< 1%
False rejection
18+
tested defect classes
15.000 p/h
inspection speed
Customer's Problem
Defect detection in production is currently done manually at two points on the production line and requires highly skilled specialists. This leads to high scrap rates, long inspection times and limited scalability. In addition, there is no efficient way to continuously improve quality assurance. To increase productivity and reduce error rates, an automated, learning solution was required that enables more precise and faster inspection.
Our solution
We have developed an AI-based inspection solution that extends the existing inspection machine with a powerful ML infrastructure. Using a specially developed interface, image data is captured, annotated and processed in an MLOps environment so that the models can be continuously optimized and expanded to include new types of defects. Customer specialists can directly participate in improving the AI by annotating data and specifically retraining models. This enables highly precise defect detection with a low error rate, while at the same time accelerating the inspection process and reducing scrap.