AI DinoTracker app uses machine learning to identify dinosaur footprints with near-expert accuracy

• Photo from Pixabay

Scientists are turning artificial intelligence into a new tool for palaeontology, launching an AI-powered app that can identify dinosaur species based on ancient footprints left behind tens of millions of years ago.

The system, now available as a free app called DinoTracker, was developed by an international research team led by scientists from the University of Edinburgh and Helmholtz-Zentrum. 

According to the researchers, the AI’s classifications align with those of human experts about 90 percent of the time.

Rather than relying on footprints that were already labelled by scientists—a method that risks inheriting past errors—the team took a different approach. 

They trained the system using around 2,000 unlabelled footprint silhouettes, allowing the AI to independently assess similarities and differences based on patterns it identified on its own.

Dinosaurs are archosaurian reptiles belonging to the clade Dinosauria that lived primarily during the Mesozoic era and are characterized by an upright limb posture; birds are their only living descendants.

“When we find a dinosaur footprint, we try to do the Cinderella thing and find the foot that matches the slipper,” said Prof. Steve Brusatte, a co-author of the study. “But it’s not so simple,” he noted, explaining that footprint shape is influenced not only by the dinosaur’s anatomy but also by ground conditions and movement.

Users of the app can upload a footprint silhouette, view the seven most similar tracks in the database, and even manipulate individual features to see how subtle changes affect classification results. 

Paleontology is a branch of science that explores the history of life on Earth through the study and interpretation of fossils—preserved remains or traces of ancient plants, animals, and microorganisms embedded in rock layers.

Writing in the Proceedings of the National Academy of Sciences, Brusatte and his team explained that earlier AI models were trained using dinosaur footprints that had already been classified as belonging to specific species. 

The researchers noted, however, that if those original classifications were inaccurate, the resulting AI analyses would inherit the same errors.

“You never find a footprint and alongside [it] the dinosaur that had made this footprint,” said Dr Gregor Hartmann, the first author of the new research from Helmholtz-Zentrum in Germany. “So, no offence to palaeontologists and such, but most likely some of these labels are wrong.”

According to Hartmann, the tool still requires expert oversight, especially to confirm factors like geological age and sediment type, but its clustering closely mirrors established scientific groupings.

However, not all experts are fully convinced. Dr. Jens Lallensack of Humboldt University of Berlin, who has previously applied AI in the analysis of dinosaur footprints but was not part of the research team, pointed out an important limitation of the new approach. He noted that the system may be focusing on features that are not directly tied to the actual shape of the animal’s foot.

According to Lallensack, the bird-like appearance of some tracks could simply reflect how a typical theropod foot pressed into soft sediment. “They are not evidence for an early appearance of birds,” he said.

SOURCE: The Guardian