Cartoon Characters Based On Faster R-CNN Identification Research

As a unique stage art form in China,opera has a history of hundreds of years.Due to the development of cultural diversity,traditional opera has gradually declined.In recent years,the produced opera animation is easy to be accepted by people,so that the cartoon characters of the opera can be used to attract people to like and understand the opera.Therefore,the combination of drama cartoon characters and target recognition technology to classify and identify targets can help people quickly recognize each drama cartoon character and cultivate more audiences and actors for this art category,which is of great significance for the inheritance and protection of drama.At present,the research on target classification and recognition has achieved gratifying results in many aspects,but there are few people involved in the detection and recognition of drama cartoon characters.In this paper,the Faster R-CNN algorithm based on feature extraction network Res Net50 is used Detection and recognition,and then merge Feature Pyramid Networks and improve the network,achieved a good recognition effect.The main research contents are as follows:(1)Production of data setThe data set used in the experiment in this paper is based on searching for related pictures on the Internet and shooting and collecting opera dolls.Because the number of pictures collected is limited,data enhancement technology is used to expand the number of pictures;then each picture is manually marked.Produced the first complete opera character data set.(2)Based on Faster R-CNN algorithm,the character recognition of drama cartoons fused with Feature Pyramid NetworksIn this paper,VGG16 based on feature extraction network is used,and the original Faster R-CNN algorithm is improved: replace VGG16 with Res Net50 network,and then integrate the feature pyramid network to improve it.The comparative analysis of the character recognition results of various categories shows that the improved model after fusion of Feature Pyramid Networks has a greater improvement in recognition accuracy.(3)Style transfer technology,rich functionsAfter the Faster R-CNN algorithm merges the Feature PyramidNetworks to identify the cartoon characters of the opera,combined with the style transfer technology,the cartoon color of the cartoon characters of the opera is rendered,which adds the easy-to-accept Chinese style ink painting.Combined with Django,the system prototype for realizing the cartoon character recognition in the opera.The innovation of the thesis: the first opera cartoon character data set was produced,which provided convenience for the follow-up study of opera character recognition and application.Using the classical algorithm models Faster R-CNN and Res Net as the carrier,the Feature Pyramid Networks is fused,and the network model is improved and optimized,and good results have been achieved in the recognition of drama cartoon characters.