Convolutional neural network on-a-chip promises always-on face recognition: Page 2 of 2

June 15, 2017 // By Julien Happich
Researchers at the Korea Advanced Institute of Science and Technology (KAIST) have developed a Convolutional Neural Network Processor on silicon that they combined with a custom made image sensor to perform face recognition with a 97% accuracy while only drawing 0.62mW.

Dubbed K-Eye series, the face detection and recognition device can be used with a smartphone via Bluetooth, and can operate for more than 24 hours on one battery charge. The device which could be manufactured as a dongle combines an analogue processor with a CMOS image sensor to distinguish the background area from the area likely to include a face. Then a digital processor detects the face only in the selected area, making the system very effective in terms of frame capture, face detection processing, and memory usage.

In effect, the K-Eye series has been designed as a very low power (less than 1mW) "Always-on" image sensor that can determine if there is a face in its camera range. It only capture frames and set the device to operate when a face has been detected, then the CNNP performs its recognition task.

In a demonstration video, K-Eye is used as a smartphone add-on that allows users to authenticate themselves by looking at their camera, allowing financial transactions. In another example, a policeman wears it around his neck and by accessing a registered database, can automatically check information about the persons he comes across. KAIST also believes the K-Eye could displace other forms of authentication on smartphones, including today's passwords, fingerprint, or iris authentication.

KAIST - www.kaist.edu


Vous êtes certain ?

Si vous désactivez les cookies, vous ne pouvez plus naviguer sur le site.

Vous allez être rediriger vers Google.