TTZ Günzburg employee Jens Gebele researches hybrid intelligence at the University of Oulu

The aim of the project is to combine machine learning methods with logical reasoning. While deep learning (DL) models have high computing power and pattern recognition capabilities, they often lack transparency and explainability. By integrating symbolic knowledge—such as from the Facial Action Coding System, which describes which facial muscle movements are associated with certain emotions—a system can be created that checks and improves its own predictions.

During his stay, Gebele worked on preparing data sets, developing initial model architectures, and evaluating the well-known BP4D database, which contains extensive information on facial expressions. The results showed that there are significant gaps in the annotations and differences between tasks and emotional responses – an indication that hybrid approaches combining machine learning and human expertise are crucial to making emotion recognition more robust and comprehensible.

In addition to scientific work, personal exchanges and Finnish culture also shaped the stay. “The openness and curiosity of the researchers in Oulu made the collaboration particularly enriching,” emphasizes Gebele. The cooperation with Dr. Yante Li and Prof. Guoying Zhao will continue: Further experiments are currently underway, and a joint scientific publication is in preparation.

With this project, TTZ Günzburg is underlining its commitment to conducting state-of-the-art AI research. The findings will form the basis for successful technology transfer.

Contact
Jens Gebele