Deep learning is at the core of the current AI revolution and represents the main research focus of the professorship. How can common practical tasks like quality control, defect classification, disease detection, object recognition and segmentation be automated using deep learning? This includes research into the development of the following tasks: 
  • Anomaly detection research focusses on applications when there are only negative samples available, and examples of positive samples are missing. This can be used to detect anomalies, unknown damage and other deviations from normal situations. 

  • Synthetic data research focusses on generating digital twins using pre-trained generative models which are fine-tuned to generate data from a few exemplary images, text prompts or 3d graphics. 

  • Few-shot and zero shot research focusses on handling situations when there are only a few images per class available or even for situations where there are now images available. 

  • Explainable AI aims to give feedback on where the models were basing their decisions on. This research focusses on how to prevent the model from overtraining on small amounts of data and has broader applications in signaling bias and alleviate the black-box nature of AI. 

Further research lines focus on improving data quality though the use of sophisticated image acquisition set-ups and automated image-data mining and image querying strategies. 

Discover our other research lines

Want to learn more about the professorship Computer Vision & Data Science and our research lines?