- Machine learning and citizen science combine to protect wild animals
- Accurate data on species population necessary for conservation efforts
- Animal data must be securely managed to deter poaching
The 7.6 billion humans living on planet Earth consume an enormous amount of natural resources. As we drill for oil, clear forests for livestock, and alter the climate with carbon emissions, we push many other species toward extinction.
But could we now use the science and technology that have improved living conditions for humans to slow or reverse the harm we’ve done to the animal kingdom?
Tanya Berger-Wolf, a computational ecologist at the University of Illinois at Chicago, believes we can. Wildbook, created by Berger-Wolf and a team of passionate researchers and engineers, combines the power of citizen science, computer vision, and machine learning to identify and track animal populations in the wild.
To preserve the planet’s biodiversity, conservationists need accurate information about wild animal populations. Data on the distance traveled by whales or the survival rate of turtle hatchlings is not easy to collect or verify. Scientists can’t attach GPS devices and tracking tags to all of the world’s animals. But with Wildbook there’s an easier way.
The result of collaboration between Berger-Wolf, computer scientist Charles Stewart of Rennselaer Polytechnic, data architect Jason Holmberg, and Princeton evolutionary biologist Dan Rubenstein, Wildbook was deployed for the first time in January 2015 in Kenya in partnership with the Lewa Wildlife Conservancy and Wildlife Direct.
Volunteers and researchers in 27 cars with 55 cameras collected nearly 10,000 photographs of the common plains zebras and Masai giraffes in Nairobi National Park in just 2 days. This rapid acquisition of photos was made possible by crowdsourcing the task. The researchers then used Wildbook's computer vision algorithms to identify all the animals recorded and match them against images of known animals in its database.
The algorithms work by comparing the stripe or spot patterns in the new images with those of animals in the database. This census-taking, dubbed the Great Zebra and Giraffe Count (GZGC), showed that Wildbook could be effectively applied at scale and across large geographical areas.
“People engage in this domain easily because the price of entry is so low. All you have to be able to do is take a picture,” says Berger-Wolf. “We’re finding very willing citizen scientists and volunteers.”
In January 2016, a second effort, the Great Grevy's Rally (GGR), collected over 40,000 photographs with the help of 200 people including school children, tribal chiefs, and an ambassador to Kenya.
The Grevy’s zebra is an endangered species whose numbers were previously counted by wildlife service officials driving in a limited area and counting any zebras they saw. The population estimates resulting from Wildbook and GGR are the most accurate to date. The IUCN Red List of Threatened Species now uses Wildbook data on the Grevy’s zebra as its official numbers.
Today, there are over a dozen active installations of Wildbook including Giraffespotter, MantaMatcher for tracking manta rays, and the Polar Bear Library. Wildbook’s open source software is free to download. Researchers who need hosting or customization of the software can get help from Wildbook’s non-profit, Wild Me.
But there is a downside to maintaining large-scale collections of accurate population and geographic data on wild animals. Poachers could hack the systems and use the GPS-tagged data to find and kill the animals conservationists are trying to protect.
Berger-Wolf says that having the image of an animal along with its location and the time the photo was taken is great for scientists, but also “golden for poachers and wildlife criminals.” She adds, “By building Wildbook, I do not want to be responsible for driving an animal to extinction.”
To prevent this abuse, Wildbook has partnered with Trusted CI to develop an Identity Management and Access Control framework to ensure that Wildbook’s data doesn’t fall into the wrong hands. The Wildbook team is also collaborating with Ross Anderson, a world leader in cybersecurity research from the University of Cambridge, to write a policy for animal privacy.
A new project in the works for Berger-Wolf is a partnership with the World Wildlife Fund (WWF) known as IoT, or the Internet of Turtles. It’s an effort to study turtle populations without resorting to the traditional flipper tagging method. Another upcoming challenge is finding a way to identify individual elephants even when the animals are obscured by the elements. The researchers hope to learn how to adapt algorithms used for one species to also identify others.
Thanks to Wildbook, the future of the planet’s animals is looking brighter, but time is running out.
“If we don’t do anything, we’re going to not have elephants by 2030,” says Berger-Wolf. “That’s not that far into the future.”
“If we can slow that down and engage people in conservation and protection rather than poaching, that’s huge. It’s just amazing to me that we can do all that with just a few thousand lines of code."