Traditional Chinese Medicine Recognition Based on Target Detection

Area of Science:

• Computer Vision
• Pharmacognosy

Background:

• Traditional Chinese Medicine (TCM) is extensively used in China.
• The wide variety of TCMs presents significant challenges for accurate visual identification in pharmacies.

Purpose of the Study:

• To evaluate the effectiveness of deep learning-based target detection models for identifying diverse TCMs.
• To compare the performance of mainstream object detection models in the context of TCM recognition.

Main Methods:

• A dataset of 100 common TCM images was collected from pharmacies.
• Three mainstream target detection models (Faster RCNN, SSD, YOLO v5) were trained on the TCM dataset.
• Model performance was evaluated using key metrics such as mean average precision and frames per second (FPS).

Main Results:

• The YOLO v5 model demonstrated superior performance in recognizing various TCMs.
• YOLO v5 achieved a mean average accuracy of 94.33% and an FPS of 75.
• This model also showed advantages in handling small targets and issues of occlusion, with fewer parameters.

Conclusions:

• Target detection technology, particularly YOLO v5, shows significant potential for automated TCM identification.
• This approach can address the visual identification difficulties in TCM settings.
• The findings support the broad application prospects of AI in TCM detection and quality control.