CELTES 3D aims at automatic pattern recognition on 3D models of Celtic coins. To achieve this, deep learning methods are mobilized to help identify similar patterns between two coins and check whether they come from the same "corner". While the identification of hallmarks enables us to better understand the engraver's style and the way in which the object was made, this work also contributes to determining essential information in economic history.
But identifying the same motifs for several objects is a very long and tedious job, and requires a great deal of expertise. The objects may be timeworn or broken.
Computing resources are essential... Those of GENCI's Jean Zay supercomputer, hosted and operated at IDRIS were used for this pioneering project.
To better understand Sofiane Horache's work:
- My thesis in 180 seconds:
- Interview on BSmart: https://www.bsmart.fr/video/8929-smart-tech-partie-28-septembre-2021
- Mines Paris PSL press release: https://www.minesparis.psl.eu/Presse/Communiques-de-presse/-Quand-l-8217-Intelligence-Artificielle-tente-de-percer-le-secret-des-tresors-archeologiques-Mines-Paris-PSL-presente-les-travaux-de-recherche-de-Sofiane-Horache/461
Quart de statère région Armorique