Analysis Of Japanese Animation Diet Culture Elements Based On Convolution Neural Network

With the proposal of the concept of interdisciplinary integration,the relevant methods of using advanced science and technology to solve the problems existing in cultural research gradually began to appear.As one of the most favorite leisure and entertainment methods of contemporary youth,animation has good media attributes.The use of animation for cultural communication is not only clever,hidden,but also has strong penetration.As a big food country,how to avoid the invasion of foreign culture and carry out the inheritance and dissemination of local culture has become one of the hot research topics.Traditional research methods mainly rely on the relevant research of cultural scholars.In the research process,researchers need to watch a large number of animation videos,collect and classify animation images,and finally conduct cultural analysis.This way not only costs researchers a lot of time and labor costs,but also the collected and classified images are subjective and uncontrollable,which ultimately reduces the efficiency of processing and analysis.Convolutional neural network is a research hotspot in the field of machine learning.It has a wide application prospect and great economic value in image retrieval,image processing,image classification and so on.Therefore,the application of convolutional neural network model to diet image classification in animation will provide convenience for relevant scholars.At present,the recognition and classification of real food images based on convolutional neural network has been carried out steadily and achieved good results.However,because the food images presented in animation depend on the subjective creativity of comic writers,have non objectivity and high degree of abstraction,the effect of applying the traditional convolutional neural network model to the classification of animation food images is not very good.Therefore,this paper studies the Densenet network model and proposes the corresponding efficient classification algorithm.The main innovations of this paper are as follows:1.In view of the fact that there is no daily diffuse diet data set specially used for image classification at present.This paper constructs a daily diffuse diet image data set Ja food-1100,which contains 11 common categories;In order to improve the robustness of the model,this paper proposes a diffuse image dataset Ja photo-1200,which contains12 common categories of non dietary foods.2.In order to improve the accuracy of the model in the classification of Japanese anime food image data set,after comparing several classical convolution neural network models in detail,this paper proposes an adensenet5 network model based on Densenet network.The idea of the algorithm is:(1)the convolution kernel size of 5 is introduced into the dense layer of Densenet,expand the receptive field of the algorithm and enhance the depth of the network model;(2)After the dense layer of Densenet,the convolution attention mechanism module is introduced to optimize the feature representation of the image,select and pay attention to the more important image features,and further extract the deeper information in the image.The experimental results show that compared with other classical convolutional neural network models,adensenet5 network model improves the classification accuracy in the two data sets proposed in this paper,and has achieved better results in the test data set proposed in this paper.