Design And Implementation Of A Chinese Painting Classification System

Chinese painting,as a unique form of traditional painting art in China,plays an important role in the art filed in world.In addition to paying attention to the aesthetic feeling of the pictures,viewers should also learn about the related information of the paintings,including how to classify the paintings from a professional perspective.However,the classification of Chinese painting requires professional knowledge and literacy,it is difficult to classify Chinese painting by ordinary viewers without knowledge of Chinese painting.This thesis designs and implements a Chinese painting classification system to help users automatically classifying the Chinese paintings they have taken.This thesis describes the design and implementation process of the Chinese painting classification system in detail,including the investigation of existing related software and Chinese painting classification methods,and the introduction of related technologies.The requirements analysis,design and implementation of the system are described next,and finally the system is tested.The main work of this thesis includes:(1)A Chinese painting data set labeled according to the two classification criteria of theme and painting technique was established,and the deep learning model based on VGG-16 was trained based on the data set.In order to solve the problem of small training sample size,the transfer learning method is used to train the model.The experimental results show that for the classification task of Chinese painting in this thesis,transfer learning can effectively improve the classification effect of the model.(2)Designed and implemented a Chinese painting classification system based on C / S architecture.The client of the system is implemented based on the iOS operating system,and is mainly responsible for UI interaction,image acquisition,and display classification and matching results.In the design and implementation of the client,the work of the UI interface and other tasks,such as image quality detection,image processing and network communication work,are separated by GCD technology,so that the interactive interface can work smoothly.In order to improve the accuracy of image classification,the image checking module is designed and implemented to detect the image quality and the result is fed back to the user.The down-sampling calculation method is used to solve the problem that the full-image brightness detection in a whole image costs too much time.(3)The main functions of the server side of the system include: receiving client requests,classifying images,matching images in the system database,and returning classification and matching results.In the design and implementation of the server,a three-layer architecture scheme is proposed.The scheme uses the message proxy module based on Redis database as the middle layer.It solves the problem that the model is loaded multiple times when web server directly calls the deep learning framework caused by the concurrent request status.(4)The implemented system is tested,and the test shows that the system achieves the design requirements.The system’s design and implementation in this thesis can not only be used to assist users in the classification of Chinese painting,but also offer some reference for other image classification engineering practices based on deep learning.