Introducing AI that identifies the faces of bears and cows, promising in wildlife protection and livestock tracking
by
In recent years, the accuracy of AI that recognizes and identifies people's faces has improved, and it is used everywhere, such as unlocking smartphones and monitoring by law enforcement agencies. Face recognition by AI does not seem to be a proprietary patent only for humans, and CNN of overseas media reports that AI that recognizes the faces of wild bears and cows has appeared.
Automated facial recognition for wildlife that lack unique markings: A deep learning approach for brown bears --Clapham ---- Ecology and Evolution --Wiley Online Library
https://onlinelibrary.wiley.com/doi/10.1002/ece3.6840
Face recognition isn't just for humans — it's learning to identify bears and cows, too --CNN
https://edition.cnn.com/2020/11/22/tech/face-recognition-bears-cows/index.html
Many people can't tell the bear's face and can't identify a particular bear among multiple bears. However, Melanie Clapham, a postdoctoral fellow at the University of Victoria, Canada, said, 'I use individual characteristics to identify bears. For example, some bears have scratches on their ears and nose.' As mentioned, there are various individual differences between bears. According to Clapham, many people look at the bear's overall appearance and try to identify it, but bears can get fat before hibernation, lose weight after hibernation, and dramatically during the year. Because it changes, it is difficult to identify even if you look at the whole body.
Identifying individual bears is very important, as it can be useful for species research and protection, Clapham said. Being able to identify each bear makes it easier to deal with specific bears that cause problems, such as individuals descending into the trash cans and attacking farm animals.
Clapham began thinking that the facial recognition AI used to identify humans could be applied to bears as well, in 2017 on a platform called Wildlabs.net , where conservationists and the technical community interact. Participation. There, he met Ed Miller and Mary Nguyen, who are based in Silicon Valley, and the three of them worked together to start developing 'bear face recognition AI.'
by David Ellis
In developing the bear face recognition AI, the research team improved the existing AI software ' Dog Hipsterizer ' that can identify the dog's face and add a mustache and a hat, and developed an AI that can recognize the bear's face. thing. By letting this AI learn a large amount of bear face photographs, we aimed to develop an AI that can identify each individual.
For machine learning, the research team Canada British Columbia of the night Strait inhabit and Arizona Brooks River grizzly bear to collect a total of 4674 pieces of photos of, 80% of them to the training of the face recognition system, and the remaining 20% Was used to measure the accuracy of the system.
The bear face recognition AI called ' BearID ' developed by the research team is said to be able to identify grizzly bears that have already learned AI with an accuracy of 84%. A grizzly bear tracking project using BearID reportedly tracked 132 individuals, allowing the research team to track them for a longer period of time, cheaper and more non-invasive than implanting RFID tags in their ears. It states that.
Attempts to identify and track individuals using face recognition AI are becoming more widespread, not only for wild animals, but also for livestock raised on farms. Joe Hoagland, a rancher in Kansas, has worked with Professor KC Olson of Kansas State University to develop a cattle tracking app called CattleTracs .
With CattleTracs, anyone can take a picture of a cow, link it to GPS coordinates and time, and save it in an online database. The photos of each individual stored in the database are collated with previous photos to find out which producer the cow was born from and how it came to its current location. I can.
Trained with more than 135,000 photos of young beef cattle, Cattle Tracs seems to be able to identify even first-time cows with 94% accuracy, and RFID that does not work well if the cows are in close contact with each other. It is more accurate and better than the tracking used.
Hoagland argues that CattleTracs can be used to easily track contact between individuals when investigating infectious diseases that spread among livestock. 'With CattleTracs, you can track sick animals to find the cause, isolate them, and do contact tracking. Everything that is said about the new coronavirus can be done with animals,' he said.
Not limited to bears and cows reported by CNN this time, there is a possibility that AI that identifies individuals can be developed for all organisms for which many photographs are available. On the other hand, animal facial recognition technology does not have the same 'privacy issues' as humans, but individual identification AI and training data may be abused by malicious people. For example, poachers may use facial recognition AI training data to track the animals they are looking for and use them for capture and hunting, so it is necessary to take measures so that only specific people can access the data. ..
Related Posts:
