ResSepSE-TransNet: An Efficient Deep Learning Model for the Classification of Common Chinese Medicinal Herbs

Chinese medicinal herbs are a vital part of traditional medicine, and accurate classification is essential for advancing TCM research and promotion. However, the large variety and complex morphology of herbs pose challenges to efficient classification. This study focuses on 12 common herbs and proposes ResSepSE-TransNet, an improved deep learning model based on ResNet50. It employs Depthwise Separable Convolutions to reduce parameters and computation, integrates a Squeeze-and-Excitation (SE) module to enhance feature representation, and utilizes transfer learning to improve feature extraction. Experiments on a self-built herb image dataset show that ResSepSE-TransNet achieves <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{9 8. 9 6 \%}$</tex> accuracy on the test set, outperforming ResNet50 and other baselines. The model also excels in parameter efficiency and inference speed, suitable for resource-limited scenarios. These results demonstrate that ResSepSE-TransNet offers an effective and practical solution for intelligent classification of Chinese medicinal herbs.