Robert Grey
3 min readJan 14, 2020

--

3D from 2D — AI face analysis used to launch the Snapchat of Asia

Naver Inc., to many known as the ‘Google of Korea,’ is clearly a major player in the peninsula’s technology scene. Developer of the widely-used messaging app Line, Naver was looking for the next product to compete with other global players such as Snapchat and Instagram. The only problem was that in 2016, very few 3D enabled phones were in use, a technology necessary for the virtual stickers using augmented reality (AR) and photographic filters driving the virality of such apps. The other option would require 3D face analysis, though only two companies owning 3D face analysis technology existed at the time, Snapchat and Instagram’s parent company, Facebook.

SNOW Camera marketing image from Google Play.

Naver found their answer in the form of SNOW Camera, originally developed by Camp Mobile. In order to compete with Snapchat’s AR features and stickers, two former Samsung employees were tasked with developing the 3D face recognition, which relies on machine learning and convolutional neural networks, or AI. This allowed nearly all anyone with a regular smart phone newer than Galaxy S3 to use all of the features they expected from Snapchat, but in a local app.

The key technology behind SNOW Camera is a facial landmark estimating algorithm. Face landmark estimating algorithms allow the camera to infer the exact face pose by estimating the position of the important areas of the face, such as the eyebrows, eyes, nose, mouth, and jaw lines. Face frontalization and face landmark detection is also used to increase the face recognition accuracy. This algorithm analyzes a 3D face, head pose, and a face image’s scale through proprietary technologies. This is achieved through 106 points of detection which estimates the major facial areas with clear landmarks.

3D face analysis by Alchera Inc. running on a Galaxy S7 Edge

Shortly after the implementation of 3D face analysis technology, SNOW Camera was launched. Touted as a Snapchat copy by Business Insider, the success of SNOW Camera could not be denied. Bolstered by partnerships with BTS and more than 40 million downloads within its first year, the success attracted much attention, having been acquired by Naver and spun-off as its own company, and even raising $50 million in investment from SoftBank and Sequoia China.

Today, SNOW Camera is as popular as ever, and they are expanding their AI-enabled technologies to other AR applications such as the 3D avatar app ZEPETO.

South Korea is quickly becoming a leader in artificial intelligence, sparked by the government's AI framework initiative and technology leaders starting their own enterprises. Robert will continue to cover AI-related developments in the Korean market. For more news and insights on how AI is being used in South Korea, please follow the Dreamtime publication.

--

--

Robert Grey

Robert focuses on detecting wildfire through vision sensors via AI at Alchera Inc., a visual AI company empowering society-impacting industries through AI.