Research on Image Classification of Chinese Medicine Based on Improved Deep Learning Technology

Abstract

Automatic classification of traditional Chinese medicine (TCM) images using deep learning has important application value as widely used in medical, teaching, and other fields. However, there is a lack of data collection of TCM images in domestic databases, and there are only a few studies on the classification of TCM images using deep learning techniques. To solve this problem, we collected 1000 TCM images of mango and almond leaves, both of which are difficult to distinguish. Comparison of this dataset with improved residual networks and deep neural networks such as CNN, VGG16, and VGG19 shows that the best results and the classification accuracies were over 95%. The deep neural network is ideal in two complex TCM image backgrounds and has some future application prospects.