Research And Application Of Children's Graffiti Automatic Scoring

Graffiti is the beginning of children's self-expression.Graffiti behavior is a product of visual experience,body and finger muscle movement coordination,which can reflect the physical and mental state of children at that time.Children's graffiti products are also a hot issue in the design of entertainment products.However,the traditional children's graffiti products can only provide interactive rendering function,but can not analyze and evaluate the quality of children graffiti,and lack of interaction.In order to solve this problem,based on the theory and technology of deep learning,this paper proposes to evaluate the quality of graffiti automatically by image convolution theory.The main research work is as follows:(1)The convolution network structure of graffiti image is improved and the accuracy of graffiti quality score is improved.A hybrid pooling function is defined for children graffiti images to extract color features.Firstly,based on Inception-BN network,the mean pooling function is used to extract the global color distribution of graffiti image.Then the main color information of graffiti image is obtained by using the maximum pooling function.Finally,the network is optimized by Fine-Tuning method,and the optimal scoring model is obtained.In the experimental part,a dataset containing more than 30,000 graffiti images is created,and the performance of the model is analyzed.The effectiveness of the network structure is verified by single category and multi-class graffiti scoring experiments.In addition,the subjective score and the model scoring effect are analyzed and compared,and the accuracy of the algorithm is verified.(2)Based on the research of automatic graffiti image scoring,the requirement analysis of children interactive graffiti system is carried out,and the system framework is designed.According to the business process of the system requirements,the functional design structure,the hardware topology and the corresponding user use case diagram of the children interactive graffiti system are given.It provides support and basis for the development and implementation of the system.(3)On the basis of the above graffiti scoring research and system requirement analysis,the children interactive graffiti system is designed and implemented.Using Unity platform as the development tool and Mxnet as the deep learning computing framework,the children's interactive graffiti system is designed and implemented.The main functions of the system include graffiti module,communication module,scoring module,graffiti mapping module,animation module and user interaction drawing module.The system realizes graffiti drawing function through MobilePaint plug-in,solves the problem of image work transmission through Socket network communication,realizes the deployment of scoring module through dynamic link library technology,and updates the map file through XML file.The interaction of flick diagram is realized by collision detection technology.The children's interactive graffiti system has been officially launched,and the product runs stably in practical application scenarios.At present,the products have been installed and deployed in China,the United States,Turkey and other countries and regions,which is very popular with children.