Facebook FB +1.16%said it had developed a way to recognize people in photos even if their faces are obstructed, reports blogs.wsj .
“We can easily picture Charlie Chaplin’s moustache, hat and cane or Oprah Winfrey’s curly volume hair,” Facebook said in a paper presented earlier this month. “Yet, examples like these are beyond the capabilities of even the most advanced face recognizers.”
But such isolated features are sufficient for Facebook’s latest image-recognition technology. The social network identified individuals with 83% accuracy using a method dubbed PIPER, an acronym for pose invariant person recognition. The method took into account body poses, partial views of faces and other cues.
Clothing isn’t always a good indicator of identity, researchers said, because multiple people may wear the same outfit and people change clothes.
Facebook tested PIPER on a data set collected from public Flickr albums and consisting of more than 37,000 photos with more than 63,000 images of people. Some 2,350 people were represented in the collection.
Slightly more than half the photos contained images of faces at a resolution suitable for recognition.
The findings were first reported in New Scientist.
Source: blogs.wsj
No comments:
Post a Comment