Abschlussarbeiten
Auf dieser Seite werden regelmäßig Themen für mögliche Abschlussarbeiten veröffentlicht, wir sind auch offen für Ihre Ideen.
Wenn Sie Interesse haben an einem der Themen, senden Sie bitte eine E-Mail an: info-ttz@hnu.de mit einem Kurzprofil über sich und dem entsprechenden Thema.
Aktuelle Themen
Please note that all topics are discussion base and may be adjusted
Computer Vision
- Which role does image resolution play in computer vision? Investigating the influence of image resolution on the performance of a Convolutional Neural Network model.
- Which role does contrast in images play in computer vision? Investigating the influence of image contrast on the performance of a Convolutional Neural Network model. Background to the topic: The contrast in female faces is higher than in male faces (Russell, 2022).
- An empirical ML experiment: Can data inconsistencies be combat by merging ML models trained with different data sets of the same domain (here: facial emotion recognition)?
- Investigating data label bias / errors in (training and/or test sets of) visual datasets.
- Vision Transformers: What advantages do the have over Convolutional Neural Networks?
- Development of a person re-identification (reID) system on the edge (NDVIDIA Jetson Nano) / in multicamera environment
Hands on: Computer Vision in use of tire retreading. Development (and implementation) of a ML concept for the detection and classification of damage spots on pre-processed scrap tires.
Reinforcement Learning
Investigation of the use of Reinforcement Learning in context of explainable AI – Common use cases and potential for improvement.
Hands on: Using Reinforcement Learning for training a humanoid robot (NAO).
Model-related
- Generative Adversarial Networks: Identifying different architectures, their advantages, disadvantages and differences.
Neuro-Symbolic Artificial Intelligence: What is state-of-the art? What are current and future trends?
Confident learning: an approach to assess the data label quality.
Data-related
- Data Architecture: Comparative Analysis between Data Mesh and Data Warehouse / Data Lake. Systematic Overview of Explainable Artificial Intelligence.
Abgeschlossene Arbeiten 2021
- "KI kann was! - Eine empirische Studie zur Konzeption eines KI-Canvas zur Evaluierung der Realisierbarkeit von KI in KMU"
- "Potenzialanalyse und Ableitung von Handlungsempfehlungen zur Belebung der Günzburger Innenstadt"
