No escape: AI tracks individual chimps across video archives
n n
KYOTO–When scientists here set out to apply artificial intelligence to the study of animal behavior, they weren’t monkeying around.
rn
Researchers from Kyoto University and Oxford University have developed a face recognition model using AI technology that can identify and track chimpanzees in video footage with a high degree of accuracy.
rn
The deep learning technique, they believe, will provide a fully automated approach to analyzing vast amounts of archival data on animals and allow for more thorough analyses by ethologists.
rn
Deep learning is an AI system technology that uses an artificial neural network, allowing for automated “learning” of data.
rn
The study was published in the U.S. journal Science Advances on Sept. 5.
rn
Kyoto University established a field site for studying chimpanzees in Guinea, West Africa, in 1976 for observing groups of wild chimpanzees.
rn
Research there over the years has produced volumes of video footage capturing social behavior such as adults using stones as a tool and infants learning from them, but it has been impossible for researchers to manually process or conduct a large-scale longitudinal analysis.
rn
The face recognition model using AI technology was able to recognize the faces of 23 chimps across 50 hours of footage collected from 1999 to 2012, with an accuracy of 92.5 percent, compared with 20-42 percent for humans.
rn
“The research result has revealed a new way to conduct field research,” said Tetsuro Matsuzawa, distinguished professor of primatology at Kyoto University, who led the team.
n n