Image Archive on Former German Colonies to be Made Accessible with Artificial Intelligence // The new project “Visual Analytics for Images from Colonial Contexts” (VABiKo) is a German Research Foundation-funded collaboration between Frankfurt University Library and Philipps-Universität Marburg
The central historical image archive of the German colonial and colonial revisionist movements is one of the largest photographic collections of its kind and plays an important role in the reappraisal of German colonial history. This image archive will be made accessible on a public internet portal using automated image analysis processes and deep learning methods. Its position in both colonial and Nazi contexts necessitate an approach that is both historically aware and ethically responsible.
The still existing part of the image collection belonging to the “German Colonial Society” (1887 – 1936) and subsequent propaganda organizations such as the “Reichskolonialbund” (1933 – 1943), comprises around 45,000 image carriers. In the late 1940s, these were handed over to the Frankfurt City and University Library. From the perspective of the original collecting institutions, many of the images relate thematically to the alleged cultural and economic “development” of the German colonies on the African continent, in Oceania, and in China. However, other European and non-European regions are represented as well. In the late 1990s, this was the first photographic collection from colonial contexts in Germany to be digitized and presented online. In the early 2000s, it was expanded to include other image collections such as those of the Sam Cohen Library in Swakopmund, Namibia. Due to insufficient metadata, access for researchers has so far been severely restricted.
The project “Visual Analytics for Images from Colonial Contexts” (VABiKo) was launched in order to fundamentally rethink possibilities of access to these images. The project is a cooperation between the J. C. Senckenberg University Library in Frankfurt and the “Multimodal Modelling and Machine Learning” working group (Prof. Ralph Ewerth) at Philipps-Universität Marburg. It is funded by the German Research Foundation (DFG) for a provisional period of three years. The project aims to improve the archive’s usability through semi-automatic indexing and additional metadata. This includes new possibilities for extended search and online presentation, such as simplified searchability with the help of deep learning-supported image analysis and interactive time-space visualizations. In addition, this will make it possible to compare the image collection with illustrations in publications from colonial contexts. In this manner, the original contexts of use and the instrumentalization of these images become evident. By automatically recognizing recurring image patterns relating to people, objects and places, it will be possible to display the places where the photographs were taken on maps with an additional timeline. A general improvement in data management will ensure the project’s sustainability for future generations of users.
The resulting multilingual internet portal should be publicly accessible by the end of 2027 and support the reappraisal of German colonial history. The collection contains depictions of various forms of violence within an imbalance of power between the colonizers and the colonized. For this reason, ethical questions concerning presentation and use are being discussed with colleagues, researchers and stakeholders from the countries of origin. Despite the originally propagandistic contexts, the photographs have great potential for reinterpretation by the aforementioned interest groups. The aim is to achieve a balance between transparent accessibility and a respectful approach to the people photographed.
About Johann Christian Senckenberg University Library (UB JCS)
With its extensive holdings and collections, the JCS University Library is one of the most important academic libraries in Germany. It combines the functions of a university library with numerous regional responsibilities, acts as an academic library for the city of Frankfurt and the Rhine-Main region as well as a principal library within the supra-regional literature and information supply. Link: https://www.ub.uni-frankfurt.de/
About the Multimodal Modelling and Machine Learning research group (Prof. Ralph Ewerth) at Philipps-Universität Marburg
Prof. Ewerth’s research group at the Department of Mathematics and Computer Science at Philipps-Universität Marburg was founded in April 2025. It is one of three newly established AI professorships that were created together with the Hessian Centre for Artificial Intelligence (hessian.AI). The working group concentrates on the development and improvement of machine learning methods for analyzing images, videos and multimodal information. Another focus is on methods for improving (multimodal) search engines using generative AI models and on interdisciplinary applications.
Press release from Goethe University Frankfurt.