Vanishing archaeological landscapes in Mesopotamia:CORONA imagery site detection (arxiv.org)

Researchers have successfully enhanced a deep learning model to automatically identify archaeological sites in Mesopotamia using historical CORONA satellite imagery. By retraining a Bing-based convolutional network with these vintage grayscale images, the study focused on the district of Abu Ghraib near Baghdad, an area severely altered over the past 50 years. The results revealed a remarkable increase in detection accuracy, achieving over 85% Intersection over Union (IoU) values and a general accuracy of 90% in identifying archaeological locations. This advancement is significant for the AI/ML community as it showcases the potential of integrating deep learning with historical datasets to uncover archaeological sites that have been lost to urban development and environmental changes. Notably, the retrained model successfully identified four new archaeological sites that had eluded traditional detection methods, highlighting AI's capacity to aid in cultural preservation efforts. This study illustrates the transformative impact of AI technologies in fields like archaeology, enabling researchers to detect and study landscapes that are at risk of vanishing due to human activity.