Chinese traditional visual cultural symbols recognition based on the combination of depth hierarchy features and SVM

Abstract

Chinese traditional visual culture symbols (CT-VCSs) is formed in the tradition and has the characteristic of Chinese unique ideological and cultural connotation. It is a visual cultural heritage of Chinese culture. So the research on CT-VCSs has important practical significance. In this paper, it is mainly about the recognition and classification of CT-VCSs based on machine learning. We make use of the advantages of both shallow learning and deep learning, and then we combine them together to achieve the purpose of recognition. In other words, we regard the traditional convolutional neural network as a feature extractor and extract the features of two full-connected layers as depth hierarchy features. At last, we take them into SVM with Histogram Intersection Kernel function for classification. Experimental results show that the recognition effect is very good.