Based on machine learning for personalized skin care products recommendation engine

摘要

With the economic development and the aging trend, the use of cosmetic products has expanded rapidly. In an ever-expanding skin care market, facial skin care product was the most popular product of skin care product. However, thousands of skin care products are available in the market. With endless options, shoppers are confronted confused and tired. Because everyone's skin condition is not the same, using unsuitable skin care products can damage the skin. Frequent problems with face skin are wrinkles, spots, acne vulgaris, pores, etc. The causes of facial lines, such as dryness, facial expressions, aging, etc., are cause different shades and different types of wrinkles. Therefore, it is very important to know your skin quality and use skin care products correctly. According to the application of different levels of image processing, it can be divided into image classification, positioning, object detection and object segmentation in the field of image vision. This paper will focus on the application of machine learning and deep learning algorithm development on human face and skin intelligence recommendation platform. That uses YOLOv4's novel object recognition algorithm to detect key features in face images, and intercept sub-images of regions of interest (ROI) as input information for multi-label models. Each sub-image detects the defective part through the YOLOv4 identifier of the second layer, and calculates the ratio of the pixel area of the local block to the main body to evaluate the correlation between feature parts and degree to establish a reference for the optimization of subsequent multi-label model. The skin condition classification uses the image processing algorithm to preprocess automatically remove, reduce noise, enhance, normalize and extract features to obtain the feature vectors of the sub-images for training the multi-label classification model. The prediction results of machine learning can provide suitable maintenance knowledge and product recommendations for users to recommend the suitable skin care products and maintenance ingredients for the user's skin condition.