Visual inspection is often performed manually when visual understanding on a human level is required and the classical methods of computer vision may not be reliable. This may be due to a wide variability between the parts inspected or to inconsistent conditions under which the products are inspected. One possible solution to this problem is to use machine vision in combination with deep learning. With this modern approach, the system can be trained to distinguish between faulty and non-faulty products or even to classify different types of defects, based on a large data set of training examples. Two main topics are being researched: defect detection using supervised learning and anomaly detection using unsupervised learning.
The investment is co-funded by: