Lists

▼

Friday, June 16, 2017

KAIST “K-Eye” Recognizes Faces

KAIST research team led by Prof. Hoi-Jun Yoo developes CNNP chip (CNN Processor), that runs AI algorithms with ultra-low power, and K-Eye, a face recognition system using CNNP. The system was made in collaboration with a start-up company, UX Factory.

The “Always-on” image sensor, can determine if there is a face in its camera range. Then, it can capture frames and set the device to operate only when a face exists, reducing the standby power significantly. CNNP achieved 97% face recognition accuracy while consuming only 0.62mW of power. Both chips were developed by Kyeongryeol Bong, a PhD student under Prof. Yoo and presented at ISSCC in February.


A Youtube video demos the K-Eye operation:



Update: EurekaAlert also publishes a die photo:

at

5 comments:

  1. AnonymousJune 16, 2017 at 10:17 PM

    They are South Koreans, right?
    The camera on the chest of a cop in the video will be very welcomed by Russian, Chinese, and neighboring North Korean governments.

    The video must have been made by the student authors themselves. They might need some civic education even though they might be brilliant engineers.

    ReplyDelete
    Replies
    1. Eric R FossumJune 17, 2017 at 1:59 PM

      Body-cams for police have become quite common in the US in the past ~5 years. Incorporation of facial recognition will be a hard to control and will bring both benefits and problems. Niqabs and burkas might come into general fashion for men and women, or be completely outlawed. Who knows?

      Delete
  2. AnonymousJune 17, 2017 at 5:40 PM

    my problem is, with mesh network of these unnoticeable cams, soon they will know where I go, when I go there, and for what possible reasons I would be there.

    ReplyDelete
  3. AnonymousJune 20, 2017 at 10:00 PM

    very nice idea. I like how he went further and discussed applications of the technology. Let's see if this will go to market.

    ReplyDelete

All comments are moderated to avoid spam and personal attacks.