Navigating a new digital interface: using automated image recognition to identify pottery in the ArchAIDE project
Gabriele Gattiglia, Francesca Anichini, Holly Wright
EAA Barcelona, September 6, 2018
Re-membering the Bard: Revisiting The Compleat Wrks of Wllm Shkspr (Abridged)...
Navigating a new digital interface: using automated image recognition to identify pottery in the ArchAIDE project
1. Gabriele Gattiglia1,
Francesca Anichini1, Holly Wright2
1University of Pisa, 2University of York
EAA Barcelona, September 6, 2018|
Navigating a new digital interface: using
automated image recognition to identify
pottery in the ArchAIDE project
2. Navigating a new digital interface: using automated image
recognition to identify pottery in the ArchAIDE projectPartners
University of Pisa (coordinator)
Dipartimento di Civiltà e forme del sapere
CNR –Istituto di Scienza e Tecnologie
dell’Informazione
INERA srl
University of Barcelona
Fac. de Prehistòria, Història Antiga
i Arquelogia
BARAKA
ELEMENTS
University of York
Archaeology Data Service University of Cologne
Institut für Archäologie
University of Tel Aviv
School of Computer Science
An EC H2020 project
Call Reflective 6, RIA
(Research and Innovation
Action)
Duration: 36 months
Project Started on June
1st, 2016 and end on May
31st 2019
3. ArchAIDE aims to support the classification and interpretation work of the
archaeologists with innovative computer-based tools, able to provide the user with
features for matching of each discovered sherd over the huge existing ceramic
catalogues.
from to
Navigating a new digital interface: using automated image
recognition to identify pottery in the ArchAIDE project
4. recognition requires:
• complex skills and since it is heavily
dependent on human inspection and
interpretation it is a very time
consuming activity;
• a boundless bibliography, fragmented
and incoherent, whose consultation is
long and fatiguing also when is
available a well furnished library …
Navigating a new digital interface: using automated image
recognition to identify pottery in the ArchAIDE project
We want to innovate the archaeological practice, introducing a modern
computer-aided approach
but we want to keep as much as possible unchanged the overall methodology,
to ensure easy adaptation and impact in the archaeology domain.
5. Navigating a new digital interface: using automated image
recognition to identify pottery in the ArchAIDE project
10. Appearancebasedrecognition
1 3
Ranking of
pottery classes
by relevance
2
Shapebasedrecognition
Navigating a new digital interface: using automated image
recognition to identify pottery in the ArchAIDE project
14. This project has received funding from the European Union’s Horizon 2020 research
and innovation programme under grant agreement N.693548
The views and opinions expressed in this presentation are
the sole responsibility of the authors and do not necessarily
reflect the views of the European Commission.
[Thank you for your attention]
www.archaide.eu
15. Navigating a new digital interface: using automated image
recognition to identify pottery in the ArchAIDE project
We are at stand 43 (in the courtyard)
Notas do Editor
One of the main goals of ArchAIDE project is to support the work of archaeologists towards an automatic classification (and description) of pottery sherds.
This can be reached by taking into account the features that archaeologists compare in order to assign a sherd to a typology. While some of them are hard to be inserted in an automatic system (due to the necessary background and high-level reasoning of archaeologists), there are two descriptors that could be used: shape and appearance. Both of them are used on field, with a different degree of importance w.r.t. different classes of ceramic. Shape and appearance information has been structured in a number of catalogues, which are the main reference during the classification process.
From the technical point of view, there are two main aspects that have to be taken into account:
• The information that will be extracted from the catalogues (mainly from the reference drawings, but possibly also from the text) has to be extracted also from the images acquired on-site. Hence, it’s necessary to select a proper set of features, and also to provide guidelines for the acquisition of images when excavating.
• The classification will be performed by an automatic system that has to be “trained” using real or synthetic data. Hence it is necessary to be able to provide enough input data so that the system is able to learn how to classify in a robust way.