Volumes / Journal / 37.1 / Palombini et al.

AIOH: Artificial Intelligence Operator for Heritage. The case study of nuraghe detection

Augusto Palombini, Edoardo Baiocchi, Saverio Giulio Malatesta, Romeo Lanzino, Marco Raoul Marini, Paolo Rosati

«Archeologia e Calcolatori» 2026, 37.1, 229-254; doi: 10.19282/ac.37.1.2026.12

Abstract

Detecting archaeological features from satellite images is an increasingly vital tool in archaeology, offering a non-invasive means to discover and protect ancient structures. Among these, nuraghes, which are tower-shaped stone structures from the Bronze Age unique to Sardinia (Italy), present a significant challenge due to their varied sizes, forms, and environmental contexts. This paper introduces a pioneering approach employing an Artificial Intelligence (AI) driven method for detecting nuraghes in satellite images, specifically employing a Deep Learning (DL) model based on Vision Transformers (ViT), which is an advanced neural network architecture that excels in processing image data. The system, called ‘Artificial Intelligence Operator for Heritage’ (AIOH), is trained on custom data retrieved for this work: the dataset, NuragAI, includes thousands of satellite images from Google Maps with manually annotated nuraghe presence. It is designed to train the model to identify subtle and complex characteristics of nuraghe structures. The effectiveness of this approach was quantitatively assessed through extensive validation, with very high results on the training and testing set and interesting results also on data, detecting nuraghes in various terrains and under different lighting conditions. A preliminary on-site field campaign, to test results, was also carried out, confirming the model efficiency.

Figures

Palombini et al. 2026, fig. 1 - thumbnail — Fig. 1

Palombini et al. 2026, fig. 2 - thumbnail — Fig. 2

Palombini et al. 2026, fig. 3 - thumbnail — Fig. 3

Palombini et al. 2026, fig. 4 - thumbnail — Fig. 4

Palombini et al. 2026, fig. 5 - thumbnail — Fig. 5

Palombini et al. 2026, fig. 6 - thumbnail — Fig. 6

Palombini et al. 2026, fig. 7 - thumbnail — Fig. 7

Palombini et al. 2026, fig. 8 - thumbnail — Fig. 8

Palombini et al. 2026, fig. 9 - thumbnail — Fig. 9

Palombini et al. 2026, fig. 10 - thumbnail — Fig. 10

Palombini et al. 2026, fig. 11 - thumbnail — Fig. 11

Palombini et al. 2026, fig. 12 - thumbnail — Fig. 12

Preview