a research team coordinated by the University of Jaen how he designed a system that recreates would be a complete object made of ceramic pieces Found in the upper valley of Guadalquivir. this application recognizes the archaeological remains of the piece and suggests how it might have been in the beginning. This tool allows archaeologists to more accurately reconstruct the remains found in their excavations.
In previous research, experts developed a comprehensive catalog and classification of Iberian ceramic fragments. Now they have created a tool that commands the automatic discovery of any new object or archaeological remains. To do this, they use criteria set by archaeologists to determine the shape: open or closed, with or without a neck, spherical or oval.
IberianGan, one more step
Now they’re going one step further by training the system they call it. IberGan, thus providing a complete picture of the ceramic object to which a particular piece belongs. In the article “Reconstruction of Iberian pottery using productive adversary nets” published magazine scientific reports, experts explain the model used and confirm its validity. Archaeologists confirm the quality of the reconstructed specimens and the usefulness of the tool, which will significantly facilitate the task of assembling the remains.
In excavations to find archaeological remains, ceramics are normally found in fragmented form, requiring on the one hand each item to be cataloged, on the other hand the entire object inspected and reconstructed by hand. “IberianGAN solves the stone problem by automatically comparing between known pieces. The best match in the dataset is the solution of that ceramic,” he says. Discover the Foundation researcher at the University of Jaén Manuel Lucenathe author of the article.
The system has already cataloged the different possible parts in an extensive database. It also creates new parts with the same characteristics as real parts. In this way, the tool creates mounting possibilities for creating the full ceramic.
Experts presented both the data and the source code to the use of the scientific community with the configuration and different approaches analyzed in the study. https://github.com/celiacintas/vasijas/tree/iberianGAN with the intention that they may be reproduced and extended to other investigations.
Self-learning artificial neurons
Thus, the tool learns to distinguish which item is real and which is not, through what is known as deep learning. The model used is based on productive competing networks or antagonistic networks, GAN for its English abbreviation. These are two opposing neural networks. They ‘play’ against each other, with one having a ‘True’ card and the other a ‘False’ card. One always wins what the other loses.
Thus, to generate probabilities, the system is trained by creating real images and comparing them with other non-existent images. The tool discriminates the unreal, but has the ability to generate something that looks real from the cards it receives from its opponent.
By simulating ceramics created from others already assembled by archaeologists, the researchers trained the network and replicated the breakage they would have suffered if they were in an area. That way, they already had the necessary items for the tool to learn on its own any possible combination to display the actual item.
In addition to verifying its accuracy with archaeologists, the experts compared the results obtained through this deep learning model with others, confirming that antagonistic networks are ideal for training Iberian ceramic reconstruction tools. On the other hand, they propose expanding the system, including the creation of three-dimensional images from the same profiles, or even recreating the decoration that found containers can have. They also suggest applying it with other types of relics, such as bones, shells, and even buildings.
The work was funded through the FEDER Andalusia Operational Program 2014-2022, the University of Jaén, the Center for Advanced Studies in Information and Communication Technologies (CEATIC) and the Iberian Archaeological Research University Institute. The research team consists of researchers affiliated with the University of Jaén Research Institute of Iberian Archeology, CONICET (Argentina) Patagonia Institute of Social and Human Sciences, and the Department of Electrical and Computer Engineering of the National University. South (Argentina).
#Andalusian #smart #technology #reconstruct #archaeological #remains #Jaén