Machine Learning Based Vision System For Tablet Elegance
Webinar Overview
Functional or cosmetic coatings are applied in the pharmaceutical industry to aid performance or appearance of tableted products and the application of machine vision for automated inspection has been gaining attention with recent advances in computing power and high-resolution sensors.
Traditional manual inspection is both time and resource intensive for large batches. Defects on tablet coating are not only an indication of product quality, but also of paramount importance to process understanding. Machine vision utilizing deep learning techniques can provide an automatic and fast tool to facilitate both inspection and diversion from the good product stream based on defined visual CQA’s (Critical Quality Attributes). The goal of the vision system currently under development is to become an alternative option for labor intensive visual AQL (Acceptable Quality Limit) testing for tablet elegance and, by examining a larger proportion of the batch, extract more information from the types of defects and their frequency of occurrence.
Two approaches to machine vision of tablet inspection will be compared. The first is supervised training that requires labeling defects in training datasets/images and can generate defect specific information, which can show advantages in commercial manufacturing with large batch and relatively fixed formulas. The second is unsupervised training that uses only defect-free tablets, leading to an efficient method development. The advantages and considerations of each approach will be discussed along with a case study to demonstrate the results from both scenarios. Finally, the presentation will discuss the workflow, modeling approaches of the vision system, and its benefits for continuous manufacturing.
Optimising tablet release with hybrid fusion testing.
Some companies are now adopting at-line or off-line non-destructive testing for tablets, usually using a single spectroscopic method. But what might be possible through combining different non-destructive testing techniques, and how could this influence the analytical testing of Oral Solid Dose products in laboratories?
This webinar was originally hosted by the Community for Analytical Measurement Science (CAMS) as part of FutureLab: Chemistry in the Digital Age Online Webinar Week. To learn more about CAMS, visit CAMS-UK.
Wednesday, October 15th, 2025
11:00 EDT | 16:00 GMT
Speakers
Yong Mei
Senior Analytics Scientist at Pfizer
Yong Mei is a Senior Analytics Scientist with seven years of experience applying data science to advance pharmaceutical R&D. A specialist in manufacturing data analysis, He has worked on supporting aspects of continuous manufacturing and pioneering machine learning based visual inspection methods.