The LG affiliate competed with 340 teams from businesses and research teams around the world at the Adversarial Vision Challenge hosted by the Neural Information Processing System. It finished fourth, trailing behind teams from Carnegie Mellon University, Tsinghua University and Montreal College of Information and Technology.
LG CNS was the only business team among the top five winners, the company noted.
The firm participated in the three-month-long competition as a defender against attackers that launched attacks by finding vulnerabilities in AI-based image analyses.
LG had to build accurate AI algorithms to judge the images of various objects that were damaged with noise by competitors, the firm explained.
“The core technology of AI image analysis is to discern how much noise is in the image,” a company official said. “LG CNS has outstanding image analysis technologies using deep learning algorisms.”
The AI image analysis technologies are expected to contribute to reducing accidents involving self-driving cars by increasing the accuracy of object recognition, the official added.
By Song Su-hyun (song@heraldcorp.com)
[코리아헤럴드] [K-POP HERALD] [페이스북]
<ⓒKoreaHerald(www.koreaherald.com)무단전재 및 재배포 금지>
