The quality inspection of drug manufacturing must be operated as prescribed with little to zero margins for errors; otherwise, consequences could be fatal. With the advancement of technology and digital devices, taking photos of unidentified drugs or pills to prevent erroneous phenomena has become significantly easier. This area of research falls under the broad category of information retrieval, specifically focusing on image detection and recognition. This paper proposes a detection system to assess pill quantity and quality based on the blister pack surface. The designed system is integrated with the SOTA deep learning network (YOLOv8) to autonomously perform the inspection process. We establish the dataset by capturing the blister pack surface images under different light conditions and applying image processing techniques for pre-processing steps to enrich the data. The augmented dataset is then used for training the designed deep-learning model. The experimental results demonstrated that the proposed model not only achieves high detection accuracy but can also be effectively implemented in real-time industrial classification tasks.