Collection insight and interconnectivity through artificial intelligence image analysis: A collaboration with the National Archives of Estonia

Abstract

Descriptive metadata information is paramount to creating a holistic historical record for researchers, scholars and the general public, as well as other institutions to use both internally and externally. Metadata contextualise image-based collections, providing insight into the individuals, objects and stories contained within them, allowing for the sharing and cross-referencing of information and collections. However, gathering descriptive metadata for image collections is time-consuming and costly to generate, often relying the knowledge of a single person. And yet, if researchers do not extract, catalogue and link this information to the images, valuable context can be lost, obscured by the sheer volume of images that one would have to review manually in order to find specific content. This paper demonstrates how the application of artificial intelligence (AI) image analysis offers an alternative to the traditional labour-intensive manual approach and provides important identifications within collections that allow for further research, collaboration and accessibility. The paper describes how, in partnership with the National Archives of Estonia, we digitised over 100,000 film frames and glass-plate negatives, collected from 22 different institutions around the country, and applied AI image analysis to the digital assets in order to take the first step in successfully identifying individuals within the collections, unlocking their stories, interconnecting these institutions’ collections and the broader cultural heritage community.