Computational Analysis of Visual Features from Digitized Manuscripts

The study of ancient written artifacts has led to an ever-growing repository of digital data, encompassing raw images of artifacts and automatically generated outputs from advanced acquisition techniques. Manual analysis of this data is not only time-consuming but also prone to human error and bias. Consequently, the integration of novel machine learning and AI methodologies offers significant potential to streamline and enhance the efficiency of artefact research. Computer vision, in particular, provides opportunities beyond handwriting text recognition to address complex challenges in manuscript studies. As a computer scientist, exploring the boundless possibilities of computer vision to resolve research questions in this field has been an enriching experience. This lecture delves into key challenges encountered throughout this journey, highlighting ongoing collaborative projects. These projects illustrate the wide range of diverse research topics, including detecting specific handwriting styles on palm leaves from south India, measuring the density of sieve prints in Himalayan paper-making, addressing the complexities of multilingual texts and varied writing tools in Rilke’s Notebooks from Austria, and even restoring undertexts in Georgian palimpsests. These projects demonstrate the transformative potential of computer vision in advancing manuscript studies.