Abstract
Tongue coating is a crucial diagnostic indicator in traditional Chinese medicine, intuitively reflecting the body's physiological and pathological conditions. However, traditional visual inspection methods are highly susceptible to subjective bias, often resulting in diagnostic deviations and inconsistencies. To address these limitations, this study proposes an intelligent tongue coating diagnostic model based on an enhanced YOLOv13. The model integrates a hybrid architecture of swin transformer and YOLOv13, effectively capturing global contextual and local textural features for fine-grained recognition and analysis of tongue coating characteristics. Experimental results show that the enhanced model substantially outperforms the original YOLOv13 in fine-grained feature extraction and boundary localization, establishing a reliable foundation for the objectification, standardization, and intelligent advancement of tongue diagnosis in traditional Chinese medicine.