An Approach for Generating Microscopic Images of Scarce Traditional Chinese Medicine Cells

Artificial microscopic identification remains the predominant method in traditional Chinese medicine (TCM) authentication. This technique, which depends heavily on manual expertise, is time-consuming and lacks consistency, thereby hindering the high-quality development of the TCM industry. Fortunately, deep learning has made significant advancements in the field of image recognition. However, unlike conventional image recognition tasks, the acquisition of microscopic images for Traditional Chinese Medicine (TCM) identification is costly, and issues such as image scarcity and imbalanced dataset distribution pose challenges to the direct application of deep learning techniques in TCM microscopic identification. This article starts with expanding the image scale of scarce traditional Chinese medicine cells and proposes an approach for generating microscopic images of scarce traditional Chinese medicine cells. This approach generates microscopic images in a condition guided manner, then optimizes the discriminator through Wasserstein distance to enhance the model fitting ability, and finally uses Differentiable Augmentation Strategy enhancement to expand the image generation range, achieving the expansion and enrichment of the diversity of scarce traditional Chinese medicine cell microscopic images. The experimental results of YOLOv8 model training show that our CGNGAN+DA approach has improved precision by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{2 3. 0 4 \%}$</tex> compared to the CGAN+DA approach.