Natural compounds such as biological colorants (biocolorants) have long been employed for the purpose of dying textile and represent a crucial ingredient in the mass market textile industry. However, industry-wide standards for commercialization of biocolorants are still lacking. Thus, it is beneficial to establish a database including compositionally diverse biocolorants. Moreover, it could be used as a tool to identify and authenticate biocolorant in textile end-products, to ensure their quality and safety, thereby supporting the growth of the biocolorant industry. Efficiently managing the databases and analyzing bio-dyed products data typically requires experts to organize and refine data collected from bio-based dye and pigment production. In this process, it not only requires researchers to have a certain understanding of botanical taxonomy but also knowledge about biology and chemistry.
As one part of the BioColour consortium project, our goal in this research is to take advantage of unsupervised learning for cluster analysis, to discover possible clusters of bio-dyed textile in the absence of ground truth labels or other knowledge of expert domain. This work aims to apply different approaches for unsupervised learning. Specifically, we use agglomerative clustering, Fuzzy C-means, OPTICS as well as a well-known artificial neural network (ANN), namely self-organizing maps (SOM), resulting in an investigation that combines data visualization and cluster analysis. In summary, we apply AI techniques to discover hidden clusters emerging among products colored using biocolorant, here specifically bio-dyed textile samples, and show the potential of clustering techniques in this application domain.
Writer: Zongyue Li, Master of Science, University of Eastern Finland