Skin Type Prediction and Product Suggestion using Machine Learning and Image Processing

Skincare diagnostics are part of an emerging field of research that seems quite promising for personalized skincare recommendations in view of machine learning and image processing. This paper focuses on the classification of skin types, namely oily, dry, normal, and combination, using facial images through CNNs, aiming at improving skincare decisions and beauty product recommendations. In this work, several architectures of CNNs were compared including MobileNet-V2, EfficientNet-V2, and ResNet-V1; among these, the best performance was delivered by EfficientNet-V2. It combines image preprocessing and feature extraction of skin features through MTCNN with weather data to provide customized skincare routines. Such a combination of image processing with environmental factors makes this system strong and specific for the user in terms of skin type classification and skincare management.