Digging into and re-using image data for archaeology

Christopher Power, Andrew Lewis, Helen Petrie, Julian Richards, Katie Green, Mark Eramian, Ekta Walia, Brittany Chan, Isaac Sijaranamual, Maarten de Rijke


Each year thousands of archaeological field studies are undertaken. One of the resources with the largest potential to help archaeologists in their work is the hundreds of photographs that are taken during field studies. These photos, which can now be labelled with captions and uploaded to repositories direct from the field, could open new possibilities for re-use in many research tasks. Unfortunately, the potential of these resources has not yet been realised. Due to time pressure and lack of personnel, most photos do not have appropriate content-related metadata associated with them. While it is possible to identify what collection an image is from, or where it was taken using GPS coordinates, knowing what is actually in the image is often impossible. Even if tools were available to provide such content-related metadata, it is unlikely that the person power would be available to provide this metadata for the thousands of photos that already exist in digital form. The DADAISM Project is addressing this issue by using a mixed-initiative approach, where the deep domain knowledge of the archaeologist can be used to identify a number of key features in an image, and then automated processing can identify images that are similar, digging into the image data and extracting relevant information from the content. These identified images can then be re-used by archaeologists for their research, or even automatically labelled with appropriate content-related metadata. The new data created from this labelling can then be published to improve the robustness of searches by other archaeologists during their research. This paper will present preliminary results from the DADAISM Project on the identification of images in two specific archaeological domains, flint tools and Anglo-Scandinavian brooches, and will also present the interactive system to enable archaeologists to work with the DADAISM image identification system.