Tencent AI research lab achieves record face recognition rate in MegaFace challenge

April 17, 2017 - 

Tencent’s artificial intelligence research lab, YouTu Lab, has achieved a record-setting 83.29 percent recognition rate in the MegaFace challenge (Challenge1/FaceScrub identification), an internationally recognized facial recognition test managed by Paul G. Allen School of Computer Science & Engineering at the University of Washington.

This marks the second time YouTu Lab has set a world record, following its record in the Labeled Faces in the Wild (LFW) face database challenge in March.

However, unlike LFW, which compares faces in photos to verify identities, MegaFace measures the facial recognition rate in relation to the scale of the key index.

As a result, MegaFace is one of the most reliable and popular frameworks of reference in assessing facial recognition performance, particularly on a massive scale.

With the data training based on 2 million Western faces and 4 million Asian faces, YouTu achieved an 83.29 percent recognition rate in the test to recognize a million faces.

Tencent YouTu Lab achieved 83.29 percent, followed by DeepSense V2 (81.3 percent), Vocord-dee pVo1.2 (80.23 percent), GRCCV (77.28 percent), and SphereFace – Small (75.77 percent).

YouTu Lab said its technology has helped nearly 10 families in Fujian province locate their missing members within 24 hours of their disappearance since March.

In the facial recognition test, the recognition rate for an interference set of one million faces at scale has become the most effective solution when searching for missing persons and wanted fugitives.

The technology is used in a wide range of applications, including Tencent’s social networking site Qzone, the online financial platform WeBank, the integrated mobile transportation platform Didi Chuxing, and mobile communication service provider China Unicom.

Leave a Comment

comments

About Justin Lee

Justin Lee has been a contributor with Biometric Update since 2014. Previously, he was a staff writer for web hosting magazine and website, theWHIR. For more than a decade, Justin has written for various publications on issues relating to technology, arts and culture, and entertainment. Follow him on Twitter @BiometricJustin.