A deep convolutional network-based scoring model for the beauty of Asian women

Abstract

The invention discloses a method for scoring model of Asian women's face value based on a deep convolution network, which includes the following four steps: sample data collection step; step of building a convolution network model; step of building a random forest model: based on a face recognition library The technology marks the position of the key points of the face, extracts the position coordinates of the key points of the face, traverses the cycle to generate a group calculation point composed of some key points of the face, constructs and optimizes the random forest model; based on the booster trees algorithm, the convolutional network model and the random forest are fused model to complete the fusion model stage of the final appearance score. It can simultaneously integrate Asian women's unique face shape, facial features, skin color and other key point information with strong distinguishing ability, and integrates the respective advantages of convolutional network and random forest model to improve the accuracy of the prediction of facial value score. The effect of the model is remarkable, and the average are closer to the real value.

Description

technical field [0001] The present invention relates to a deep convolutional network, in particular to a scoring model method for the appearance of Asian women based on a deep convolutional network. Background technique [0002] With the rapid development of social economy and the progress of the times, people's living standards are constantly improving, and more women have higher requirements for their own appearance. As a result, various appearance evaluation software has emerged as the times require. Many women hope to objectively and quantitatively evaluate the beauty of their appearance through certain technical means. However, there are two main types of existing appearance scoring techniques: appearance scoring based on public voting and appearance scoring based on machine learning algorithms. Taking the mean or mode, etc., has the interference of various artificial factors, and the accuracy fluctuates greatly; the latter mainly uses machine learning or deep learning...

AI Technical Summary

Problems solved by technology

[0003] In order to solve the current situation of the existing Asian women's appearance evaluation, the appearance evaluation effect is not good enough, only face features are extracted, there are much fewer features than image classification, and the evaluation accuracy is not ideal, etc. The present invention provides a method that can simultaneously integrate Asian women Unique face shape, facial features ratio, skin color, etc. have a strong ability to distinguish Asian women's facial value scoring model method based on deep convolutional network

Images