Innovation

Facial-recognition software may help save Earth's most endangered mammals

lemur

 (Natalee Phelps)

Lemurs are in trouble. In 2012, the International Union for Conservation of Nature listed the Madagascar natives as the most endangered mammals on the planet. But a team of researchers have developed a system called LemurFaceID, which uses facial-recognition software to spot lemurs in their natural habitat.

"The original inspiration for developing LemurFaceID was a desire to develop a noninvasive tool that would help us ID and track lemurs," Stacey Tecot, University of Arizona assistant professor and senior researcher of the project, told Digital Trends.

To minimize invasiveness, Tecot and her colleague, George Washington University's Rachel Jacobs, decided not to capture or tag their subjects, but soon found that such a hands-off approach made collecting sufficient datasets difficult. "I'd explored using dye via a gentle water gun, but didn't get very far with that," Tecot said. "Another researcher had used photographs but I didn't have the resources for a good camera, and my photos weren't useful."

Jacobs, on the other hand, managed some success using photographs to double-check her team's visual assessments. "Because of that success, automating lemur identification seemed possible, and we began to explore the different biometric tools out there. We kept coming back to facial recognition. Connected with Anil Jain and his students gave us the confidence that this could work."

Along with his students, Jain, who teaches biometrics at Michigan State University, used hundreds of lemur photos that were captured by Tecot, Jacob, and their team, feeding the images into facial-recognition software until the system was able to identify the 80 individual red-bellied lemurs depicted in the images.

"We hope the system will help inform the study and conservation of lemurs in several ways," Tecot said.

First, the system may be used to train researchers, assistants, and technicians who are new to studying lemurs. "It's not a replacement for learning IDs for long-term research," Tecot said, "but it can be a good check that a lemur is actually who we think it is during data collection."

Second, LemurFaceID can help keep track of populations, much like a census, by spotting individuals new to the group or identify lemurs that have been trafficked for sale elsewhere. "Part of conservation is determining population viability and status," she added. "We plan to test this system in the future alongside traditional census methods, and reassess the conservation status of the species."

Finally, the system may even be used to engage the public in conservation efforts by making the lemurs that much more relatable. "Tourist guides and visitors to the park … could have a more personal connection with the lemurs, and the park and wildlife more generally, by either contributing to the database and helping us keep track of the population, or snapping a photo and learning more about the individuals," Tecot said.

The researchers have published a paper detailing their work in the journal BioMed Central Zoology.