Tracking animals with facial recognition
BearID isn’t the first conservation project to track wildlife using facial recognition – and, given its success, it won’t be the last. Here are three other projects that are turning the controversial tech into a tool to protect nature.
Chimpanzees
Researchers at the University of Oxford and Kyoto University trained facial-recognition AI to identify chimpanzees in Guinea, West Africa, using 50 hours of footage of 23 primates. The system was accurate 92% of the time, but to further test the system, the team used it on 100 still images that were also given to human researchers. In that trial, the facial recognition took 30 seconds to accurately identify the chimpanzees 84% of the time versus the humans who only managed a 42% success rate – and took 55 minutes to lose that badly. It’s hoped that the system will work on other primates and help researchers understand how social networks change in groups of the animals. The researchers followed up the facial-recognition system with a neural network that can identify animals from their body alone, helpful for when the primates turn their backs on the camera.
Lions
Kenyan conservation group Lion Guardians has developed a database of feline images to track specific individuals. Dubbed LINC – or Lion Identification Network of Collaborators – the project captures photographs of the big cats and analyses them with facial recognition to track their movements. While that’s much easier than sedating a lion and fitting it with a GPS transmitter, the photos still have to be taken from within 100ft. LINC looks at two areas of a lion’s face to identify it: its whisker patterns and its facial features. The data is helpful to a variety of wildlife groups, as lions have a massive range, travelling across borders and through different conservation areas. Find out more at linclion.org.
Salmon
Monitoring endangered species via facial recognition makes sense, but Norwegian fisheries giant Cermaq is trialling a similar idea to watch over the health of salmon on its farms. The BioSort system separates each and every fish of 150,000 in a farm into a chamber when they come up to the surface, where a 3D sensor scans its face for unique markings. The aim is to improve the health of fish, so if the camera also spots diseases such as sea lice, the specific fish will be quarantined for treatment. The aim is to reduce unnecessary handling of fish, which Cermaq says can cause stress for the animals. Find out more at cermaq.com.