Map identification
75M maps from the Internet: https://huggingface.co/datasets/sraimund/MapPool
Map patterns and content description
Text annotation recognition, scene classification, super-resolution reconstruction, and style transferring of 20k maps: https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/ZBXJD5
Map features recognition, map scale interpretation, turn-by-turn navigation with more than 2000 samples: https://github.com/ungquanghuy-kddi/CartoMapQA-Dataset
1200 geographic thematic maps annotated for visual question-answering and legend detection tasks: https://huggingface.co/datasets/koukeft/visqam
Classification of historical map patches
Railspace and buildings in Great Britain: https://zenodo.org/records/7147906
Historical map segmentation
Paris maps: https://zenodo.org/doi/10.5281/zenodo.4817661
Paris and World maps: https://zenodo.org/doi/10.5281/zenodo.5497933
Swiss railroads and vineyards: https://doi.org/10.3929/ethz-b-000691430
Taiwanese maps: https://doi.org/10.6084/m9.figshare.25560276.v1
U.S. Natural map features: https://github.com/ASUcicilab/GeoImageNet + https://www.sciencebase.gov/catalog/item/5ec6aad082ce476925eddbc6
Vietnamese maps: https://github.com/h-oneohone/islandMapVN-dataset
Text extraction from historical maps
David Rumsey Collection: https://searchworks.stanford.edu/view/rc349kh8402 + https://doi.org/10.5281/zenodo.10608900
Napoleonic Cadastre Collection: https://doi.org/10.5281/zenodo.10610731
U.S. Spot elevations: https://osf.io/9pgdf/?view_only+=+=c294460771114afc9ba0253bc2d8fc0d
Geolocation
OpenStreetView: https://huggingface.co/datasets/osv5m/osv5m
Eye Tracking
OnMapGaze: https://zenodo.org/records/11520659
EyeCatchingMaps (saliency): https://zenodo.org/records/10619513
If you like any dataset to be added, please fill in the form below. Please note that the dataset needs to be used with a deep learning model for a cartographic purpose. Ideally, the dataset is split into training and testing data, and some evaluation metrics are given.