The LandCover.ai (Land Cover from Aerial Imagery) dataset is a dataset for automatic mapping of buildings, woodlands, water and roads from aerial images.

Dataset features

• land cover from Poland, Central Europe ¹
• three spectral bands - RGB
• 33 orthophotos with 25 cm per pixel resolution (~9000x9500 px)
• 8 orthophotos with 50 cm per pixel resolution (~4200x4700 px)
• total area of 216.27 km²

Dataset format

• rasters are three-channel GeoTiffs with EPSG:2180 spatial reference system
• masks are single-channel GeoTiffs with EPSG:2180 spatial reference system

We provide split.py to split images into 512x512 pieces, and following files: train.txt, val.txt and test.txt containing lists of pieces used for training, validation and testing respectively.

Version 1

We added roads and refined some annotations.

• classes: building (1), woodland (2), water(3), road(4)
• areas: 1.85 km² of buildings, 72.02 km² of woodlands, 13.15 km² of water, 3.5 km² of roads

Paper Download

Version 0

Note: This version was not peer reviewed and has not been officially published. If possible, please use updated Version 1.

• classes: building (1), woodland (2), water(3)
• areas: 1.85 km² of buildings, 72.22 km² of woodlands, 13.25 km² of water

Paper Download

To cite our work, please use the following:

@InProceedings{Boguszewski_2021_CVPR,
    author = {Boguszewski, Adrian and Batorski, Dominik and Ziemba-Jankowska, Natalia and Dziedzic, Tomasz and Zambrzycka, Anna},
    title = {LandCover.ai: Dataset for Automatic Mapping of Buildings, Woodlands, Water and Roads from Aerial Imagery},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
    month = {June},
    year = {2021},
    pages = {1102-1110}
}

This dataset is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

If you encounter any problem or have any feedback, please contact landcoverai@linuxpolska.pl