There is now a new tool that can be used to fight the illegal trade in chimpanzees. The same facial recognition software that social media sites use is now being adapted to recognize trafficked animals, and this algorithm will be used to find the faces of these apes online.
For an entire year, the BBC investigated the smuggling of chimpanzees and found that they often end up as performers in commercial zoos or as pets to wealthy owners. Baby chimps are so popular, that they can be sold for as much as $12,500.
The idea surged after conservationist Alexandra Russo was investigating online ape trafficking and came across posts on social media sites such as Facebook and Instagram where chimpanzees were being offered for purchase.
Related: These AI-powered cameras can sense poachers and save wildlife
“We were spending more and more time looking through the depths of the internet; it’s like a rabbit hole. You don’t know where to look, you click around pretty aimlessly until you start to find things that look suspicious. So, I thought there must be a more efficient way to do this,” she explains. “I began discussing the possibility of using some kind of software that could automatically find ape faces in online searches.”
Russo then contacted Conservation X Labs, a non-profit group, and met with Dr. Colin McCormick, who is an expert in computer vision. The two ended up creating “ChimpFace.”
The software works by detecting images that show a chimpanzee, and then it identifies the individual. They trained the algorithm to recognize different chimps using facial structures. Russo says that it is important to get as many images as possible of different positions, facial expressions and altered lighting.
The algorithm uses a database of 3,000 different ape face images, and different chimpanzee conservation groups have contributed photos.
Experts say that this new technology could help bring down large-scale criminal networks. Facebook and Instagram are already starting to shut down wildlife trafficker accounts.
Image via PublicDomainPictures