Hollow glass: the industry that demands zero defects.

The hollow glass markets, particularly pharmaceutical, food, and cosmetics sectors served by Tiama, impose exceptionally strict standards: the slightest micro-defect—particle, crack, imperfection, or opacity—can compromise the health safety of the contents or harm the product’s aesthetic appeal. Meanwhile, pressure to improve sustainability is intensifying, pushing producers to constantly optimize their processes.

Tiama meets new requirements

A specialist in hollow glass inspection, Tiama — a subsidiary of Kestrel Vision (a globally recognized expert in solid packaging inspection) — develops quality control solutions capable of automatically detecting a wide variety of defects in glass containers (bottles, jars, vials), while ensuring traceability and intelligent data exploitation. Its systems operate throughout the production chain, from the hot-end (molten glass) to the cold-end (solidified glass), with a dual goal: guaranteeing quality while maximizing industrial performance.

Client

Benefits

Using existing equipment, deep learning improves performance and inspection precision. Furthermore, the developed solution offers better tolerance to mechano-optical variations and guarantees improved sensitivity and processing repeatability across different machines. Finally, AI classification provides industry-specific qualified data (defect classes) across the entire image spectrum, unlike image processing systems based on thresholding.

Tiama quickly identified it: using artificial intelligence to automatically classify glass defects and improve production decisions is a strategic lever to better manage the glassmaking process, which is inherently inconsistent (mold wear, temperature variations, etc.) and, more generally, for quality control across the industry.
In this context, the company called upon Neovision as early as 2017 to support its vision.

Simulation of a range of glass bottles (visual generated by AI)

Development of an initial AI model validating the potential

This initial collaboration took the form of a project aimed at evaluating the contribution of Deep Learning to defect detection.

Creation of an advanced database

Neovision conducted a study validating AI’s potential to address these challenges, while highlighting an essential prerequisite: the need for a rich and reliable dataset. Tiama then decided to invest in creating its own base of annotated images, a strategic foundation for all its future AI developments.

Positive feedback from the first trials

Neovision took charge of developing and integrating AI models capable of automatically detecting and classifying a wide range of glass defects, while respecting inference time constraints.

Key figures

<100ms

Inspection time per bottle

+24

images traitées par bouteille et par minute

These AI models are embedded in the machines and tested at several pilot customer sites. The initial feedback is positive.

Simulation of a glass bottle with manufacturing defects (visual generated by AI)

AI enters Kestrel Vision’s strategy

Strategic roadmap prescription

Beyond its subsidiary Tiama, Kestrel Vision sees the opportunity to expand this AI approach group-wide, establishing it as a true strategic lever in service of its vision for innovation and Industry 4.0. All subsidiaries—whether they operate in glass, PET, or metal—are now involved.

The stakes have risen: how can artificial intelligence be integrated into each entity’s offerings while capitalizing on existing expertise and synergies?

Neovision offers tailored support: the co-construction of a 3-year strategic AI roadmap for Kestrel Vision.

This approach is based on several key analyses:

• • Mapping of AI use cases within the group and subsidiaries, evaluating their business value, technical feasibility, and technological maturity;
  • Analysis of the external environment, including traditional competition and new entrants with AI-based products;
  • Analysis of AI trends and product impact analysis;
  • Proposal for Data & AI Governance;
  • Prioritization of high-potential projects and projection of resources to be mobilized.

This work allows Kestrel Vision to lay the foundations of a clear, structured, and shared AI strategy, perfectly aligned with its ambitions for innovation, industrial excellence, and differentiation.

Empowering the client

The upskilling of Tiama’s teams becomes a central pillar of Neovision’s vision: building, with and for its clients, a progressive and structured autonomy in service of a sustainable transformation.
From the initial phases, Neovision involves Tiama’s technical teams in the projects and trains them in the various stages: data annotation, model training, performance analysis, production integration…
Tiama can now be autonomous; the company no longer positions itself as a mere user of AI solutions, but as a key player capable of leading certain developments.

Neovision as strategic and operational support

Neovision continues to support Kestrel Vision as a strategic partner by providing its consulting expertise. The team also steps in for operational support when Kestrel Vision’s internal priorities require reinforcement capable of handling specific developments with agility and autonomy.

Testimonial

For several years, Neovision has been supporting us in identifying, evaluating, and industrializing deep learning technologies. Their expertise has allowed us to choose solutions that are both tailored and state-of-the-art, while defining an effective strategy—from initial data management to the deployment of models in production.

Thanks to this partnership, we remain at the forefront of defect detection and classification, allowing our customers to improve the quality and management of their manufacturing processes. We are thus gradually replacing our “expert” vision systems with more flexible solutions, better suited to their needs.

Our collaboration with Neovision has proven essential to successfully completing this technological transformation.

Sébastien ROMAN
AI Technological Platform Director
Kestrel Vision