Supervised and unsupervised textile classification via near-infrared hyperspectral imaging and deep learning

Recycling textile fibers is critical to reducing the environmentalnimpact of the textile industry. Hyperspectral nearinfraredn(NIR) imaging combined with advanced deep learningnalgorithms offers a promising solution for efficient fiber classificationnand sorting. In this study, we investigate supervisednand unsupervised deep learning models and test their generalizationncapabilities on different textile structures. We shownthat optimized convolutional neural networks (CNNs) and autoencodernnetworks achieve robust generalization under varyingnconditions. These results highlight the potential of hyperspectralnimaging and deep learning to advance sustainable textile recyclingnthrough accurate and robust classification.