Ceramic Image Recognition And Its Application Based On Convolutional Neural Network

With the diversified development of society,personalized customization has become the pursuit of fashion for young people,and ceramic products carrying thousands of years of Chinese culture have become the target of young people’s customization.In order to facilitate many young people to customize ceramic products,the author’s team launched a personalized customization platform for ceramic products.In order to increase the user stickability of the ceramic customization platform,this paper applies the image recognition technology to give users a memorable scene story on the customized ceramic products,so as to give users the best experience.Image recognition technology based on convolutional neural network has been widely used in other industries.In this paper,the convolutional neural network technology is applied to the recognition of ceramic decorative patterns,and combined with the personalized customization platform of ceramic products,a small program called "porcelain language story" is developed.In this paper,ceramic images in the Internet are collected by web crawler technology,and then unsuitable images are eliminated,and effective images are processed by image enhancement technology,which makes the image samples more diverse and solves the problems of imbalance of image categories and insufficient samples.With the development of the research on convolutional neural network,a variety of convolutional neural network models have been produced.Each model has its own design considerations,and the model structure is also different.In this paper,Alexnet model,VGGnet model and Googlenet model were selected for experimental research,and the image recognition accuracy based on the three models was compared.In combination with generalization,model parameters,training time and other factors,VGG16 of the VGGnet model with the best performance was selected for the subsequent image location multi-task learning.In order to further improve the accuracy and efficiency of image recognition,this paper also tries to reduce the retrieval range by ceramic type classification,and to remove background interference information by image positioning and cutting method,which has achieved satisfactory results.In the end,We Chat applet is used as the application development platform,and the above technologies are used to identify the images uploaded by mobile devices and match the corresponding ceramic products,so as to obtain the story content of the product and display it.