A Tongue Color Classification Method in TCM Based on Transfer Learning

In the field of TCM tongue diagnosis, the most important feature is the tongue color. Recently, the deep learning method can be used to classify the tongue color, which has been one of the important research directions for the objectification of tongue diagnosis. However, if there is a small amount of training data and low hardware equipment configuration, directly use of the current deep learning methods to classify the tongue color will not achieve satisfactory results. In order to solve these problems, this paper proposes a tongue color classification method in TCM based on transfer learning. This method transfers the model parameters trained on the ImageNet data set to the task of tongue image feature extraction and classification, splits of the tongue image data set is used for training and testing. We have used the obtained depth network to classify the color features of tongue image. Experiments illustrate that this method has good accuracy and short training time, and can better complete the task of tongue color classification.