Image Classification Based on Convolutional Neural Network and Support Vector Machine

Automatic classification of images is a key task in many areas, including information retrieval, scene detection, internet data filtering, medical applications, etc. When directly operating on the image, the traditional classification method is difficult to achieve good results due to the high-dimensional characteristics of the data. To this end, in this paper, a novel method of image classification named CNN-SVM is proposed. Specifically, we first preprocess original images to extract the vital features and reduce redundant features by using the convolutional neural network (CNN). Then, the cross-validation (CV) is adopted to optimize the penalty parameter $C$ and the kernel parameter σ for the support vector machine (SVM). Finally, the extracted features will be input to an optimized SVM model. In order to validate the superiority of our proposed algorithm, we select a hybrid image set taken from Caltech 265 image archive as our experimental data. The experimental results reflect that CNN-SVM has higher classification accuracy.