Herbal Medicine Image Recognition Algorithms Based on Convolutional Neural Networks

Herbal medicine is the basis of Chinese medicine treatment, and the correct identification of herbal medicine is the most fundamental issue in Chinese medicine. Traditional Chinese herbal medicine recognition relies on personal experience and has a low accuracy rate, which seriously affects the therapeutic effect of Chinese medicine. In this paper, we propose a convolutional neural network-based herbal medicine image recognition method, which aims to improve the accuracy of herbal medicine recognition and enhance the therapeutic efficacy of Chinese medicine treatment. Specifically, a dataset of 10 different types of Chinese herbal medicines is first collected using methods such as web crawlers. In view of the small amount of data, this paper expands the dataset with samples. Multiple convolutional neural network models are compared on these expanded datasets in terms of accuracy, precision, recall, and other aspects. The comparison results show that the DenseNet121 model performs best in these metrics.