
Amid the rapid advancement of artificial intelligence, the pharmaceutical industry is undergoing a profound technological transformation. Long characterized by its complexity and high barriers to entry, every stage of the pharmaceutical value chain—from drug discovery and development to manufacturing, regulatory compliance, and supply chain management—presents significant challenges. In recent years, the rise of artificial intelligence (AI), machine learning, and deep learning has been reshaping this sector. AI technologies are poised to be game-changers, shortening R&D cycles and enhancing the efficiency of clinical trials, while also driving intelligent upgrades in manufacturing processes. Against this backdrop, visual inspection technology, a critical component in ensuring drug safety and quality, is unlocking new possibilities through the integration of AI.
The characteristics of drug forms such as suspensions and lyophilized formulations cause traditional visual inspection systems to misidentify harmless bubbles or minor cosmetic defects as contaminant particles, thereby increasing re-inspection rates and generating additional costs. To address this pain point, Stevanato Group has launched the SG Vision AI platform, specifically designed for the pharmaceutical industry. This platform combines extensive industry expertise with advanced deep learning models to provide customers with customized, high-quality inspection solutions while minimizing false rejection rates.
SG Vision AI can analyze every image of vials or cartridges, accurately distinguishing between air bubbles and harmful particles, thereby significantly reducing the false rejection rate. In a real-world project, the platform was trained using 30,000 samples, with 75% allocated for model training, 10% for the validation set, and 15% for the test set. Subsequent system validation with an additional 6,000 samples achieved an outstanding false rejection rate of only 0.21%, demonstrating exceptional detection performance and reliability. This achievement not only enhanced inspection capabilities on the production line but also generated substantial cost savings for pharmaceutical manufacturers.
The unique advantages of SG Vision AI lie not only in the precision of its deep learning models but also in the platform’s overall design and ease of application. Built on Microsoft Azure, the platform complies with U.S. CFR 21 Part 11 and EU GMP Annex 11 regulatory requirements, ensuring secure and controlled data during upload, sharing, and processing. The built-in annotation assistant facilitates rapid defect classification and ensures end-to-end data traceability, significantly shortening the workflow from initial annotation to final review. Furthermore, the platform provides intuitive statistical and visualization tools, enabling users to monitor and analyze the performance of each inspection stage in real time. Even with data stored in the cloud, operators can conveniently access and manage images, facilitating efficient collaboration.
In terms of cost and operational efficiency, SG Vision AI eliminates dependence on local storage space and high maintenance costs through cloud-based deployment, while maintaining data encryption and integrity. This enables pharmaceutical companies to reduce total cost of ownership (TCO) while ensuring production safety and compliance. The entire platform is designed with user experience at its core, offering a graphical interface and intuitive workflow to make the visual inspection process smoother and more user-friendly. Continuous expert support is provided throughout the deployment and operation phases to ensure ongoing optimization of model performance.
Looking ahead, as the pharmaceutical industry’s demand for personalized therapies and high-quality medicines continues to grow, the value of AI applications in visual inspection will become increasingly prominent. By deeply integrating advanced artificial intelligence technologies with pharmaceutical expertise, Stevanato Group is driving the pharmaceutical industry into a new era centered on data-driven, intelligent management.