Medical-Pills Detection Using YOLOv11: A Proof of Concept Study for Pharmaceutical Automation

The pharmaceutical industry is tasked with ensuring the production and distribution of medications that meet stringent quality and safety standards, yet it grapples with significant challenges in automating critical processes such as quality control, pill sorting, and inventory management. These challenges arise from the inherent complexity of identifying medical pills, which vary widely in shape, size, color, and imprints, often requiring meticulous human intervention that is both time-consuming and error-prone. This study introduces a proof-of-concept (POC) dataset and a cutting-edge YOLOv11-based computer vision model tailored for medical-pills detection, with the overarching goal of advancing automation within pharmaceutical workflows. The dataset comprises 115 meticulously labeled images, split into 92 training and 23 validation samples, and serves as the foundation for training our YOLOv11 model, which achieves an exceptional mean Average Precision (mAP@ 0.5) of 0.995. We rigorously evaluate the model’s performance using a suite of analytical tools, including precision-recall curves, F1-confidence curves, confusion matrices, and bounding box visualizations, providing a comprehensive assessment of its capabilities. The results underscore the transformative potential of AI-driven solutions in pharmaceutical applications, such as automated sorting, defect detection, counterfeit identification, and real-time inventory tracking. However, we also acknowledge limitations, such as the dataset’s modest size and the controlled conditions of our experiments, which temper the generalizability of our findings. This work establishes a foundational resource for researchers and industry practitioners, offering both a dataset and a high-performing model to catalyze the development of scalable, efficient systems for healthcare automation, while transparently outlining areas for future improvement.