SNUAILAB: “Vision AI predicts the landing point of a 2.7-gram table tennis ball at 112 km/h”
Wins project for video AI-based player action recognition and ball tracking models for sports matches
Vision artificial intelligence (AI) platform company SNUAILAB (CEO Myungho Yoo) announced on the 16th that it has won a project for video AI-based player action recognition and ball tracking models for sports matches from CLOIT (CEO Lee Young-soo), the cloud affiliate of ITCEN Group. The project will run until April next year together with Seoul National University's Computer Vision Lab (CVLab).
SNUAILAB was co-founded in early 2019 by Seoul National University's technology holding company, current SNU professors specializing in AI, and key researchers from Samsung Advanced Institute of Technology and Samsung Techwin (now Hanwha Vision). It has AutoCare, a deep learning automation platform that generates, classifies, and searches AI-based video data as metadata, performs automated training, and finds a model's optimal parameters.
It also provides differentiated services across image preprocessing, using super-resolution and deblurring algorithms and 3D pose estimation algorithms.
Through this project, SNUAILAB will research deep learning AI models for analyzing table tennis matches and validate their commercialization. It will build a pose recognition dataset for players and datasets for detecting and tracking players and the ball, and develop and validate a pose recognition model and a model that predicts where the ball will land. A table tennis ball weighs only 2.7 grams but is fast enough to reach up to 112 km/h. The system must capture a ball moving freely across a table 2.74 m wide and 3.5 m long.
A SNUAILAB official said, “A variety of AI technologies, such as AI-based match analysis, refereeing, and result prediction, have recently been making their way into sports. We expect a differentiated AI sports match operation system to emerge when cloud specialist CLOIT's long-standing know-how in match operation systems is combined with SNUAILAB's video and image correction, 3D pose extraction, and object tracking technologies.”
[ETNews, reporter Lee Kyung-min]