Recent research has paid increasing attention to printed portraits as an independent genre. Current investigations focus on the conventions by which social groups were represented and portrayed, the function of the portrait in the particular setting in which it was displayed and the practice of portrait collecting as such. Because prints of portraits have often come down to us devoid of their original contexts and because their appearance is powerfully influenced by producers, techniques, models and iconographic traditions, the research world is in urgent need of an image-similarity search engine that is optimised for portraits. The search system should be able to identify the publication context of portraits that have come down to us in isolation, while also making it possible to find similarities between portraits of different people that go beyond the verbally explicit. The HAB has a collection of c.32,000 portraits that have been catalogued in depth and digitised as well as numerous portraits in its book holdings that have not been properly catalogued. These collections will provide the material basis for the projected portrait image search. This process will be supplemented by the analysis of portraits in digitised books held by other institutions.
The image-similarity search will be developed in the context of intense cooperation between humanities scholars and IT developers. In the process, image-recognition techniques will be systematically used to answer hermeneutical questions, while at the same time the conclusiveness of machine-identified similarities will be evaluated. Sound search methods will be developed further and made available via the user interface. In addition to classical image-processing approaches, advanced algorithms such as convolutional neural networks (CNNs) will be used. The image search will be made available both on a public web interface and in an app for mobile devices. The app will provide image identification and comparative research into portraits using the camera function, thereby providing a tool for working with early modern portraits that is easy to use and can be deployed worldwide.
In cooperation with Stiftung Universität Hildesheim (University of Hildesheim), Institut für Informationswissenschaft und Sprachtechnologie (Institute for Information Science and Natural Language Processing, IWIST), Prof. Thomas Mandl
PURL: http://diglib.hab.de/?link=105
Funding: The State of Lower Saxony via VolkswagenStiftung
Duration: January 2020 – December 2022
Project participants: Dr Hartmut Beyer (contact), Dr Hole Rößler (contact), Dr Nina Niedermeier (team member), Sebastian Diem (team member, Hildesheim)
