Deep learning is the way to identify and classify pills, making a big impact on healthcare and pharmaceuticals. Patients will get the right medications more accurately by automating this process. These deep learning systems analyze unique pill features like shape, color, size, markings, and any imprinted text to match each pill to its correct medication identity. However, there are still challenges, such as limited datasets and pills that look very similar, which can affect the identification accuracy. In this study, these challenges were tackled by developing deep learning models specifically for pill image classification. A dataset of 994 images of 20 common pill types from a public pharmaceutical database was used and splitting them into training, validation, and test sets in a 70:15:15 ratio using Jupyter-Notebook. Three deep learning models which are YOLOv6, YOLOv8, and Real Time Detection Transformer (RTDETR) were tested and compared by measuring their performance metrics such as precision, mean Average Precision (mAP), and recall. The results were promising, especially for YOLOv8, which achieved a precision of 98.72%, outperforming models like RTDETR and YOLOv6. However, for recall and mAP50, RTDETR took the lead with 100% and 99.09%, respectively. This high level of accuracy suggests that YOLOv8 could significantly reduce medical errors by providing real-time, precise pill identification. However, since no pill should be missed, the high recall rate of RTDETR also plays a vital role. In environments where it is vital to detect every pill, such as in automated packaging and dispensing in pharmacies, RTDETR’s perfect recall score ensures that no relevant pill goes undetected. Integrating artificial intelligence into pill recognition helps minimize the risk of human error, prevent patients from taking the wrong medication and makes life easier for healthcare professionals by freeing them up to focus on more critical tasks and speeding up the often-tedious process of pill identification.