A human–AI collaboration workflow for archaeological sites detection

Abstract

This paper illustrates the results obtained by using pre-trained semantic segmentation deep learning models for the detection of archaeological sites within the Mesopotamian floodplains environment. The models were fine-tuned using openly available satellite imagery and vector shapes coming from a large corpus of annotations (i.e., surveyed sites). A randomized test showed that the best model reaches a detection accuracy in the neighborhood of 80%. Integrating domain expertise was crucial to define how to build the dataset and how to evaluate the predictions, since defining if a proposed mask counts as a prediction is very subjective. Furthermore, even an inaccurate prediction can be useful when put into context and interpreted by a trained archaeologist. Coming from these considerations we close the paper with a vision for a Human–AI collaboration workflow. Starting with an annotated dataset that is refined by the human expert we obtain a model whose predictions can either be combined to create a heatmap, to be overlaid on satellite and/or aerial imagery, or alternatively can be vectorized to make further analysis in a GIS software easier and automatic. In turn, the archaeologists can analyze the predictions, organize their onsite surveys, and refine the dataset with new, corrected, annotations.

The Synergy between Artificial Intelligence, Remote Sensing, and Archaeological Fieldwork Validation

“…Building on these insights, the algorithm proposed in our previous work was refined on the basis of the empirical knowledge gained from fieldwork validation. The data resulting from this validation, including information on True Positives and False Positives, enriched the datasets used to fine-tune the machine learning models for future archaeological surveys [9,38]. We used slope maps derived from LiDAR-derived DTMs to eliminate inferences in high-slope regions.…”

“…This is crucial in archaeology, where manually identifying archaeological features from remote sensing data can be timeconsuming. Archaeologists are employing these algorithms to detect archaeological sites and features in various data types [8][9][10][11][12][13], such as LiDAR data, multispectral, hyperspectral, and satellite aerial imagery. By allowing them to identify and map potential archaeological sites or features, the output of such algorithms allows archaeologists to plan targeted surveys, saving both time and resources.…”

Validating predictions of burial mounds with field data: the promise and reality of machine learning

“…Today, however, more and more remotely sensed datasets are becoming available, requiring scalable and reproducible approaches to their analysis. Manual inspection is not without merit, and expert supervision is essential toor may even outperformautomation (Bennett et al, 2014;Casini et al, 2023;Quintus et al, 2017). Human observers, however, cannot manage the growing volume of remotely sensed data, or harness all the data contained in modern, global, high-resolution, multi-or hyperspectral imagery.…”

“…Even many successful ML applications require significant human intervention either in the form of crowdsourcing or specialist work (see, e.g. Casini et al, 2023;Doyle et al, 2023;Verschoof-van der Vaart and Lambers, 2019). As a result of challenges like these, Casana (2020), goes so far as to reject ML approaches entirely, instead favouring "brute force" manual inspection of satellite imagery by experts.…”

“…Use of a pre-trained CNN potentially obviates the need to have large, high-quality, and representative datasets for ML training, and promises to bring CNN approaches within the reach of smaller-scale projects, or projects that hope to detect or monitor varied archaeological phenomena in heterogeneous landscapes. Relatively few projects applying pre-trained models, however, offer a sustained discussion of failures, limitations, or shortcomings of the approach (for exceptions see Casini et al, 2023;Sech et al, 2023).…”