Textiles and Data: Application of Computer Vision to Cross Collection Characterisation of Historic Silk Textiles

This paper presents the results of my PhD research which applies computer-based imaging technologies to examine historic silk production evidence. My program combines high-resolution images with a computer vision software application to measure identifiable quality and workshop characteristics for weft-faced compound weave figured silks attributed to Mediterranean workshops between ca. AD 600-1200.

For a variety of reasons, research for this category of textiles has slowed in recent years. While essential to protect fragile textiles from damage, the consequence of conservation standards has been reduced collections access. Resource constraints and changes in museum practices mean that many institutions now focus on exhibitions rather than research. At some institutions, large textile collections built up on the heels of the antiquarian era now languish. Even at well-resourced institutions, there is little opportunity for research.

Dramatic advances in digital imaging provide opportunities for the development of new methods for investigation and documentation. My research protocol combines a research grade digital microscope with a custom-built stage to perform precise digital ‘sampling’ for measurement of textile attributes including yarn characteristics, textile structure, density and pattern unit features. The computer vision application aids in error detection, providing a form of ‘industrial inspection’ for ancient textiles. The outcome is a set of objective and reproducible measurements enabling specific comparison of attributes across different collections.

By using my portable equipment setup, I was able to record 125 silk fragments in ten different collections in North America and Europe. Analysis demonstrates patterns of work practices and imitative pattern reproduction among workshops. Results also help to re-unify textiles divided in antiquity or after excavation. In the future, this methodology could provide the basis for a shared database of images available to a broader community of researchers as well as supporting the work of conservators.