Facial aesthetics evaluation method based on visual facial form judgment and storage medium

Abstract

The invention discloses a facial aesthetics evaluation method based on visual facial form judgment and a storage medium. The facial aesthetics evaluation method comprises the steps of S1, performing mark point positioning on a facial image; s2, performing interpolation modeling on the facial contour mark points through a B spline function; s3, calculating facial contour curvature features according to the interpolation curve; s4, extracting geometric features according to a facial expression research theory; s5, extracting texture features through a Gabor filter; s6, extracting features such as skin color and hair through a convolutional neural network VGG; and S7, fusing the four features, and evaluating facial aesthetics by using a support vector machine (SVR). According to the method, mixed features are extracted in combination with B splines, Gabor filtering and a deep learning algorithm to describe facial aesthetics. And the final features cover the whole face information.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to a method for visually judging facial shapes and a method for evaluating facial aesthetics based on the method. Background technique [0002] At present, the research on facial aesthetics evaluation at home and abroad is not comprehensive. Facial aesthetics assessment methods are mainly divided into traditional methods and deep learning methods. Traditional methods are mainly based on manual features. The deep learning method is mainly to build an end-to-end CNN model. [0003] Manual features mainly include local features such as face geometry, texture, and color. Among these features, geometric features are one of the most commonly used features in the study of face aesthetics. They are derived from the positions of the marker points, the distances between the positions, and the ratio of the distances. Therefore, accurately obtaining the landmarks of facial contour...

AI Technical Summary

Problems solved by technology

However, a deep learning model is an end-to-end model that lacks aesthetic interpretation

[0005] Therefore, we combine the traditional method with the deep learning method, which not only overcomes the problem of low robustness of the traditional method, but also solves the disadvantage of the difference in the interpretation of the deep learning model.

Images