Seongheon Kim  Minsu Chae  Hwamin Lee

Vol. 1,  No. 1, pp. 1-12, Aug.  2024
10.23246/AAIRJ.2024.01.01.02

Pill Classification YOLO Deep Learning Image recognition Pharmaceutical Safety

PDF

This study addresses the critical issue of medication misuse through the development of an automated pill classification system using deep learning. To prevent adverse drug reactions caused by pill misuse, we developed a deep learning model for accurate pill classification using YOLOv8. We trained the model on a diverse dataset from AI-Hub, consisting of 69 pill types, with preprocessing techniques like image rotation and color conversion. The model achieved high accuracy scores on training (0.977), validation (0.974), and test sets (0.973), demonstrating its ability to generalize effectively. Minor performance discrepancies between datasets were observed, potentially indicating overfitting or the need for further model refinement. This research contributes a robust automated solution for pill identification, with implications for improving pharmaceutical safety and reducing medication errors. Future work will focus on enhancing the model's accuracy and exploring real-world applications to optimize pill classification.