WildTrack AI Reads Animal Footprints to Protect Endangered Species

WildTrack, a nonprofit founded in 2004 by zoologists Sky Alibhai and Zoe Jewell, has built an AI system that identifies endangered animals from nothing more than a photograph of a footprint. Its Footprint Identification Technology, or FIT, classifies the species, individual, sex, and age class of an animal with more than 90 percent accuracy. That matters because traditional monitoring methods either depend on scarce professional biologists or require capturing animals and fitting them with collars, tags, or implanted sensors, a process that is expensive, limited to a few individuals, and can be physically and socially harmful to wildlife.

FIT is the only end to end platform for monitoring endangered species using footprints, according to the organization. Data is cheap to collect, accessible to non-experts, and completely non-invasive. A ranger on patrol with a phone can contribute exactly the kind of data that once required a dart gun and a helicopter.

Ancient Skills, Modern Algorithms

The idea behind FIT grew out of years spent working with indigenous trackers, particularly the Ju'hoansi of the Kalahari, whose skills at reading footprints inspired a formal science. Where a human tracker reads subtle contours in sand or mud, FIT extracts dozens of measurements and shape features from a digital image of a footprint and feeds them into customized mathematical models that match the print against a reference database.

The pipeline works in stages. First the model classifies which species made the print. Then it narrows down the sex and age class. Finally, at the finest grain, it attempts individual identification, effectively asking whether this specific cheetah or rhino has been seen before. WildTrack has reported individual identification accuracy around 95 percent, and broader classification accuracy above 90 percent across the species it studies.

From Rhinos in Namibia to Tigers and Polar Bears

To date, WildTrack has developed provisional or full FIT algorithms for at least 13 species. The list includes black and white rhino, Bengal and Amur tiger, cheetah, puma, lowland and Baird's tapirs, and polar bears. Field projects have spanned multiple continents, from monitoring black rhinos with the Namibian government to identifying giant pandas in China with collaborators at Duke University and Duke Kunshan University, where footprint images were used to determine individual identity and sex.

The cheetah work illustrates why the technology is needed. The cheetah population has declined by 93 percent over the past century, and the species has disappeared from 76 percent of its former range in Africa. Cheetahs rarely prey on domestic animals, yet many are killed on the assumption that they threaten livestock, and smuggled cheetah cubs face brutal odds, with only one in six surviving the journey out of Africa. Reliable population data, gathered cheaply from footprints, gives conservationists the evidence base to push back.

Scaling Up With SAS and AWS

The early versions of FIT leaned on statistical models built with analytics firm SAS. Deep learning capabilities were added to recognize footprint images and patterns in ways that mirror indigenous trackers, but at far larger scale and pace. More recently, WildTrack has worked with Amazon Web Services, whose cloud infrastructure underpins a modern AI pipeline that ingests footprint images from the field and processes them at volume.

The volume question is central. Through its ConservationFIT initiative, WildTrack aims to crowdsource footprint images from rangers, researchers, and communities around the world. As the organization puts it, ordinary people cannot dart a rhino, but they can take an image of a footprint. That opens conservation data collection to a far wider group than professional biologists, and it is precisely when the data inflow becomes too large to manage traditionally that machine learning earns its keep, performing high-volume classification reliably and without fatigue.

The scientific method is advancing alongside the engineering. A 2026 paper in Scientific Reports examined wildlife track classification through a human-in-the-loop method paired with explainable AI, a design that keeps expert trackers in the loop and makes the model's decisions interpretable. The study noted FIT applications spanning Asian otters, black rhinoceros, cheetah, and white rhinoceros, among others.

Why Non-Invasive Monitoring Matters

The standard alternatives to footprint monitoring each carry costs. Trail cameras can capture images along game trails but are limited by expense and by the difficulty of detecting animals at low densities. Acoustic monitoring records and identifies animal calls but covers restricted areas because of cost and the technical expertise required. Environmental DNA can generate rapid species-level assays but depends on a small number of specialist laboratories. Collaring and tagging, meanwhile, can only reach a handful of individuals and carries documented physiological and behavioral risks.

Footprints solve several problems at once. The substrate is free and everywhere. The data collection requires a camera rather than a capture team. The animal never knows it has been counted. And because the skill is rooted in tracking traditions that stretch back millennia, the approach offers indigenous communities a way to have their expertise recognized, valued, and applied at global scale, something WildTrack describes as a restorative model for conservation.

The Road Ahead

WildTrack's bet is that the next phase of conservation monitoring will be crowdsourced and AI-assisted. Every protected area with sandy roads and motivated rangers can become a data collection site. As reference databases grow and models improve, footprint-based census work could give conservationists population estimates for species that are currently counted by guesswork or expensive aerial survey.

A phone. A footprint in the sand. A model trained on the accumulated wisdom of indigenous trackers. This is conservation technology at its most elegant, replacing the dart gun with the camera and scaling human knowledge that took thousands of years to accumulate.