Detecting Pornography with Naver AI... "Accuracy 99.5%"

[Asia Economy Reporter Bu Aeri] Naver announced on the 29th that it will apply the artificial intelligence (AI) 'X-eye 2.0' that blocks the exposure of obscene images in search results starting in July.

Launched in 2017, 'X-eye' was created through AI training by classifying 4 million images accumulated on Naver by type. It filtered harmful photos and videos in real time with an accuracy rate of 98.1%.

The newly introduced X-eye 2.0 applies advanced technology. It includes the high-performance image recognition model 'ReXNet' with improved speed and accuracy, and the data augmentation technique 'CutMix' used to address data shortages during AI model training.

The accuracy of X-eye 2.0 is 99.5%. In addition to existing obscene and adult images, the company explained that its excellent judgment accuracy for provocative expressions allows it to be utilized in various services.

Yoo Bong-seok, Head of Naver Service Operations, said, "To create a healthy internet usage environment, Naver has continuously worked to block various harmful content in line with social changes, including obscene materials, adult content, illegal filming, and hate speech," adding, "In the future, we plan to release X-eye's open API technology so that startups and businesses with insufficient technology can utilize it in various fields."

Jung Seok-geun, CEO of Naver Clova CIC, said, "The various advanced research results underway at Naver Clova are important cases that have successfully contributed to solving social problems beyond services and businesses," and added, "We will continue to collaborate to ensure that our self-developed advanced AI technology helps in real life."

This content was produced with the assistance of AI translation services.