Automated Geospatial Annotation of Historical Photographic Collections (FSGR-Historic)

The project examines (semi-)automatic georeferencing for historical image collections via advanced computer vision methods.

Key data

Description

The objective is to systematically examine the feasibility of (semi-)automated georeferencing of historical photographs using modern computer vision methods and to georeference a portion of the Fotostiftung Graubünden archive.

Central to the study is the question of how reliably AI models can suggest accurate locations despite changes in photographic techniques, landscape, and the built environment.

This will be evaluated based on the hit rate on a manually labeled dataset.