Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).
Monitoring and analysing archaeological destruction in Afghanistan using satellite imagery
Charlotte Fafet, Titien Bartette, Jonathan Chemla, Mehraïl Saroufim
Iconem, France
Afghanistan’s rich cultural heritage has been severely threatened by decades of political instability, conflict, and environmental degradation. Home to some of the world’s most important archaeological sites, the country has seen its heritage increasingly jeopardised by looting, urban expansion, and large-scale infrastructure projects. In this challenging context, the French Archaeological Delegation in Afghanistan (DAFA), established in 1922, has long been at the forefront of efforts to document and protect the region’s cultural legacy.
Due to the recent geopolitical shifts and the withdrawal of international presence, DAFA has been forced to suspend its on-the-ground activities since 2021. To continue its mission remotely, DAFA partnered with Iconem, recognised for its innovative approach to heritage preservation. Whilst its core business lies in creating high-resolution 3D models, Iconem adopted a different approach by using satellite images to monitor landscape and structural changes. This new approach is particularly relevant in regions affected by war or instability, where on-the-ground access is no longer possible.
This collaboration emphasises the importance of remote sensing technologies to bridge the gap left by the absence of physical fieldwork. Our project focuses on monitoring the destruction of archaeological sites in Afghanistan and aims to identify and document the impact of looting and large-scale infrastructure projects on these sites. Iconem's expertise in recording and monitoring the conservation state of cultural heritage sites complements DAFA’s extensive historical data, allowing for the continuous observation of these critical locations.
This presentation will outline the project’s framework, approach and methodologies adopted for monitoring these destructions. We will share preliminary findings, considering the study is continuously updated with new data. Through several case studies – such as Bactria, the Qosh Tepa canal, the TAPI pipeline, and the site of Mes Aynak – we will explore the ongoing challenges and opportunities in safeguarding Afghanistan's endangered cultural heritage, and the potential future directions for expanding this critical work.
The ALCEO Project: Machine Learning and Remote Sensing for Looting Detection
Gregory Sech, Riccardo Giovanelli, Giulio Poggi, Marco Fiorucci, Arianna Traviglia
Istituto Italiano di Tecnologia, Italy
Looting of archaeological sites is a significant global threat to cultural heritage, resulting in irreversible damage to invaluable historical landscapes and the dispersion of cultural goods. Traditional monitoring methods are often inefficient and require specialized expertise, limiting their effectiveness in preventing or responding to these illicit activities. The ALCEO project addresses this issue by combining machine learning (ML) and remote sensing technologies to detect and monitor looting activities, providing an efficient solution to protect archaeological sites.
The project focuses on the use of satellite imagery and deep learning models to automatically detect looting pits across various archaeological landscapes. ALCEO has been applied to several key sites, including Arpinova, Cerveteri, Aquileia, Morgantina (Italy), Dura Europos and Ebla (Syria), and Aswan (Egypt). These sites are diverse in terms of terrain and environmental conditions, offering a robust testbed for the system's capabilities.
Developed to address the needs of law enforcement agencies and cultural heritage institutions, the ALCEO system automates the detection process by comparing sequential satellite images to identify changes over time. The system consists of two main components: the "Data Management Sub-system," which generates change detection datasets, and the "Modelling Sub-system," which manages the training, evaluation, and inference of the deep learning models. A geodatabase containing over 6,000 looting traces from various locations has been created to train the system, with the deep learning model employing a fully convolutional Siamese network architecture to detect new looting pits.
ALCEO has demonstrated strong performance, achieving an Intersection over Union (IoU) score of 0.7145 across test sites. Additionally, the system produces risk maps that enable stakeholders to monitor and mitigate looting activities, playing a vital role in preserving global cultural heritage.
