The digital preservation of cultural heritage is an important and challenging task
for the research community. Reconstructing historical objects, which do not exist
anymore, in the form of digital 3D models makes it possible to visualize them and
present them to the public. The reconstruction process as well as the visualization lead to a deeper understanding of the lost historical objects. But the process of the digital reconstruction is complex and time consuming as diverse sources have to be consulted and interpreted. Therefore, in this paper the latest technology in the feld of artifcial intelligence (AI) is used to support researchers in the feld of Digital Humanities: A Transformer deep learning model based on questions answering methods is introduced to assist to digitally reconstruct historical objects in 3D. It implies a new dimension of data availability, which supports the knowledge process by making large amounts of data qualitatively accessible.
To demonstrate the potential of Transformers two examples of historic objects were
selected: the architecture of Sophienkirche in Dresden, a church stemming from the
13th century, which was destroyed in World War II, and a feline incense burner of the
Tiwanaku culture, which fourished between AD 200 and 1100 in the Central Andean
Highlands of Peru and Bolivia.
First the two historic objects are shortly introduced followed by an overview of the
history and state of the art of Transformers. In the next step the research methodology of the project is presented. The following chapter shows the results and evaluates them.
Then the educational value of Transformers is explained. The last chapter summarizes and discusses the results of the project and discusses the value of transformers in context with the reconstruction process and the learning environment.