Tech

Conservation and livestock tracking could redeem facial recognition

Of course, there’s still a risk involved.

A grizzly bear sits in the grass looking toward the camera. A box and triangle are overlaid on their face.
BearID Project

Facial recognition for humans is a hot mess, but there’s still hope for this impressive technology. Instead of giving corporations and law enforcement agencies ways to surveil us, many conservationists and those in the animal husbandry world are looking to track animals with the tool, according to CNN. By using noninvasive cameras, the potential to save species and monitor diseases is limitless.

From the forest to the farm — Bear biologist Melanie Clapham and developers Mary Nguyen and Ed Miller created BearID after connecting in an online community for conservationists and technologists. Using 4,674 images of grizzly bears Brooks River and Knight Inlet in British Columbia, Canada and a modified AI that traditionally adds mustaches to dogs, the team is able to recognize individual bears 84 percent of the time.

Meanwhile, in Kansas, Joe Hoagland is building CattleTracs with a team from Kansas State University to help different parts of the animal husbandry chain monitor cattle. The app will allow users to take dated, geolocated pictures of cattle to establish its database. Future photos of a cow will be able to get matched with earlier photos. The initial AI model used 135,000 images of 1,000 young beef cattle with a 94 percent successful identification rate.

Both teams prefer their AI systems over traditional tracking methods like RFID tags, which are invasive and unreliable when close to other tags. BearID uses a system of camera traps, so the bears don’t even have to deal with the stress of humans. Oh, to be a grizzly bear.

A building trend — These projects don’t exist in a vacuum; those in animal wellbeing-oriented fields are making strides with facial recognition. RadioLab’s Latif Nasser explored various facial recognition applications in an episode of his Netflix show Connected, including health tracking at a pig farm in Scotland. Meanwhile, a Chinese company hopes to tackle livestock monitoring in general, and Alphabet’s Tidal is helping fish farmers work more sustainably. Rescue organizations are even working to build a database of pets around Atlanta to help owners find their lost furry friends.

The catch — The main concern about facial recognition lies with wild animals, but researchers are also having a tough time telling pigs apart. Animals that travel in packs and/or are often spotted when obscured by foliage can make it difficult to add useful photos to databases. It’s also possible for poachers to take advantage of these databases to track future trophies. Basically, no matter your standing in the animal kingdom, geolocation features could be the death of us all.