Paper published in Computers & Industrial Engineering
Can unsupervised deep learning models compete with their supervised counterparts for visual defect detection while meeting the high safety standards of industrial quality assurance? The short answer is “yes, they can” as we could demonstrate in a recent study published in Computers & Industrial Engineering (CAIE) in collaboration with the Software Development Center (SDC) Dresden (Volkswagen AG).
Special credit to my former student, Justus Zipfel, for taking on this highly relevant topic and thereby strengthening the relationship with Volkswagen AG. Many thanks also to the entire team for the fruitful collaboration: Felix Verworner (SDC), Marco Fischer (SDC), Uwe Wieland (Hochschule für Technik und Wirtschaft), and Mathias Kraus (FAU Erlangen-Nürnberg).
The full paper is available via open access and can be found here and on ResearchGate.