![Archaeologists often face major challenges when trying to connect new discoveries with information from old books [Pixabay]](http://media.assettype.com/newsgram%2F2025-06-12%2Fwxbedaw4%2Fistockphoto-1312320636-612x612.jpg?w=480&auto=format%2Ccompress&fit=max)
Archaeology: Archaeologists often face major challenges when trying to connect new discoveries with information from old books: How can the findings of two hundred years of archaeological research be combined with new data? AutArch opens up completely new avenues here. It is based on neural networks that researchers have trained to independently detect, analyze, and relate common archaeological “objects” in catalogues, such as images of graves, human remains, pottery, and stone tools. AutArch does not only locate the data, but combines them to extract meaningful information.
"When analyzing a grave drawing, for instance, the software detects the north arrow and the associated scale – and can use this to calculate the actual size of the grave and its orientation”, explains Dr. Maxime Brami, who led the project at Mainz University. For archaeologists, this means they can use AutArch to automatically generate vast amounts of data, spread across many publications, to answer specific questions about the past and compare it, for instance, with 3D scans of artefacts in museum collection.
“Previously, researchers had to manually extract information from images, which takes a lot of time and involves tedious tasks like resizing, reorienting, and reformatting the images”, explains Kevin Klein, software developer at JGU and first author of the study. AutArch automates the entire process. Although it uses AI, the results are never black box. A user-friendly interface allows researchers to check and adjust the automatically extracted data, ensuring accuracy and accountability.
The software is widely applicable and scalable
AutArch is scalable and can serve the needs of the ever-growing field of digital humanities. Antoine Muller, a Palaeolithic researcher and one of the authors of the study, says “the methodology is applicable to virtually any material, as long as the shape, size, and/or orientation of an object holds technological, functional, or chronological significance”. Not only can it be applied to any material, but it also grows with increasing demands. “This development represents an important step forward in the application of artificial intelligence in archaeological research,” Brami summarizes. “It has the potential to fundamentally transform data access and analysis.” AlphaGalileo/SP