Corpus Nummorum - Object Detection Coin Dataset

This Object Detection dataset is a collection of ancient coin images from three different sources: the Corpus Nummorum (CN) project, the Münzkabinett Berlin and the Bibliothèque nationale de France, Département des Monnaies, médailles et antiques. It covers Greek and Roman coins from ancient Thrace, Moesia Inferior, Troad and Mysia. This is a selection of the coins published on the CN portal (due to copyrights). 

This dataset contains 506 different classes with about 179.000 coin images (approx. 29.000 unique coins). The classes come from four different categories: persons, objects, animals and plants. The coin images were assigned to the classes using our NLP pipeline. For this purpose, our Named Entity Recognition and Relation Extraction were performed on every coin's description (separated into obverse and reverse). Each coin image assigned to this description was then copied to the folder of the predicted classes. A coin image can therefore also be assigned to different classes. The file name contains both the coin id and the coin type of the CN database. Whether the image belongs to a coin obverse or reverse can be recognized by the suffix obv or rev. An "sources" csv file holds the sources for every image. Due to copyrights the image size is limited to 299*299 pixels. However, this should be sufficient for most ML approaches.

Due to the numerically different occurrences of the individual entities, the data set is not balanced. In addition, a class can contain very different representations of the same entity. Therefore, some classes can be difficult to train. Unfortunately, we cannot provide any annotations for the data set.

During the summer semester 2024, we held the "Data Challenge" event at our Department of Computer Science at the Goethe-University. Our students could choose between the Object Detection dataset and a Natural Language dataset as their challenge. One team opted for the Object Detection challenge. We gave them this dataset with the task to use to try out their own ideas. Here are their results:

Multilabel Classification as Backbone for Object Detection

Now we would like to invite you to try out your own ideas and models on our coin data.

If you have any questions or suggestions, please, feel free to contact us.