Geolocating Arctic Archive Photos with AI: Revisiting Places from Century-old Photographs
Abstract
This talk presents an experimental application of AI to Arctic cultural heritage: attempting to geolocate historical black-and-white photographs from the 1913-1915 Norwegian Reindeer Grazing Commission archives without using any ground truth geographical metadata. This poses as both a cultural heritage challenge and an interesting AI research problem.
Nikiforos will demonstrate a multimodal approach which combines CLIP and Zero-Shot model classification to predict estimated photograph locations. The results are inconsistent but also revealing: when conditions align, the system achieves remarkable accuracy (essentially finding the exact same location photographed 100 years apart). But the approach also fails unpredictably, exposing bottlenecks in historical image geolocation.
Through detailed case studies of both successes and failures, Nikiforos will explore what works, what doesn't, and why. Beyond technical challenges, the project also raises important questions about applying AI to indigenous cultural materials ethically and respectfully.
Speaker
Nikiforos Staveris is a Master's student in AI at Umeå University, originally from Athens, Greece. In September 2025 he was selected by the UmU Arctic Centre to participate in the interdisciplinary course "Arctic Future Pathfinders", organized by UiT, which sparked his interest in how AI can be implemented ethically and for the benefit of Arctic communities. This project on geolocating historical Sami photographs is a direct result of that experience.
Please note. The subtitle is generated by AI and may contain some errors.