Since 2010, Facebook has used facial recognition software to
automatically tag your friends in photos. Now one researcher has come up
with an algorithm that tags photos based on the relationships that people in images already have with each other.
The algorithm uses the name and location of existing photo tags to
build a "relationship graph," where personal connections in the images
are calculated. That makes it faster and more efficient at tagging
pictures compared to what's currently used by sites such as Facebook and Flickr.
For example, if a father and daughter appear in tagged images
consistently, untagged photos featuring them can be tagged
automatically. If the daughter is in an image with both of her parents,
but the father is untagged, the algorithm is able to recognize him based
on their other images together.
University of Toronto engineering professor Parham Aarabi has been
working on the technology since 2005. He emphasizes the advantages
of the algorithm over Facebook's face-recognition photo tagging system.
"The goal of face recognition is to find or generate a new tag when there is none," Aarabi told Mashable.
"By using the tag locations and generating a relationship graph, we are
essentially extracting the meaning from photos without significant
computational analysis — since we only analyze the name and location of a
particular tag, not the pixels of the entire image."
Aarabi has not yet been in touch with Facebook or other photo sharing
sites about the algorithm technology. It will be patented by the United
States Patent and Trademark Office on Dec. 17.
Subscribe to:
Post Comments (Atom)
0 comments:
Post a Comment