Attempts to identify barnyard animals like cows and pigs via facial recognition are complicated by the software's difficulty in distinguishing between animals, as it was originally engineered for human faces.
Also frustrating efforts is a limited database of available animal faces, compared with human-image databases, which slows the rate at which the software learns.
In addition to slower differentiation time, facial recognition software for animals requires high-definition photos or more profile images from varying angles.
Researchers said animal faces need far more reference points than humans to produce algorithms that can execute similar differentiation levels.
Also challenging is the difficulty of photographing animals, which tend to avoid or attack the camera.
From The Wall Street Journal
View Full Article - May Require Paid Subscription
Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA
No entries found