Researchers have developed a free, AI-driven app that can identify which species of ancient dinosaur left multi-million-year-old footprints still visible today.
The team behind the new dinosaur footprint identification app, which has successfully identified ancient species by their fossilized tracks with over 90% accuracy, said that they intentionally made it free and easy to use to entice professional paleontologists and amateur hobbyists alike.
The researchers also hope the app will help scientists discover previously unidentified dinosaur species and spark new interest in the history of these extinct, ancient reptiles.
“Our method provides an unbiased way to recognize variation in footprints and test hypotheses about their makers,” explained study author, Dr Gregor Hartmann of Helmholtz-Zentrum research centre in a statement announcing the app’s release. “It’s an excellent tool for research, education, and even fieldwork.”
Dinosaur Footprint App Fills Critical Identification Gap
In recent decades, discoveries of ancient dinosaur footprints have captivated both scientists and the public. Still, conclusively assigning these ancient tracks to a specific dinosaur species has proven challenging.
Some efforts to identify which animal left the multi-million-year-old track have pointed to fierce carnivores, while others have suggested gentle plant-eating dinosaurs as the source. There are even suggestions that some of these ancient footprints could have been left by early bird species.
According to the team behind the new dinosaur footprint identification app, current methods require researchers to “manually compile computer databases” of previously ascribed tracks. Because this method relies heavily on previous identifications, the team said experts believe it could “introduce bias.”
To fill the gap in dinosaur footprint identification, a team led by the Helmholtz-Zentrum research centre in Berlin, in partnership with the University of Edinburgh, has completed and released DinoTracker.
How Scientists Taught the App to Learn
According to the team’s published study, the AI model that drives DinoTracker was initially trained on a database of nearly 2,00 fossil footprints. Notably, the species that left the footprints in this database had been conclusively identified by other methods, providing the new AI model with a high level of starting-point accuracy. The model was also trained to evaluate millions of potential variations within these identified footprints, mimicking realistic changes such as edge displacement and compression.
After using advanced algorithms to allow the computers to further “train themselves” on the different features of each footprint in the database, the research team tested its ability to recognize minute variations in the tracks that could enable species identification.
As hoped, DinoTracker AI found numerous “key” features of footprint variation that could bridge the gap in identification. These included variations in the spread of the toes, the position of the heel, and the size of the contact area the foot made when impacting the ground. The AI was also able to differentiate the amount of weight the extinct animal placed on different parts of the foot when it left the fossilized tracks.
DinoTracker App’s Dinosaur Footprints Identification Test Reveals 90% Success
After successfully training their AI model, the research team tested its ability to identify dinosaur species from images of their footprints. According to the team’s statement, the app’s species classification base don a footprint image “achieved around 90 percent agreement” with previous classifications of the same footprints made by human experts. The researchers note that this level of accuracy remained high “even for contentious species.”
When the researchers analyzed the data, they found an intriguing link among several of the oldest tracks, made more than 200 million years ago. According to their statement, these similar tracks shared “uncanny features” with tracks made by both extinct and modern birds.
Although not conclusive, the team suggests these footprints may be evidence that birds evolved millions of years earlier than scientists previously believed. They also note that the DinoTracker’s findings could also mean that some early dinosaur species had feet that “coincidentally resembled those of birds to a high degree.”
Another unexpected finding made by DinoTracker involved a set of 170-million-year-old fossilized dinosaur footprints discovered on the Isle of Skye in Scotland. According to the team, these tracks may have been left behind by the oldest relatives of duck-billed dinosaurs “known from anywhere in the world.”
Solving Something That Has ‘Stumped Experts for Over a Century’
When discussing the implications of their dinosaur footprint identification app, the team said that accurately modeling the lives and movements of these ancient creatures requires knowing which species left the tracks found across several parts of the world.
“This study is an exciting contribution for paleontology and an objective, data-driven way to classify dinosaur footprints – something that has stumped experts for over a century,” explained Professor Steve Brusatte, Personal Chair of Palaeontology and Evolution, School of GeoSciences. “It opens up exciting new possibilities for understanding how these incredible animals lived and moved, and when major groups like birds first evolved.
“I think [DinoTracker] is a fantastic and fruitful use for AI,” the professor added.
For researchers and enthusiasts who have identified possible dinosaur footprints and would like to add them to the project, they can simply download the DinoTracker app and upload the image for an “instant analysis.”
“[DinoTracker] gives everyone the opportunity to become their own fossil footprint investigator,” the researchers concluded.
The study “Identifying variation in dinosaur footprints and classifying problematic specimens via unbiased unsupervised machine learning” was published in Proceedings of the National Academy of Sciences.
Christopher Plain is a Science Fiction and Fantasy novelist and Head Science Writer at The Debrief. Follow and connect with him on X, learn about his books at plainfiction.com, or email him directly at christopher@thedebrief.org.