GeoWerkstatt December: Teaching AI to Understand Historical Maps

15th December 2025

Results of the map segmentation model. The source and target images are first encoded to deep features by the Image Encoder. Then the Prompt encoder merges the information from the source label and the encoded source image to capture the information of what patterns should be segmented. This information then guides the Mask Decoder to generate the segmentation of the target image that matches the patterns showcased in the source image-label pair. The Model is trained with Cross Entropy and Dice Loss.

Historical maps are treasures of the past — they record landscapes, settlements, and land use patterns that no longer exist today. To unlock this valuable information, researchers are turning to artificial intelligence (AI). This GeoWerkstatt project aims to train AI models to automatically recognize and locate different elements like forests, rivers, or towns on historical maps.

Read more about the project:

GeoWerkstatt project of the month December 2025

Info GeoWerkstatt - Insights into research

he GeoWorkshop appears monthly here on this website. Scientists take turns reporting on their current research projects in the field of geodesy and geoinformatics. The Geodetic Institute (GIH), the Institute of Geodesy (IfE), the Institute of Cartography and Geoinformatics (IKG), and the Institute of Photogrammetry and GeoInformation (IPI) are all involved.

All GeoWerkstatt articles are available via the archive: All GeoWerkstatt articles

Read more about Studying Geodesy and Geoinformatics

Published by emm

Go to list