After every vacation, we ask ourselves the same question: How can we group the countless photos stored on our cell phone in a meaningful way so that we can find them again quickly later? For example, you remember a photo you took during a hike in the Alps showing marmots. If you are lucky, you still know the place or date when you took the picture. This makes it easy to find the photo using the automatically saved shooting date or, if this is activated, the GPS localization via the map view.
But what do you do when the photo collection - as is the case in the city archive - has grown to one million images and information about the location and date of the photos is scarce?
The city archive houses numerous historical photo collections such as the press photo archive of the daily newspaper "Der Bund" and the estate of Bund photographer and photo editor Hansueli Trachsel (1951-2019). There is also photographic documentation from municipal departments such as the building inspectorate, civil engineering office and municipal gardening department. A systematic search of these holdings for specific subjects in albums, card index boxes and hanging files is very time-consuming, if not impossible.
In order to improve access to the photo collections, the city archive is in the process of digitizing them with financial support from third parties and publishing them in the online archive catalog. Manual content indexing remains important even with digitized photos (for example, when naming the photographer).
But what happens to a stock of several thousand images for which no information is known about the location and time they were taken? If these photos are to be made available to the public, new innovative solutions are needed.
Archipanion, a web-based solution for researching digital photo collections, has been available to the city archive since the beginning of April 2024 (stadtbern.archipanion.com). The service is provided by the company 4eyes. It uses artificial intelligence to group photographs according to probabilities based on visual criteria. It is based on the retrieval engine «vitrivr» from the "Databases and Information Systems" research group at the University of Basel.
With Archipanion, around 8,000 digitized photos from the collections of individual photos, the Bümpliz local archives and the Bern municipal gardening department can be sorted by individual terms such as "bears", "car" or "people in the snow" and displayed on the screen. You can also search for texts that appear on the photos. Clicking on a photograph also opens up the option of displaying similar image content or switching directly to the online archive catalog of the city archive. There, the "Show all" button displays the descriptive information for the selected photograph.
Intelligent search functions such as Archipanion are at the beginning of a whole series of new technical developments. These open up additional possibilities for the public to find the documents, plans and photographs stored in the city archive more easily and use them for other purposes. In the near future, it will be possible to search for prominent personalities in historical photo collections using facial recognition - as is already common practice in smartphones today. Automated recognition of objects and places of interest using existing encyclopaedia entries such as Wikipedia is also conceivable.
In another application that already exists today, photos of buildings in the city of Bern can be located on the city map using georeferencing and retrieved together with other building-related information such as building permits (map.bern.ch/stadtplan).