Based in Queensland, Australia, the Dinosaur Storm National Monument is known for its website dinosaur hurry. These footprints suggest that a large predator of 6 meters long was stalking 150 other dinosaurs the size of a chicken. They would belong the australian oven value tyrannosauropus.
These footprints were first discovered in the 1970s. Recently, Dr Anthony, a palaeontologist from the University of Queensland, reanalyzed the tracks dinosaurs using artificial intelligence (AI). He discovered that what was considered a predator is in fact a herbivore.
The research has been published in Journal of the Royal Society interface and includes collaborations between Australian, German and British researchers.
AI assistance was essential in identifying the species
In this study, the researchers used an AI program called “Deep Convolutional Neural Networks”. It is a trained machine learning model with over 1500 kinds of dinosaurs theropod or ornithopod type. So the algorithm is qualified to identify the fingerprints of the Dinosaur Storm National Monument. The results showed that the dinosaur in question was a herbivorous ornithopods.
That explains the lead author of the study AI assistance was essential to get his team out of trouble.
“In our three-person research team, one person supported carnivores, one person doubted and one person supported herbivores. To verify our science, we therefore decided to ask five experts for clarification and use AI. The AI turned out to be the big winner, outperforming all the experts by a margin of about 11%. »
Dr. Jens Lallensack, lead author from Liverpool John Moores University in the UK
A promising recognition method
Various factors are of influence the shape of the trackssuch as the anatomy of the animal, its behavior or the nature of the surface it has walked on. Weathering also plays a major role in their state of preservation. To analyze the anatomy of these types of shapes, scientists use qualitative methods rigorous.
This learning model makes it possible overcome the limitations of previous qualitative methods. It is also able to adapt to a huge complexity of data and can cover different application areas.
The team hopes to further enrich and supplement the database of fossil dinosaur tracks conduct further research using this technology.