Scene Image Classification Based On Deep Learning

Scene classification is a hot topic technology in the field of computer vision,machine learning and pattern recognition.In recent years,with the rapid development of image sensor and internet,the way of image acquisition and transmission have been remarkably expanded.For example,the quantity of images stored by the popular image sharing websites,Flickr,exceeds six billion,and the number of active users of Instagram has break through 100 million.Facing massive image data,the traditional manual classification method has been unable to meet the increasing demand.Therefore,it is necessary to investigate the scene classification methods to automatically annotate the category of images to improve image retrieval efficiency and expand the visual applications.The main contents of this thesis are as follows:1.For the problems of traditional scene classification methods need artificial design features,this thesis compares the three different deep network models,including AlexNet(8 layers),VGG16(16 layers),ResNet(152 layers).Using the Places2 data sets for training and testing,the AlexNet network model has a faster convergence speed on the premise of ensuring high classification accuracy.2.On the basis of the above work,this thesis proposes an IK-AlexNet network model based on global-feature,the main idea is to improve the size and quantity of the convolution kernel of the AlexNet network model.Through the Places2 data sets,the IK-AlexNet model has obtained better scene classification results.3.Because LSTM network has the characteristics of understanding the localfeature in the scene,this thesis proposes an IK-AlexNet-LSTM network model merging the global and local features,in order to further improve the accuracy of scene classification.Using the Places2 and LSUN datasets for training and testing,the accuracy of the scene classification is further improved.