Stories
Slash Boxes
Comments

SoylentNews is people

posted by janrinok on Wednesday April 23 2014, @06:47PM   Printer-friendly
from the they-want-know-everything-about-everybody dept.

Computer scientists have developed a new face recognition algorithm called GaussianFace which outperforms humans for the first time in challenging real-world conditions. The algorithm normalize each face into 150 x 120 pixels by transforming it based on five features: the position of both eyes, the nose and the two corners of the mouth. The accuracy is 98.52% compared to human accuracy at 97.53%. There's a database at umass.edu that captures much of the face variation by labeling faces in the wild which have 13 000 faces of almost 6000 public figures collected from web.

This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 2) by VLM on Wednesday April 23 2014, @07:23PM

    by VLM (445) Subscriber Badge on Wednesday April 23 2014, @07:23PM (#35091)

    "which have 13 000 faces of almost 6000 public figures collected from web"

    1) The average person has 2.1 faces? WTF Harry Potter's first novel? (Sorry if I just spoiled it for the one dude on the planet who hasn't read the book / seen the movie year, oh and also Darth Vader is Luke's father, BTW)

    2) "collected from web" So the story sounds like we can tell identical twins apart, but the reality is they can tell the difference between goatse and tubgirl and the feels meme guy. Well, OK.

    • (Score: 3, Informative) by TK on Wednesday April 23 2014, @07:44PM

      by TK (2760) on Wednesday April 23 2014, @07:44PM (#35100)

      From TFA:

      The task in facial recognition is to compare two images and determine whether they show the same person.

      So the algorithm can tell you that a screen cap of Luke Skywalker calmly listening to Obi Wan in ANH is the same person in a screen cap of Luke shouting "that's not true, that's impossible!" with a scrunched up face at the end of TESB.

      Not entirely clear from the article and the abstract, is if they gave the algorithm pair of images to pass/fail, or a cache of 13,000 images and told it to find all the matches.

      In regards to your second point, there actually are two faces for the "feels guy", a wrinkly human and an anthropomorphic frog.

      --
      The fleas have smaller fleas, upon their backs to bite them, and those fleas have lesser fleas, and so ad infinitum
      • (Score: 2) by JeanCroix on Wednesday April 23 2014, @07:52PM

        by JeanCroix (573) on Wednesday April 23 2014, @07:52PM (#35104)

        with a scrunched up face

        Not to mention scarring from the car accident...

    • (Score: 2) by BradTheGeek on Wednesday April 23 2014, @08:06PM

      by BradTheGeek (450) on Wednesday April 23 2014, @08:06PM (#35110)

      Hmmm, Star Wars and Goatse reference in one post about facial recognition.

      Scans Goatse pic

      "That's not the face you're looking for."

    • (Score: 3, Funny) by maxwell demon on Wednesday April 23 2014, @09:04PM

      by maxwell demon (1608) on Wednesday April 23 2014, @09:04PM (#35142) Journal

      1) The average person has 2.1 faces? WTF Harry Potter's first novel? (Sorry if I just spoiled it for the one dude on the planet who hasn't read the book / seen the movie year, oh and also Darth Vader is Luke's father, BTW)

      Damn, you spoiled it to me. Up to now I didn't even know that Luke was in Hogwarts! ;-)

      --
      The Tao of math: The numbers you can count are not the real numbers.
    • (Score: 1) by kaszz on Wednesday April 23 2014, @09:15PM

      by kaszz (4211) on Wednesday April 23 2014, @09:15PM (#35146) Journal

      "1) The average person has 2.1 faces? WTF Harry Potter's first novel?"

      Or out of 2.1 persons there's one public person.

      • (Score: 0) by Anonymous Coward on Thursday April 24 2014, @11:38AM

        by Anonymous Coward on Thursday April 24 2014, @11:38AM (#35469)

        good point, perhaps even a bit understated.

  • (Score: 4, Insightful) by buswolley on Wednesday April 23 2014, @08:21PM

    by buswolley (848) on Wednesday April 23 2014, @08:21PM (#35119)

    Can anyone doubt that for any skill a computer will out perform it with enough engineering? Increasingly, pattern recognition is old hat...not something at which we handily beat computers.
    The thing we have left over the machine is flexibility: A bunch of effective organic algorithms that can be selectively and effectively applied to a vast array of situations. Once, not if, computers can integrate a multitude of algorithms under context-sensitive goal-selective control, we will be beat, or we will integrate them into ourselves.

    --
    subicular junctures
    • (Score: 1) by kaszz on Wednesday April 23 2014, @09:05PM

      by kaszz (4211) on Wednesday April 23 2014, @09:05PM (#35144) Journal

      Catch is that figuring out how to make a machine to do it is a challenge in itself even if you know it can be done.

      • (Score: 2) by maxwell demon on Wednesday April 23 2014, @09:08PM

        by maxwell demon (1608) on Wednesday April 23 2014, @09:08PM (#35145) Journal

        Until we figure out how we can make a machine figure it out, that is.

        --
        The Tao of math: The numbers you can count are not the real numbers.
  • (Score: 0) by Anonymous Coward on Friday April 25 2014, @09:18AM

    by Anonymous Coward on Friday April 25 2014, @09:18AM (#35990)