Andrea Jalandoni knows all too well the challenges of archaeological work. As a senior researcher at Griffith University’s Center for Social and Cultural Research in Queensland, Australia, Jalandoni dodged crocodiles, climbed limestone cliffs and sailed traditional canoes through shark-infested waters, all to survey important sites in the Pacific, Southeast Asia and Australia. One of its biggest challenges is modern: analyzing the exponential amounts of raw data, such as photos and tracks, collected from sites.
“Manual identification takes too much time, money and specialist knowledge,” says Jalandoni. She ditched her trowel years ago in favor of more advanced technologies. His toolbox now includes several drones and advanced imaging techniques to record sites and discover things not apparent to the naked eye. But to make sense of all the data, she had to use another cutting-edge tool: artificial intelligence (AI).
Jalandoni has teamed up with Nayyar Zaidi, a lecturer in computer science at Deakin University in Victoria, Australia. Together they tested machine learning, a subset of AI, to automate image detection to aid rock art research. Jalandoni used a photo dataset from Kakadu National Park in Australia’s Northern Territory and worked closely with First Nations elders in the area. Some results of this research were published last August by the Journal of Archaeological Sciences.
Kakadu National Park, a Unesco World Heritage Site, contains some of the best-known examples of painted rock art. The works are created from pigments made from iron-stained clays and iron-rich ores that have been mixed with water and applied using tools made from human hair, reeds, feathers and chewed sticks. Some of the paintings in this area date back 20,000 years, making it one of the oldest arts in recorded history. Despite its world-renowned status for rock art, only a fraction of the works in the park have been studied.
“For First Nations peoples, rock art is an essential aspect of contemporary Indigenous cultures that directly connects them to ancestors and ancestral beings, cultural histories and landscapes,” says Jalandoni. “Rock art is not just a given, it is part of Indigenous heritage and contributes to Indigenous well-being.
An example of artificial intelligence extracting a figure from a rock art photo Courtesy of Andrea Jalandoni
For the AI study, the researchers tested a machine learning model to detect rock art from hundreds of photos, some of which showed painted rock art images and others with rock surfaces. naked. The system found the art with a high degree of accuracy of 89%, suggesting that it may be invaluable for evaluating large image collections from heritage sites around the world.
Beyond Detection
Image detection is just the start. The ability to automate many stages of rock art research, coupled with more sophisticated analysis, will accelerate the pace of discovery, Jalandoni said. The systems trained must be able to classify images, extract patterns and find relationships between different elements. All of this will lead to a deeper knowledge and understanding of the images, stories and traditions of the past.
Eventually, AI systems could be trained for more complex tasks such as identifying works by individual artists or virtually restoring lost or defaced works.
This is important because time is essential to many ancient forms of art and storytelling. In areas where many rock art sites exist, much of it is often unidentified, unrecorded and unstudied, Jalandoni says. And with climate change, extreme weather events, natural disasters, encroaching development, and human mismanagement, this inherently finite art and culture form will continue to grow more vulnerable and rarer.
Jannie Loubser, rock art specialist and cultural resource management archaeologist with conservation group Stratum Unlimited, sees another important use of AI in conservation and preservation. Trained systems will help monitor imperceptible changes in surfaces or conditions at rock art sites. But, he adds, the “truth on the ground”—standing face-to-face with the work—will always be important to understanding a site.
Jalandoni agrees that there is nothing quite like studying in person works created by artists thousands or tens of thousands of years ago and trying to understand and acknowledge the history told. But she sees great potential in combining her new and old tools to explore and document hard-to-reach sites.
Martin Puchner, author of Culture: Our story, from rock art to K-Pop (2023), sees a poetic resonance in the use of AI, the most contemporary of tools, to reveal the past.
“Even though we are heading into the future, we are also discovering more about the past, sometimes through accidents when someone discovers the cave, but also, of course, through new technologies,” says Puchner.
