Quality inspection

Context

Manufacturing (Steel)

Problem

It’s very important to detect and classify defects on products to keep the production line up 24/7 and maximize the business value of each coil.

Profit margins rely heavily on the accuracy of product quality assessment. Insufficient quality causes dissatisfied customers, or worse, a complete shutdown of the production line resulting in millions of lost revenue.

Solution

Automated detection and classification of production errors over 50 classes using 50 million infrared camera images per day. The system combines deep learning based classification with active learning components. Thanks to an expert feedback loop the system will continuously improve itself over time.

Results

  • This process significantly outperforms the previous industrial vision system
  • Increased pricing accuracy
  • Higher customer satisfaction
  • Elimination of production line outages
  • More efficient use of expert knowledge

More information?

Request White Paper