addition, our dataset contains building footprint geometry (Shapefile) and co-registered masks based on a historic OpenStreetMap extract dating to 2021. The data has been preprocessed and co-registered and is provided as a set of tiles ready for deep learning and other data analysis techniques. The tileset is available as 18.205 tiles of 2500x2500 pixels (each 40cm x 40cm) as well as as a collection of about 5.5 million tiles (210 GB, PNG) of 250x250 pixels for immediate use with computer vision models. The dataset is supposed to support image segmentation research, building footprint extraction research and further interdisciplinary research both on methods and on Bavaria. The dataset is supposed to be updated with each major re-publication of the upstream data sources to foster change detection research in the future.
The data was processed using a geospatial data processing engine in order to bring the original datasets into an analysis-ready representation suitable for adoption by communities and computational environments that do not model the geospatial aspects. Concretely speaking, images have been manually processed, the cartographic rendered mapnik has been used to transform geodatabases into georeferenced images, and Ohsome API was used to create a snapshot of OpenStreetMap information from a suitable timestamp.
The dataset is compiled from varous sources:
If you use this dataset, please cite :
Martin Werner, Hao Li, Johann Maximilian Zollner, Balthasar Teuscher, and Fabian Deuser. 2023. Bavaria Buildings - A Novel Dataset for Building Footprint Extraction, Instance Segmentation, and Data Quality Estimation (Data and Resources Paper). In The 31st ACM International Conference on Advances in Geographic Information Systems (SIGSPATIAL 23), November 1316, 2023, Hamburg, Germany. ACM, New York, NY, USA, 4 pages,https://doi.org/10.1145/3589132.3625658.