Ancient Roman Coin Recognition in the Wild Using Deep Learning Based Recognition of Artistically Depicted Face Profiles

Imanol Schlag, Ognjen Arandjelovic; Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 2898-2906

As an interesting application in the realm of cultural heritage, the challenging problem of computer based analysis of Roman coins is attracting an increasing amount of research. Herein we make several important contributions. Firstly, we address a key limitation of existing work characterized by the application of generic recognition techniques and the lack of use of domain knowledge. Our work approaches coin recognition in much the same way as a human expert would: by identifying the emperor on the obverse. To this end we develop a deep convolutional network, crafted for the specific instance of profile face recognition. No less importantly, we also address a major methodological flaw of previous experiments which are insufficiently systematic and mired with confounding factors. We introduce three carefully collected and annotated data sets, and demonstrate the effectiveness of the proposed approach which exceeds the performance of the state of the art by an order of magnitude.