Deep Discoveries: A Towards a National Collection Foundation Project Final Report

The Deep Discoveries project explored the potential of computer vision (CV) search for content discovery within and between our nation’s digitised image repositories. The research led to the design of a prototype search platform enabling cross-collection image linking by harnessing the ability of CV methods to identify and recognise visual patterns without the need for preliminary integrated descriptive metadata. Searching in this manner allows for content-linking based on attributes such as pattern, colour, and motif, and creates the opportunity for users to discover unforeseen connections between image collections across the country. The research also introduced explainable AI methods, which allow users to enter into a visual dialogue with the AI so as to refine their search tasks. During the 18-month project, research carried out by a user experience research (UXR) team from two GLAM Independent Research Organisations (The National Archives and the V&A Museum) informed the work of computer vision scientists at the University of Surrey. Using an agile working methodology and design sprints, the technological advances and UXR findings were integrated into a prototype design by consulting Interaction Design (ID) partners from Northumbria University. The Deep Discoveries project worked with four partner organisations representing different owners and creators of visual collections to open up participation in funded research to smaller organisations, to glean a better understanding of their needs, and to assess the opportunities and challenges involved in gathering visual collections for the purpose of employing CV-based search and discovery tools.