Image Recognition of Traditional Chinese Herbal Medicine Based on Deep Learning

As a treasure of Chinese culture, Chinese herbal medicine (CHM) carries rich historical and cultural connotations and plays an important role in the field of traditional Chinese medicine. Conventional identification methods consume more financial and material resources. With the development of science and technology, deep learning algorithms are widely used in CHM identification, which greatly improves the identification efficiency. Therefore, this paper presented an application study to use convolutional neural networks to recognize CHM. The tasks are divided into three categories: image classification, object detection and image generation. Firstly, this paper trained model of AlexNet, ResNet and Inception V3. The accuracy rates of 91.80%, 98.88%, and 96.20% are obtained for the three models respectively, and the data show that ResNet is the most accurate in prediction. Secondly, this study used YOLOv8 model to accurate classify and locate different kinds of herbs in the images of CHM on a real time basis, and obtains the highest accuracy of 75.72%. At last, DCGAN and ACGAN were built and trained on the same dataset for comparison to generate more realistic pictures of herbs, which is conductive to solve the problems of current datasets of CHM, which is small in number, and the existence of noise and impurity interference. Besides, improving the diversity of the dataset of CHM is the most cost-saving way. In this paper, we achieved a good result in three tasks, which showed that exploring the way of herb identification based on deep learning technology has achievable significance. At the same time, it helps to promote the further development of the digitization of Chinese medicine.