Project Details
Projekt
Face recognition for ox heads – Testing and application of computer vision for the identification of historical watermarks
Subject Area
Medieval History
Term
since 2026
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 576443479
The identification of watermarks has long been one of the established methods of scientific indexing and multidisciplinary research into pre-modern manuscripts and prints – primarily for the dating and localization of unmarked objects or for the reconstruction of genuine workshop and production contexts. However, the great similarity and variety of motifs, usually consisting of several elements, as used by European paper mills to mark their products since the 13th century, make it considerably more difficult to identify watermarks – especially the ox head mark, which is documented in thousands of versions. The fact that all verification portals for referencing watermarks are based on precombined monohierarchical classifications, in which similar motifs can be located at different points in the system, makes image matching even more complex. In response to this challenge, the project aims to test computer vision to increase the quality and efficiency of watermark identification in the cataloging and research of pre-modern manuscripts and prints. Specifically, this includes the development of an AI-driven automatic visual analysis tool that will assist in the identification of watermarks that are difficult to distinguish with the naked eye by suggesting the documents with the highest match rate. The innovative impulse of this application, which is to be designed as an intuitive web service for the functional expansion of material-related verification systems for watermarks as well as for manuscripts and prints, lies in the faceted indexing of the image marks by segmenting their partial motifs. The instrument, which is to be optimized for scientific use, is to be based on an ontology as a data model, which for the first time opens up the opportunity not only to reliably determine watermark fragments, but also to include the contents of the numerous filigranological reference portals in the automatic image comparison via mapping routines. Therefore, high priority is given to the robustness of the technology to be used against the multiple appearances in varying quality of the images to be analyzed as a result of the different watermarking methods. In addition, the project aims to improve the availability of open ground truth benchmarking data for machine image recognition and to systematically test the potential of multispectral imaging for the reproduction of watermarks in comparison to established removal methods such as rubbing, tracing and thermography in order to derive object-specific recommendations for action and technological standards for filigranological practice. The free reutilization of the results obtained for application to other image carriers – such as seals or coats of arms – is an important concern of the project.
DFG Programme
Research data and software (Scientific Library Services and Information Systems)
