Intelligent Drug Recognition System Based on Improved YOLOv7 and PP-OCRv3

The manual dispensing of drugs by pharmacists is time-consuming, labor-intensive, and has a low accuracy rate. In order to address this issue, this paper proposes an automated drug dispensing method based on improved object detection and text recognition. This method first embeds a Simple Attention Module (SimAM) into the network structure of YOLOv7, named YOLOv7_Sim, to achieve the localization and recognition of medicine bottles. Additionally, a lightweight PP-OCRv3 is employed for text recognition on liquid bags, and the similarity between the recognition results and the names of drugs on the prescription is used to determine a match. The novelty of this method lies in the combination of YOLOv7 and SimAM, which improves the accuracy of medicine bottle recognition and enhances the algorithm's robustness. Experimental results demonstrate that compared to YOLOv7, YOLOv7_Sim shows significant performance improvements, with a 3.12% increase in mAP, 0.69% increase in precision, and 2.83% increase in recall for the same type of medicine bottle. The proportion of fully recognized text on liquid bags is 90.6%. The drug recognition system, which combines YOLOv7_Sim and lightweight PP-OCRv3, has practical engineering significance for automating the drug dispensing process.