Research And Implementation Of Intelligent Photo Album System Based On Image Semantics

As the number of images increases rapidely,the production of photo albums has become more and more challenging.When there is no reasonable management process,making albums is difficult,and the quality of albums is even more difficult to guarantee.Therefore,how to build an album system to help people quickly obtain high-quality photo albums is of great significance.This article studies and explores this problem,the main results are as follows.1.Use image classification techniques to identify user image themes.Firstly,this paper studies related image classification techniques,and then combines the advantages of SENet(squeeze-and-excitation networks)and DenseNet(densely connected convolutional network)to construct the SE-DenseNet network.The comparison of this network with other classification networks proves that the SE-DenseNet network has obvious advantages under the classification task.Then this paper proposes a new scene selection model based on the SE-DenseNet network.This model uses the SE-DenseNet network,database design and matching rules of the album system to select an appropriate album background pattern for the image.Through experimental research,it is shown that the scene selection model based on SE-DenseNet proposed in this paper can help the album system to intelligently and concisely generate an electronic album which conforms to the theme of images.2.The use of super-resolution restoration technology to improve image clarity.First of all,this paper studies the super-resolution restoration technology,and then proposes an album restoration model based on SRGAN(a generative adversarial network for image super-resolution).The model uses the SRGAN network and the database design of the album system to jointly select the images that do not meet the clarity requirements in the electronic album,and then restore this part of the images.Experimental research shows that the album restoration model based on SRGAN proposed in this paper can effectively improve the clarity of the album.3.Use the album system to help people obtain high-quality photo albums intelligently,concisely,and efficiently.First of all,this paper proposes a smart album system based on image semantics in combination with the scene selection model and super-resolution restoration model,and through the analysis of user roles and needs,designs and implements a album system which has a consumer terminal system,a manager terminal system,and a large agent terminal system.At present,the intelligent photo album system has completed the realization and testing of all functions,and has been put into use.