Research On The Facial Attractiveness Of Chinese Images With Epoch Characteristics Based On Data-Driven

Facial attractiveness is a hot topic in the fields of cognitive psychology,social psychology,computer science and cosmetology in recent years.It has an important impact on people's social life.Although there are many methods for the study of facial attractiveness,scholars have mainly established facial attractiveness analysis systems based on public databases,and many studies have focused on women.This paper mainly studies the changes of facial attractive geometric features of the Chinese male and female with epochal characteristics and the composition method of average face.The main works as follows:(1)Establish a Chinese face database with the epochal characteristics.This paper download face images of Chinese male and female stars born in different times from some popular social networking sites,and then invite raters of different ages to rate attractiveness for each image,and finally find that there are different aesthetic standards in different times by analyzing attractiveness scores.(2)Analyze the changes of facial features of Chinese male and female.First,this paper studies the accuracy of KNN,LDA,SVM,MLP,and retraining Inception v3 for face classification.Experimental results show that the classification accuracy of the retrained Inception v3 model is the highest,reaching 96.5%.Finally,the face images are classified according to this model,and the results show that the heart face which conforms to the “golden ratio” is popular feature of modern Chinese male and female.When analyzing the eyebrows,this paper proposes to calculate the area,length,and average width of the eyebrows based on the contours of the eyebrows.The experimental results show that the attractive eyebrows change to thin and long with the development of the times.The results of this research can be used to explore the relationship between characteristic changes of facial attractiveness and social changes.(3)A composition method of attractive face is proposed based on the average face hypothesis.The method locates 81 feature points on the face based on the AAM algorithm.The Delaunay triangulation algorithm is used to segment a face image into several triangular regions.The Generalized Procrustes Analysis method is used to achieve face shape alignment.The local affine transformation is performed on the corresponding areas of the input image and the average face image according to the triangulation,and the average of all the original facial pixels is calculated to obtain an average face.The experimental results show that the proposed method is more feasible and the composite average face looks more beautiful and natural,which verifies the conclusion that “the average face is beautiful”.This paper uses the composite average face from the highly attractive faces born in the 1990 s as the most attractive and predictive face in the future,and this average face can be used to assist the plastic surgery.