Digital Lunch Seminar Series

Computational Visual Cataloguing: A Case Study of Rilke’s Notebooks

Hussein Mohammed and José Maksimczuk  

This talk presents a case study on Rainer Maria Rilke’s notebooks. In this work, page-level and word-level attributes, such as orientation, colour and writing implement, are detected and linked to specific locations. This process yields “computational visual catalogues” that enable corpus-wide querying and materially attentive analysis. The approach demonstrates that even the text itself contains salient visual information beyond transcription, and that layout, spatial relations and features of the writing support can be modelled at scale. Brief reference will be made to an accompanying interactive explorer that overlays predictions on images to ensure transparency of inference, while the emphasis remains on the underlying scientific methods. Overall, computer vision is positioned as a means to analyse visual structure in manuscript images, not merely a conduit to textual surrogates. As such, the preparation of edited texts can be informed by page-anchored evidence detected automatically, including layout, spatial relations, script variation and material features, that standard transcription does not capture.