A two-stage deep transfer learning traditional Chinese medicine tongue diagnosis model belongs to the technical field of traditional Chinese medicine auxiliary diagnosis and treatment. The method comprises the following steps: firstly, constructing a deep network based on a deep convolutional feature paradigm, fusing multi-scale features by using a pyramid strategy, and constructing deep abstractrepresentation of an input tongue image; then, designing two-stage deep transfer learning and obtaining the recognition capability of representative focus features in tongue image diagnosis in a targeted mode, thereby effectively solving the problem of data shortage, and reducing the training cost. On the basis, a focus examination cost function is designed, a deep migration model is trained, detection is carried out from different scales, an abnormal tongue image focus is marked, and the detection precision is improved. Finally, according to the examination result of the deep migration model,the process of combined use of multiple diagnostic methods of traditional Chinese medicine diagnosis and treatment is simulated, and real-time discrimination on abnormal tongue images is carried out,thereby improving the accuracy of diagnosis. The model designed by the invention can simulate the traditional Chinese medicine diagnosis theory, diagnose abnormal tongue images in real time, and provide clinical assistance and diagnosis and treatment suggestions for traditional Chinese medicine.