Image Recognition for Printed Early Modern Portraits as an App (PortApp)

The project is a cooperation between the Herzog August Bibliothek Wolfenbüttel (HAB) and Prof. Dr. Thomas Mandl from the Institut für Informationswissenschaft und Sprachtechnologie (IWIST) at Hildesheim University.

In the Early Modern Period (about 1500–1800) printed portraits were an influential and widespread genre of images. A broad range of social groups were depicted, and also the collecting of portraits was a widespread habit. Recently, research has begun to consider these printed portraits again as an autonomous object of interest. Studies investigate the representation and self-depiction of social and professional groups, the function of the portrait in its original context, and the practice of collecting itself. As printed portraits collected in libraries and archives have mostly been separated from their publication context and as their appearance is strongly influenced by their producer, technique or graphical model and by particular iconographical traditions, there is a need in research for an automated image search optimized for portraits. On the one hand, the search is meant to enable researchers to trace portraits to the books they were originally published in. On the other hand, it shall make visible similarities between portraits of different persons which are not evident from the usual metadata. The HAB has a collection of about 32,000 early modern portraits, which have been fully catalogued and digitized. About as many portraits, catalogued only superficially, are contained in the libraries’ books collections. The digitized portraits and their metadata serve as content for the project’s image search. This content will be expanded by digitizing portraits from books and by extracting portraits from books already digitized by other libraries.
The image recognition software will be developed by an intense cooperation between humanities scholars and information scientists. Image recognition technology can be selected for answering questions concerning content and meaning of images, whereas similarities brought to light by computerized processes can be assessed related to its historical significance. Fruitful ways of searching can be refined technically and integrated into the user interface. The search will rely on both traditional approaches to image processing and on advanced algorithms like Convolutional Neural Networks (CNNs). The portrait search will be made available as a public web interface and metadata service. In addition, an app for mobile devices will be developed within the project, enabling users to take pictures and submit them to the search directly. The app will constitute a user-friendly, advanced tool for image research which can be used by scholars interested in portraits all over the world.



Funding: Land Niedersachsen via VolkswagenStiftung
Duration: January 2020–December 2022

Contact (HAB): Dr. Hartmut Beyer, Dr. Hole Rößler