Case studies

Initially, Vooban's main task was to develop an artificial intelligence model for defect recognition. This is a supervised model, which detects and categorizes all possible anomalies on a given product.

In the second phase, we also developed the user interface and the web application where all product images are cataloged. This interface allows the company to monitor the quality assurance process and adjust their tolerance thresholds for defect detection.

Vooban also supported the company with the collaborative robotics component for in-factory handling. In a sterile room, a UR5 robot inspects the parts independently, sorting them based on their quality level. To do so, it uses a vacuum gripper tool to move the parts without damaging them, providing a complete handling and defect detection solution.

As is it usually the case with the pharmaceutical and medical industries, the products are subjected to rigorous quality standards and numerous audits and must therefore be free of defects that could compromise medical procedures.

However, these defects are often very difficult to spot because they are microscopic (sometimes less than one square millimetre) and vary greatly in appearance. Scratches, stains or black spots, for example, are never identical from one product to the next and require exhaustive verification to avoid recalls or incidents.

The company previously performed their quality control manually, an inefficient method that allowed for a high error rate.

As of today, our detection model exceeds FDA standards. aIt achieves a performance rate of over 99%, while generating less than 3% false rejections, an incredible result for an industry where anomaly detection is absolutely critical. A project like this is perfectly suited to the medical field, where strict compliance is required for product distribution and defects are often microscopic.