The Korea Herald

지나쌤

Vision AI to help find missing children

Korea's SK Holdings C&C launches artificial intelligence software API to track missing persons with facial, object recognition

By Sohn Ji-young

Published : May 14, 2018 - 15:22

    • Link copied

One of the biggest nightmares for a parent is losing their child in a crowded shopping mall. When a child goes missing, security guards typically go around on foot looking for the child, while mall employees sit with parents to shift through hours of video surveillance footage.

Such procedures -- often slow yet time-sensitive -- can now be significantly expedited when using an advanced facial identification artificial intelligence program, claims SK Holdings C&C, the information technology solutions unit of South Korea’s SK Group.

SK Holdings C&C on Monday announced the launch of the beta version of a new facial recognition artificial intelligence service called “Vision AI.” It offers application programming interface for two functions: facial recognition and object recognition.

(123RF) (123RF)

The facial recognition API will allow people to identify and track down missing children inside places like shopping malls and supermarkets by analyzing surveillance video footage. A single picture of the child is all that is needed, the company said.

The AI program studies a photograph of a person and renders a 3-D model of his or her face. This 3-D rendered image is then used to accurately identify a person’s face at various angles and positions.

“Security guards and parents will no longer have to scavenge around shopping malls or scrutinize surveillance video footage to find a missing child,” SK Holdings C&C said in a press release.

The facial recognition API was developed based on a deep learning engine developed by Korea’s National Forensic Service and is able to handle an infinite number of facial recognition tasks, according to SK Holdings C&C.

In addition to spotting missing children, the facial recognition AI is also expected to be useful for areas including product marketing or identity authentication for security at buildings or financial transactions, it said.

Meanwhile, the object recognition API can currently recognize 80 types of objects found in everyday life, including humans, animals, furniture and transportation vehicles, with room for expansion, SK said.

The object recognition API can be used for various purposes such as automatically recognizing products at the cashier and sending out traffic notifications based on the number of cars out on a road, among others.

Other potential usages for the object recognition API include examining drone footage to inspect a facility, monitoring damaged road infrastructure as well as counting the number of bacteria in the sewage water.

“With a simplified API and detailed guidelines, we hope to enable entry-level developers to develop deep learning-based video image recognition applications with ease,” said Kim Joong-su, director of SK Holdings C&C's Solution Lab in a statement.

By Sohn Ji-young (jys@heraldcorp.com