Automatic pattern recognition on archaeological ceramic by 2D and 3D image analysis: A feasibility study

The ARCADIA project aims at developing automatic image analysis and machine learning methods to facilitate the interpretation and promotion of an archaeological ceramic heritage. It consists in implementing a 3D digitalization chain of the discovered shards to extract the decoration imprinted by the potters using a carved wood wheel. The characterization of the hollow pattern from a binary map given by the 3D image then allows the automatic classification of these patterns. We propose here a feasibility study by comparing the results of automatic classification from the manual inkings made by archaeologists and those from 3D digitalizations of the same corpus of shards.