AI Applications for Industry With Little Data

Artificial Intelligence methods such as »Deep Neural Networks« have revolutionized image recognition. In the field of industrial quality control, more flexible and generic systems are being developed than with traditional methods. AI also makes it possible to optimize the quality control of structured materials or even natural materials.

A major challenge of AI methods is the need for a large amount of well-annotated, balanced data. Due to good production quality, the provision of defect images is particularly problematic. Nevertheless, we have successfully brought AI into industrial use using various approaches (synthetic data, hybrid approaches, transfer learning, etc.). The following project examples serve to illustrate what this looks like in research practice.