Wildbook: The AI That Knows Every Whale Shark by Its Spots

Wildbook is an open source software platform from the nonprofit Wild Me, now part of Conservation X Labs, that uses computer vision and artificial intelligence to identify individual wild animals from photographs, and it matters because protecting a species starts with knowing how many individuals exist, where they go, and whether their numbers are rising or falling. Traditional wildlife population studies require researchers to physically capture, tag, or track animals, work that is slow, expensive, and often impossible at the scale of an ocean. Wildbook replaces the tag with the photograph, and the photograph can come from anyone.

The original Wildbook, built for whale sharks, has grown into a family of platforms covering species from manta rays to polar bears, each using the same core trick: treating the natural patterns on an animal's body as a fingerprint.

Spots as Fingerprints

Every whale shark wears a unique arrangement of spots behind its gills, and every scar, notch, and marking adds to the signature. The pattern recognition software behind Wildbook, first as the HotSpotter algorithm and later as a trainable system called PIE, analyzes these markings to match a new photograph against every animal previously catalogued. Wild Me explains that PIE can be trained on a per species basis, with separate optimized models for manta rays, humpback whales, orcas, right whales, and more.

The result is functionally a census built from images. When two photographs taken years apart and thousands of kilometers apart match the same individual, researchers learn something that no tracking study could afford to discover: migration routes, survival rates, and population connections across whole ocean basins.

Citizen Science at Ocean Scale

Wildbook's most radical feature is who supplies the data. Scuba divers, tour operators, and beachgoers upload photographs through web platforms like the whale shark Wildbook and Sharkbook, and the system identifies the animal or logs it as a new individual. A tourist's holiday snapshot, timestamped and geotagged, becomes a data point in an international research database.

The platform has even automated discovery. Microsoft, whose cloud tools Wild Me uses, described an intelligent agent running on the whale shark Wildbook that scans YouTube nightly for videos tagged with the species, uses machine learning to determine whether relevant footage is present, and reads video descriptions to estimate when and where the shark was sighted. Social media, in effect, becomes a distributed wildlife monitoring network.

Data That Moves Policy

Photo identification at scale produces population models, and population models drive protection decisions. Wild Me founder Jason Holmberg was lead author on widely cited studies of whale shark population trajectories at Ningaloo Marine Park in Western Australia, work built on the ability to track individuals across years using Wildbook data. Whale sharks, once abundant enough to be ignored, are now classified as endangered on the IUCN Red List, with population analyses contributing to the assessment.

The same pattern has repeated for other species on Wildbooks around the world, from zebras in Kenya to whale sharks in the Indian Ocean. When governments consider fishing limits, protected areas, or trade restrictions, the quality of the population evidence frequently determines the outcome.

Open Source for the Whole Field

Wild Me gives the platform away. The software is open source, and research groups and conservation organizations can deploy their own Wildbooks for the species they study. Dozens of Wildbooks now operate worldwide, each a collaboration between computer scientists and field biologists, all feeding the same philosophy: the tools of machine learning, so often used to sell advertising, can be pointed at the problem of extinction.

The next time a dive boat full of tourists photographs a whale shark, there is a decent chance the encounter ends up in a database where an algorithm glances at the spots behind the gills and says, in effect, we know this one. Her name is in the record. She was last seen two years ago, eight hundred kilometers away. And because of that, she counts.