Scientists have created an AI-powered app called DinoTracker, which uses artificial intelligence to identify dinosaur footprints based on their unique characteristics. The innovative tool analyzes the shape and features of footprint silhouettes to determine how similar or different they are from one another.
A team of researchers, led by Prof Steve Brusatte from the University of Edinburgh, fed 2,000 unlabelled footprint silhouettes into the AI system, which was trained on a range of meaningful features such as toe spread, ground contact, and heel position. The resulting eight features revealed variations in the footprints' shapes that experts had previously overlooked.
Using DinoTracker, users can upload the silhouette of a footprint, explore seven other similar footprints, and manipulate the footprint to see how changing its characteristics affects which other footprints are deemed most similar. This allows for a more accurate identification of dinosaur species based on their tracks.
One notable finding is that the AI system supported the theory that certain bird-like footprints from the Triassic period show remarkable similarities with those of modern birds, such as Archaeopteryx. According to Prof Brusatte, this suggests that birds have an even deeper ancestry than previously thought, dating back tens of millions of years.
However, the researchers caution that their findings do not necessarily prove the existence of early bird species. Instead, it is likely that these tracks were made by meat-eating dinosaurs with bird-like feet. Dr Jens Lallensack from Humboldt University of Berlin noted that a key limitation of the system was that its features of interest were based on the way the foot sank into soft ground, rather than its actual shape.
Despite this limitation, DinoTracker represents an exciting step forward in using AI to analyze and understand dinosaur tracks. As Dr Hartmann from Helmholtz-Zentrum Germany said, "You never find a footprint without knowing which dinosaur made it," but with the help of machines like DinoTracker, researchers can make more accurate connections between footprints and their corresponding species.
A team of researchers, led by Prof Steve Brusatte from the University of Edinburgh, fed 2,000 unlabelled footprint silhouettes into the AI system, which was trained on a range of meaningful features such as toe spread, ground contact, and heel position. The resulting eight features revealed variations in the footprints' shapes that experts had previously overlooked.
Using DinoTracker, users can upload the silhouette of a footprint, explore seven other similar footprints, and manipulate the footprint to see how changing its characteristics affects which other footprints are deemed most similar. This allows for a more accurate identification of dinosaur species based on their tracks.
One notable finding is that the AI system supported the theory that certain bird-like footprints from the Triassic period show remarkable similarities with those of modern birds, such as Archaeopteryx. According to Prof Brusatte, this suggests that birds have an even deeper ancestry than previously thought, dating back tens of millions of years.
However, the researchers caution that their findings do not necessarily prove the existence of early bird species. Instead, it is likely that these tracks were made by meat-eating dinosaurs with bird-like feet. Dr Jens Lallensack from Humboldt University of Berlin noted that a key limitation of the system was that its features of interest were based on the way the foot sank into soft ground, rather than its actual shape.
Despite this limitation, DinoTracker represents an exciting step forward in using AI to analyze and understand dinosaur tracks. As Dr Hartmann from Helmholtz-Zentrum Germany said, "You never find a footprint without knowing which dinosaur made it," but with the help of machines like DinoTracker, researchers can make more accurate connections between footprints and their corresponding species.