Image Recognition of Chinese herbal pieces Based on Multi-task Learning Model

The automatic identification of traditional Chinese medicine decoction pieces is of great significance in the objectification of Chinese medicine quality control. Common methods are traditional features of artificial design and machine learning algorithms represented by convolutional networks, but both have certain limitations in certain situations. This paper refers to the concept of multi-task learning, based on neural network and supplemented by traditional features, so that the two can be merged, establish a new deep learning model, and collect 200 kinds of medicine tablets (total 30437 images). The model is in different experimental tasks. The accuracy rate is up to 86.2%, and both of them maintain a good generalization effect. The multi-task learning model can be used to identify the Chinese medicine decoction pieces, which can effectively improve the accuracy and stability of the decoction piece identification, and has good application value.