Communications of the ACM
Training Facial Recognition on Some New Furry Friends: Bears
Silicon Valley software developers have trained facial recognition technology to monitor individual bears.
Credit: Melanie Clapham
Silicon Valley software developers Ed Miller and Mary Nguyen have trained facial recognition technology to monitor individual bears, as part of a project called BearID.
They built on dog hipsterizer, a program that identified the faces, eyes, and noses of dogs in photos and placed glasses and mustaches on them, which worked reasonably well on bear faces.
Miller and Nguyen produced a training dataset for the deep learning program, using more than 4,000 photos featuring bears, then manually highlighting each bear's eyes, nose, and ears with boxes.
Melanie Clapham at Canada's University of Victoria said once trained, the system operated unsupervised to detect differences between known bears from 935 photos; it identified them with 84% accuracy.
From The New York Times
View Full Article - May Require Paid Subscription
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA
No entries found