This project has two main objectives:

  1. Create a prototype for scholars and students to help identify relevant features of manuscript illumination.
    1. Automatically detecting illuminations on a page and extract them
    2. Automatically recognising objects
    3. Exploring annotation system specific to medieval manuscripts
    4. Recommendation system for confirming or expanding annotations
    5. Classifying manuscripts by the objects, and combination of objects, they depict
    6. Combining extracted data on image with HTR output (surrounding textual data)
    7. Visualizing the images according to patterns or user queries
    8. Implementing active learning strategy
    9. Automatically recognising people, actions, subjects, based on the presence of a combination of objects and textual clues.
  2. Create Illuminator profiles using several features such as colours, saturation, edges, symmetry, objects, and other details to determine specific patterns. It includes localisation and dating details to determine specific illuminator profiles and track evolutions. The aim is also to find evolutions and patterns according to dates and places, to detect what is different across a corpus.

A project in collaboration with Dr. David Wrisley, New-York University Abu Dhabi; Christofer Meinecke, Universität Leipzig and Dr. Stefan Jänicke, University of Southern Denmark.