Tongue Color Classification Based on Convolutional Neural Network

Abstract

Tongue color classification plays an important role in traditional Chinese medicine. Tongue color is closely related to the physical condition of patients, so it can help doctors to diagnose patients accurately. However, it is difficult to distinguish between different tongue colors. Therefore, it is necessary to develop an effective method to extract high-dimensional tongue color features. Based on deep learning, this paper proposes a new method to improve the accuracy of tongue color classification. Firstly, the semantic convolutional neural network (CNN) is used to extract the tongue image from the background. Then the CNN model is used to extract the tongue color features, and the center loss function is used to enhance the feature discrimination during the training. Experimental results of different verification indexes show that the accuracy of tongue color classification can be improved by the semantic based CNN and center loss function.