Similarity Measurement of Visual Patterns in Written Artefacts
2023–2025
RFA05
In the past few years, automatic pattern analysis proved to be a powerful and useful tool for the study of written artefacts when it is developed and used properly. Machine learning and artificial intelligence approaches have been applied to visual and tabular data in order to detect, classify, and recognise various types of patterns in written artefacts. Nevertheless, the full potential of machine learning has not been utilised yet, mainly due to the lack of sufficient annotations and training data.
This project aims at developing novel methods of similarity measurements for visual patterns in written artefacts. These methods can be used for classification, detection, and recognition of visual patterns such as handwriting styles, seals, and drawings. State-of-the-art learning-based approaches are utilised, and the possibility of minimising their need for labelled (annotated) data is investigated.
The main research objectives of this project are as follows: investigating the possibility of acquiring similarity metrics from small sets of digitised written artefacts as annotated training samples; generalising these metrics to unseen classes of patterns in order to retrieve visually similar samples; and integrating these metrics into pattern classification and clustering systems.
The work achieved during the first phase of this project has yielded very promising results and demonstrated the potential of computational pattern analysis for the study of written artefacts. Therefore, new and ongoing cross-disciplinary collaborations will be carried out across several fields of research.
People
Project lead: Hussein Adnan Mohammed
Research Associate: Mahdi Jampour
Preceding projects
Pattern Recognition in 2D Data from Digitised Images and Advanced Acquisition Techniques (2019–2022)
Project lead: Hussein Adnan Mohammed