Scientists Launch DinoTracker, an AI App That Identifies Dinosaur Footprints in Seconds

AI Transforms How Scientists Study Ancient Tracks

A collaboration between Helmholtz-Zentrum and University of Edinburgh has produced DinoTracker, an AI-powered app that can identify dinosaur footprints in moments by analyzing photos or sketches uploaded through a mobile phone 1

1

2

2

. The research, published in PNAS, addresses a challenge that has stumped paleontology experts for over a century: determining which species left behind the fossil footprints scattered across ancient landscapes 2

2

5

5

.

Dinosaur footprints are the most common type of dinosaur fossil known, far outnumbering skeletal remains. Each dinosaur could only leave one skeleton but could make thousands of footprints in a single day . Yet these tracks prove notoriously difficult to interpret because their shapes reflect not just the foot structure but also how the dinosaur was moving, the softness of the ground, and millions of years of geological transformation 1

1

4

4

.

Removing Human Bias from Fossil Analysis

Gregor Hartmann, who led the project at Helmholtz-Zentrum, explained that the team sought to remove subjectivity from the identification process by developing an algorithm that could remain neutral 1

1

. Previous AI systems relied on manually built computer databases where researchers assigned tracks to specific dinosaurs, a step that could introduce human bias, especially when track identity was uncertain or disputed 2

2

4

4

.

The breakthrough came from using an unsupervised neural network approach. Researchers trained the algorithm on nearly 2,000 real fossil footprints, along with millions of simulated variations designed to recreate natural distortions such as compression and shifting edges 1

1

2

2

. Critically, the system received no labels identifying whether tracks came from theropod or ornithopod dinosaurs during training 1

1

.

Eight Key Features Drive Track Recognition

The AI system independently discovered eight key features that distinguish one footprint from another, including how far the toes spread, where the heel was positioned, how much surface area contacted the ground, and how weight was distributed across different parts of the foot 2

2

4

4

. After identifying these variations, the system compared new footprints with known fossil examples to predict which dinosaur most likely made the tracks 2

2

.

When evaluated against human expert classifications, the algorithm reached 80-93% agreement overall, including for species considered controversial or difficult to identify 4

4

. Steve Brusatte, Personal Chair of Palaeontology and Evolution at the University of Edinburgh and co-author of the work, called it "an objective, data-driven way to classify dinosaur footprints" 2

2

.

Unexpected Clues About Bird Evolution

One of the most surprising findings emerged from tracks more than 200 million years old from the Triassic and early Jurassic periods. The AI detected striking similarities between some dinosaur footprints and the feet of both extinct and modern birds 1

1

2

2

. These tracks are approximately 60 million years older than the oldest bird skeletons, the fossilized bones of Archaeopteryx 5

5

.

According to the research team, this could mean that bird evolution occurred tens of millions of years earlier than scientists have previously believed 1

1

4

4

. However, Brusatte cautioned that another possibility exists: some early meat-eating dinosaurs may have had feet that coincidentally looked very similar to bird feet 5

5

. Hartmann noted that footprint evidence alone isn't enough to rethink bird evolution, since a skeleton remains the "true evidence" of earlier bird existence 1

1

.

New Insights from Scottish Tracks

The system also offered fresh analysis of mysterious footprints found on the Isle of Skye in Scotland. These tracks were formed on the muddy edge of a lagoon around 170 million years ago and have puzzled scientists for decades 2

2

. The analysis suggests that these footprints may have been left by some of the oldest known relatives of duck-billed dinosaurs, making them among the earliest examples of this group identified anywhere in the world 2

2

4

4

.

Opening Paleontology to Public Participation

DinoTracker is available for free on GitHub and as a mobile app, allowing anyone to upload a picture of a dinosaur footprint, sketch its outline, and receive instant analysis 1

1

. Users can explore the seven other footprints most similar to their upload and manipulate the footprint to see how varying the eight features affects which other footprints are deemed most similar 5

5

.

The technology creates new opportunities to study how dinosaurs lived and moved across Earth while giving the public a chance to participate in fossil research 2

2

. In research settings, it could help screen large numbers of tracks quickly and identify patterns across sites. For fieldwork, it offers a fast way to test hypotheses on the spot 4

4

. Hartmann emphasized that the hope is for DinoTracker to be used by paleontologists and for the AI tool's data pool to grow as it's used by more experts 1

1

.