The conference Deep Learning Theory and Applications (DeLTA 2020), a sister conference of ICINCO, took place for the first time this year from 8 to 10 July 2020. Due to the Covid 19 pandemic, it was held entirely virtually instead of in Lieusant, Paris. Andreas Margraf (Department Online Process Monitoring at the Fraunhofer IGCV) and Silvan Mertes (former student at the Fraunhofer IGCV, now a research assistant at the Chair for Human-Centered Multimedia at the University of Augsburg under Prof. Dr. Elisabeth André) convinced the jury with their publication »Data Augmentation for Semantic Segmentation in the context of Carbon Fiber Defect Detection using Adversarial Learning«. The publication was initiated in partnership between the Fraunhofer IGCV and the Chair of Human-Centered Multimedia of Prof. Elisabeth André.

The presented method opens new possibilities for the development and configuration of image processing systems for continuous processes. Previous AI solutions require large amounts of training data, which can only be generated with great effort. By generating artificial data, the approach of Mertes, Margraf, Kommer, Geinitz, and Wedel allows to create and bring into production neural networks for semantic segmentation faster, easier, and with less effort. This significantly reduces the cost of providing systems for quality control.

Through its research work, the Fraunhofer IGCV, with its department Online Process Monitoring and the Chair of Human-Centered Multimedia at the University of Augsburg, places itself at the forefront of international research in the field of machine learning for the industry. The approach is transferable to different tasks in quality control, independent of the product. In the future, the researchers will continue to optimize their application to increase automation through AI systems and improve industrial production efficiency.

Please also read our recently published technical article: AI in fiber technology