The first paper, entitled “Towards Trustworthy AI: Evaluating SHAP and LIME for Facial Emotion Recognition”, is based on the master's thesis of Selina Lorch, a research assistant specializing in data analytics and artificial intelligence at TTZ Günzburg. It investigates how AI models for facial emotion recognition can be made more transparent and comprehensible. The focus is on comparing the SHAP and LIME explanation methods, which show which areas of the face - such as the eyes, mouth and cheeks - are particularly relevant for emotion recognition. The research shows that explainability is a central building block for trustworthy AI systems, while human interpretation of the results remains indispensable.
The second paper, “Assessing Sequential Databases for Spontaneous and Posed Facial Expression Recognition”, was written as part of the cooperative doctorate of Jens Gebele, research assistant with a focus on Data Analytics & Artificial Intelligence, which is being conducted via BayWISS in collaboration with Prof. Dr. Frank Schwab and Prof. Dr. Sebastian von Mammen from the University of Würzburg. This study analyzes the quality of databases containing spontaneous (real) and posed (artificial) facial expressions. The research provides practical recommendations on how to improve the quality of datasets - an important step towards more accurate and reliable AI systems.
Both research contributions emphasise the leading role of the TTZ Günzburg in the field of artificial intelligence. The knowledge gained not only provides a basis for future scientific projects, but also the opportunity to transfer innovative AI technologies to the regional economy and thus sustainably strengthen the region's competitiveness.
