A method for automatically classifying complexion in traditional Chinese medicine by applying a shallow neural network belongs to the field of computer vision. The designed shallow network has 5 layers in total, using three different layer structures, which are input layer, feature extraction layer, and output layer. The input layer consists of a convolutional layer and a rectified linear unit; the feature extraction layer consists of a 3-layer network, each of the first two layers consists of a convolutional layer and a ReLU activation function, between the convolutional layer and the ReLU A batch normalization, and add a pooling layer after the second ReLU of the feature extraction layer, the third layer of the feature extraction layer is a fully connected layer, followed by a corrected linear unit ReLU; the output layer consists of a fully connected layer , followed by a softmax classifier. The invention has obvious advantages in classification accuracy, is invariant to distortions such as zooming, translation, and rotation, has strong robustness, can effectively improve classification accuracy, and applies the theory of deep learning to the objectivity research of face-to-face diagnosis in traditional Chinese medicine. .