These five students managed to achieve better results in image recognition, particularly in identifying coin types and mints, by using the provided CN dataset of approximately 14 GB. They did this through techniques like voting and stacking, as well as a multimodal approach, surpassing the performance of the previously used VGG-16 architecture of Convolutional Neural Networks in the project.
The students, along with our colleagues Karsten Tolle, Sebastian Gampe, and Chrisowalandis Deligio, visited Berlin. They began their trip by visiting the Reichstag dome. Then they spent several hours at the BBAW and had the opportunity to study the original coin collection of the project. On the second day, the director, Bernhard Weisser, gave a guided tour through the coin exhibition of the Bode Museum and allowed a look into the impressive vault.
We are thankful for the visit and the interesting discussions we had!
Συγγραφέας: Ulrike Peter
