Classification of tongue color based on CNN

Tongue manifestation is one of the most significant basic criteria for the diagnosis of Traditional Chinese Medicine (TCM). And tongue color recognition with high accuracy will contribute to the efficiency of TCM diagnosis. The drawbacks of traditional tongue diagnosis methods are that the features need to be designed artificially. While the feature acquisition from the deep learning is a process of simulating the brain activities and learning behaviors of human beings, and it has achieved fruitful results in many aspects, including image classification, face recognition, objects detection and so on. Therefore, the method combining deep learning with tongue color classification is proposed. First, the preprocessed and enhanced images are created as a tongue image database. Then, the parameters of the traditional network are modified for tongue color classification. Finally, it is more targeted to use our own model to fine-tune our neural networks. The experimental results show that this method is more practical and accurate than the traditional one.