A Novel Strid CNN Model for Cosmetic Product Recommendation based on Skin Type and Tone

In the last few years, consumer interest in cosmetic purchasing has increased globally, especially in skincare products. Traditionally, customers have used in-store advice from beauty experts for popular product purchases. However, due to the different skin types and skin tones of every individual, sometimes it is difficult to get the correct product that is suitable for them. This indicates the need for customized and effective methods for analyzing cosmetic products suitable for individual skin types and skin tones. This research proposes a novel cosmetic product recommendation system that addresses different consumer needs using a Strid Convolutional Neural Network (CNN). The proposed method identifies customers' different skin types (oily, normal and dry) and skin tones (medium tan, fair/light, and dark/deep). The proposed method takes the identified skin type and tone as input and then recommends the best-suited cosmetic or skincare product related to their skin tone and type. The dataset used to have 3152 images of different skin types and 1305 images of different skin tones. The overall dataset forms the basis of the proposed system, which identifies these parameters to offer customized recommendations that are most suitable cosmetic products like sun protection, eye creams, cleansers, face masks, and moisturizers, according to individual customer needs. The proposed Strid CNN architecture demonstrated excellent performance, achieving a validation accuracy of 98.8% and a validation loss of 0.0356, highlighting the model's enhanced capability to provide effective cosmetic product recommendations for different skin types and tones.