The University of Liechtenstein is pleased to announce the publication of a new scientific study in the renowned academic journal Electronic Markets. The article, titled “Fine-Tuning Image-to-Text Models on Liechtenstein Tourist Attractions”, was authored by Dr. Pejman Ebrahimi, Postdoctoral Researcher at the Liechtenstein Business School, and Prof. Dr. Johannes Schneider, Professor of Data Science & Artificial Intelligence.

The study explores how effectively modern AI models can automatically generate text descriptions from images of tourist attractions – a technology with potential applications in tourism, cultural heritage communication, and accessible digital services. The research focuses on two of Microsoft’s image-to-text models: GIT and Florence-2.

Both models were specifically trained using image data from Liechtenstein – including landmarks such as Vaduz Castle, the Malbun alpine region, and various cultural events. The results show that the GIT model performs well even with relatively small training datasets. In contrast, Florence-2 demonstrates significantly better performance when larger amounts of training data are available.

This research offers valuable insights into the practical use of AI in tourism and heritage contexts. It highlights which models are best suited for specialized applications such as the automated creation of image descriptions for websites, apps, or digital travel guides.

The paper was published in Electronic Markets, a leading international journal on digital markets and information systems.

Ebrahimi, P., & Schneider, J. (2025). Fine-Tuning Image-to-Text Models on Liechtenstein Tourist Attractions. Electronic Markets, 35(55). https://doi.org/10.1007/s12525-025-00806-7