ARTIFICIAL INTELLIGENCE (AI) TOOLS FOR THE REAL – TIME DETECTION OF TABLET DEFECTS

Senthil Prabhu R.*, Umamaheswari D., Asmita B. S., Abinaya P., Akash C., Annie Panies Stany A., Bavatharani E.

ABSTRACT

Tablet defects such as capping, lamination, cracking, sticking, and coating irregularities can significantly compromise drug quality, therapeutic efficacy, and patient safety. Conventional inspection methods are largely manual, subjective, and labor-intensive, making them unsuitable for high-speed pharmaceutical production environments. Recent advancements in artificial intelligence (AI), particularly deep learning–based computer vision systems, have enabled automated and highly accurate real-time detection of tablet defects. This review examines the application of Convolutional Neural Networks (CNNs) and the You Only Look Once (YOLOv5) object detection algorithm for tablet defect identification and classification. CNNs facilitate automated feature extraction and defect classification from tablet images, while YOLOv5 enables rapid localization and multi-class detection within a single-stage framework. These AI-based systems significantly enhance detection accuracy, reduce human error, and support continuous monitoring in pharmaceutical manufacturing. Integration of AI-driven inspection systems within Quality by Design (QbD) and Process Analytical Technology (PAT) frameworks promotes improved process control, consistent product quality, and Industry 4.0–based intelligent manufacturing. This review summarizes principles, architecture, applications, advantages, and limitations of CNN and YOLOv5 models, highlighting their transformative role in modern pharmaceutical quality assurance.

Keywords: Artificial Intelligence; Deep Learning; Convolutional Neural Networks; YOLOv5; Computer Vision; Pharmaceutical Manufacturing.