An Exploratory Analysis of Evaluation Criteria for Glossiness in Base Makeup Cosmetics Utilizing Light Reflection Measurement and Machine Learning

This study investigated gloss perception in base makeup and its relationship with optical characteristics. Using foundation on artificial skin to minimize variability, we established an evaluation system combining sensory evaluation, microscopic photometry, and Convolutional Neural Networks (CNN) for luminance analysis. Results revealed two distinct perceptual criteria: one prioritizing visual attributes such as brightness, and another emphasizing textural impressions like smoothness. Optimal light-receiving angles varied by criteria; specular reflection predominated for visual-focused groups, whereas diffuse reflection was more critical for texture-focused groups. Furthermore, the luminance regions contributing to these criteria differed even under identical conditions. These CNN-based findings were consistent with analysis based on Gray-Level Co-occurrence Matrix (GLCM) features, validating the developed model. These findings indicate that gloss perception cannot be defined by a single physical indicator, and that measurement conditions need to be selected to match specific perceptual criteria.