Proposal for cuticle damage classification using deep learning in hair care

A method for classifying hair damage based on cuticle morphology was investigated. Hair cuticles, used to categorize hair damage into seven types based on a previously established classification, were observed using a metal microscope, a cost-effective tool commonly utilized in the beauty industry. In this study, we applied deep learning to cuticle images that were labeled according to damage type to facilitate automated classification. Six well-known deep learning models were evaluated, and ResNet50 achieved the highest accuracy, exceeding 0.76. The proposed methodology provides an objective means for users to identify their specific hair damage, thereby supporting the selection and implementation of appropriate hair care regimens.