A Data-Driven Approach for Fibres Recognition via Spectrophotometry

The increasing volume of textile waste presents significant environmental and economic challenges, necessitating the development of efficient automated sorting techniques to support a more effective textile waste recycling. Automated sorting is a notoriously complex task, due to deployment constraints and to the variability of textiles. To advance the work on automated textile sorting, this study investigates the use of data-driven approaches on spectrophotometer-based reflectance measurements for recognising fibres. Spectrophotometry offers significant advantages in terms of operational simplicity and reliability, making it a promising choice for use in textile sorting facilities where environmental conditions are difficult to control. Considering an extensive dataset of specifically acquired pure textile samples, in this work we leverage on AutoML solutions to determine the best architecture to discriminate between cotton and polyethylene terephthalate (PET) fibres.