Chinese Calligraphy Style Recognition Based On Deep Learning

Chinese traditional culture has a long history,unique charm and profound mass foundation.Opera,calligraphy,painting and so on are all important carriers to express and inherit Chinese excellent traditional culture.The digitalization of costume patterns and the generation of new styles are of great significance.However,the existing forms of digitalization technology are monotonous,and the artificial design is difficult and inefficient.Based on convolution neural network,this paper integrates traditional culture and digital technology,enriches the forms of traditional culture with the method of style transfer,and innovates the styles of traditional costume.Most of the existing style transfer algorithms are based on the style of Western oil paintings.The over abstraction of texture is not suitable for the expression of Chinese cultural elements such as opera,and the transferable style features are single,and the jitter produced by the frame by frame generation of video artistic output affects the output effect.In this paper,the related network structure and loss function are improved,and the main work is as follows:(1)This paper proposes to adjust the content loss function and style loss function to get stylized images of different degrees,and optimizes the selection of convolution network framework,the calculation method of content and style loss function and the selection of corresponding feature layer.(2)For the first time,the style transfer is combined with the traditional clothing image.In order to better show its artistic effect,a new loss function is added on the basis of the original loss function.Compared with the experimental results of PRISMA,the image processed in this paper is smoother,the over abstraction phenomenon of line twist is solved,and the subjective appearance is more beautiful.traditional culture.Changing the weight coefficient of different styles can make the final migration style tend to a specific style.(4)Aiming at the flicker and jitter problem of video stylization,the time consistency of each pixel is realized.The transformation output of stylized network is given to calculate the time sequence loss,which is added to the total loss function to improve the effect of video style migration.In a word,this paper integrates style transfer with traditional cultural elements for the first time,improves the generation effect,solves the jitter problem in video,designs,realizes and finally applies it.