
Monkey see, monkey recognise. That’s what an experimental app is offering to do for conservationists seeking to identify and track primates in the wild. It could even help wildlife crime investigators recognise individuals that have been killed or trafficked.
While some researchers in the field are able to identify individual primates in the small populations they are studying, recognising them quickly in other contexts is very difficult, says at Liverpool John Moores University, who was not involved in the work.
“We put camera traps out or we take pictures and we don’t have an idea if it’s the same individual all the time or different ones,” he says. This is unlike other species, such as tigers, which are more easily distinguishable by their markings.
Advertisement
It was because of this problem that primatologists first contacted and colleagues at Michigan State University. The team was tasked with building a mobile app that could be used to identify primates just from photos of their faces.
The race to find faces
However, the first challenge they encountered was a lack of data. “There’s no shortage of faces to train human face recognition systems,” says Jain, “but if I want to build a lemur face recognition system there are a limited number of lemur images.”
Over a few days, Jain’s student Debayan Deb took photographs of the primates at to produce a dataset of 3,000 images. The team also gathered thousands of pictures of golden monkeys and chimpanzees from conservationists.
A neural network learned to distinguish facial features of the primates and it proved particularly good at recognising lemurs. It achieved over 80 per cent accuracy – even when it was required to deduce that it could not identify a particular lemur because that individual had not been included in the dataset the system was trained on.
The scores for chimpanzee recognition were lower, despite the team having assembled an even larger collection of images. This was due to the poorer quality of those photographs, says Jain.
“It’s fantastic that they’re trying to do this and that the results look promising,” says Wich. Including orangutans in the app would be useful as the faces of males change dramatically as they age, he adds.
Journal Reference: