KYOTO – Researchers from Kyoto University and the University of Oxford said they have succeeded in recognizing the faces of wild chimpanzees with an accuracy of over 90 percent using artificial intelligence.
Kyoto University conducts research on wild chimpanzees in a field site in Guinea in West Africa set up in 1976.
Using the AI system’s deep-learning methods, researchers analyzed a total of 50 hours of footage recorded between 2000 and 2013 by three cameras at the site on the top of a small mountain.
The system detected some 10 million images of chimpanzee faces from the footage, according to the team, whose study was published Thursday in an electronic edition of the U.S. journal Science Advances.
The system improved its face-recognition ability as time advanced, recognizing the identities of 23 individuals with an accuracy of 92.5 percent. For sex recognition, accuracy reached 96.2 percent.
Chimpanzee faces were recognized accurately even if they were captured only partially in analyzed images. Analysis of face and other images found that two female chimpanzees came to live away from their group as they aged.
Tetsuro Matsuzawa, a distinguished professor at Kyoto University, said AI systems can process amounts of data too vast for humans to handle.
“The new analytical method for field research has clearly shown changes in a society of chimpanzees,” he said.