TCM Tongue Segmentation and Analysis with a Mobile Device

Mobile devices and applications provide increased accessibility to things like Traditional Chinese Medicine (TCM). Ideally, one would be able to seek out a diagnosis to the tongue using advanced machine learning models in a mobile device. However, such kind of models are still not yet available for the public. In this research, we investigated a way of combining machine learning with a mobile application to support tongue diagnosis for the public. For tongue segmentation, we experimented with an unsupervised approach (K-means), a supervised method and a deep learning model. There are two outcomes presented in this research. Firstly, the user can seamlessly take a picture from a mobile phone and send it over to the server for a segmentation process. Secondly, the image processed by the server would have at least some levels of autonomous analysis preformed on it with acceptable results using machine learning methods. Our current results indicate that we were able to finish some key steps in this process.