Gabriele GATTIGLIA | Francesca ANICHINI | Michael REMMY | Holly WRIGHT | Massimo ZALLOCCO
(University of Pisa, Pisa, Italy)

Keywords: Deep Learning, image recognition, shape recognition, Archaeology, Ceramic

Pottery is the most common archaeological evidence and is of fundamental importance for the comprehension and dating of archaeological contexts, and for understanding the dynamics of production, trade flows, and social interactions. Today, classification of ceramics is a time-consuming activity carried out manually, through the expertise of specialists and the use of analogue catalogues held in archives and libraries. The ArchAIDE project (2016-2019) is funded by the European Union’s Horizon 2020 research and innovation programme and aims to create a new system for automated recognition of archaeological pottery with an innovative app designed for tablets and smartphones, without changing the current overall approach and reasoning process of the archaeological community. Ceramic fragments can be photographed, their characteristics sent to a comparative collection, which activates an automated object recognition system, resulting in a response with all relevant information linked to the image of the fragment.
Deep learning algorithms have been developed for classifying pottery, both through decoration and profile. Using appearance-based and shape-based recognition, neural networks have been trained with synthetic data and more than 13.000 images of real potsherds. This process has informed the design of a reference database for pottery types, decorations and stamps to be used as comparative target by the classification tools. It has also resulted in the development of advanced OCR tools for the digitisation of paper catalogues, the automated of extraction of 3D models from 2D drawings, and the creation of a corpus of thousands of classified pottery sherds to refine the neural network training.
Currently, ArchAIDE app is a proof on concept, working with catalogues related to the Terra Sigillata (Italica, Hispanica, South Gaulish), Roman Amphorae, Italian Majolica of Montelupo, and Spanish medieval and post-medieval majolica from Barcelona.

Relevance for the conference: The app is especially created for supporting archaeological professionals and researchers during fieldwork such as urban excavations and development led archaeology.
Relevance for the session: The paper deals not only the technical aspect but also issues of paramount importance in the archaeological domain such as copyright management, open data and exploitation plan.
Innovation: The automatic recognition of archaeological potsherds represents an innovation by itself and having it at your fingertips through a mobile application is even more innovative.
• BANTERLE, F., DELLEPIANE, M., EVANS, T., GATTIGLIA, G., ITKIN, B. and ZALLOCCO, M. 2017. The ArchAIDE Project: results and perspectives after the first year. In R. SABLATNIG and B. ŠTULAR (Eds.) EUROGRAPHICS Workshop on Graphics and Cultural Heritage, pp. 161-164.
• GUALANDI, M.L., SCOPIGNO, R., WOLF, L., RICHARDS, BUXEDA I GARRIGOS, J., HEINZELMANN, M., HERVA, M.A., VILA, L. and ZALLOCCO, M. J2017. ArchAIDE – Archaeological Automatic Interpretation and Documentation of cEramics. In C. E. CATALANO and L. DE LUCA (Eds.) EUROGRAPHICS Workshop on Graphics and Cultural Heritage