Historical Norwegian Aerial Imagery Land-Cover Dataset

The Historical Norwegian Aerial Imagery Land-Cover Dataset is a large-scale collection of high-resolution, orthorectified aerial image patches paired with pixel-level land-cover annotations. Derived from the Norge i Bilder archive and previously validated rasters, the dataset includes 3.46 million 512\u00d7512 grayscale patches, each provided as a PyTorch tensor containing both the image and its corresponding label mask. Six land-cover classes are represented: Urban, Vegetation, Water, Wetland, Agriculture, and Bare. This dataset enables deep learning research on historical aerial photography and supports applications in land-cover mapping, environmental reconstruction, and remote sensing model development.