A new 3D dataset of celtic monetary dies. Nakala, collection AOrOc

We propose a new 3D dataset of 2 070 scans of coins. With this dataset, we give two benchmarks, one for point cloud registration, essential for coin die recognition and a benchmark of coin die clustering. We show how we automatically cluster coins to help experts and perform a preliminary evaluation for this two tasks.

Projet en cours

Partenaires institutionnels

AOROC - UMR8546-CNRS/ENS IRIS SCRIPTA-PSL Mines-ParisTech_PSL Musée de Bretagne- Les Champs Libres -Rennes Nakala PSL  - Paris Sciences et Lettres | université de recherche Semiologiegraphiquemonnaies

Celtic coins are minted from two dies, one for each face, the obverse and reverse monetary dies. As the surface of these dies is greater than that of the coins, it takes several coinsissued from the same die to restore the image engraved on the original dies.

Identifying coins issued from the same die is a long and difficult visual operation. The arrival of high-resolution 3D scanners makes it possible to work on point clouds (.stl) and meshes (.ply) that ignore surface effects. We use deep learning to recognize similar patterns between two coins and recalibrate the coins to see if they were struck with the same die or not.

We provide a 3D dataset of 2,070 scans of coins from the Musée de Bretagne de Rennes on which we have tested our methods of automatic coin classification.
We show how we automatically classify coins by coin to help experts and perform a preliminary assessment for these two tasks.

This dataset is currently being entered and will be online soon (February 2022).