Op-Ed
[Op-Ed] AI That Sees the World: The Next Decade of Korean Computer Vision
Seon Joo Kim, Professor of Computer Science at Yonsei University, serves as lead researcher for Yonsei, one of NAIRL’s participating universities, working on video understanding, 3D scene recognition, and sequence modeling. His column begins with the decade Korean computer vision has just completed: the Korean Conference on Computer Vision (KCCV), which started with around 100 attendees in 2014, drew 1,500 this year, and Korea now ranks third in the world by publication count at top venues such as CVPR, ICCV, and ECCV.
Yet paper counts, Professor Kim argues, are not the same as research leadership. Studies that define new problems and shift the direction of the field remain dominated by a handful of countries and global big tech, and much of the gap comes down to computing resources. With GPUs provided through NAIRL, his lab has begun work on 3D tokenization and streaming video models, foundational technologies for physical AI. Carrying out the full cycle, from problem definition to model design and training, he writes, turns students from followers of research into creators of it.
For the qualitative leap ahead, Professor Kim points to international collaboration that raises Korea’s research questions to world standards, talent development that extends into startups, and long-term research infrastructure that lets researchers pursue hard problems years into the future. Korea becomes an AI talent powerhouse, he concludes, the moment the world follows the questions Korean researchers ask, and new industries begin from those questions.
This column is part of NAIRL’s biweekly series “AI Talent Powerhouse,” published in ZDNet Korea, which examines the direction of Korea’s AI research ecosystem.