Computer Vision Lab
Minsu Cho, Suha Kwak, Jaesik Park (Computer Science & Engineering)
Computer vision is a scientific field of artificial intelligence on vision. Just as people see everything in the world through their eyes, computer vision technology aims to help computers identify and understand our visual world. While it may seem as though computer vision is merely synonymous to the human eye; these computers should be able to read, comprehend, and analyze any data entering the camera lens, upon which it makes its own judgments. This actually places it in direct connection with the human brain and explains why computer vision is considered at the very heart of the AI research that is emerging as the ‘hottest’ agenda item in the global scientific community.
The Computer Vision Lab is headed by the three AI experts, professors Minsu Cho, Suha Kwak, and Jaesik Park at the Department of Computer Science & Engineering, POSTECH. Professor Cho is engaged in the research of the correspondence of objects that appear in multiple different videos that are the same object, yet bearing different expressions. People, unlike computers, possess the innate ability to recognize a dog as dog, regardless of whether it is a long-haired one or a large-sized one. Conversely, we also recognize a specific building as it is whether it is bathed with morning sun or lit up artificially at dusk. Although, we take this very simple skill of nuanced sight for granted, computer vision requires much additional processing technology. Professor Cho’s work is to leverage AI to study such correspondence problems.
Professor Kwak specializes in video recognition. In particular, he established an unrivaled position in enabling AI to assimilate learning with the minimum amount of information to recognize content in videos. He developed AI capable of recognizing specific objects even in videos with a lot of background noise or low definition.
Professor Park’s area of expertise is in ‘three-dimensional vision’ including 3D video restoration and recognition. Just as the human eye is capable of imagining the shape of a three-dimensional object by merely glancing at its sides, the goal of 3D vision research is to enable computers to do the same. LIDAR-equipped autonomous vehicles are an example of this as they are capable of sensing distance through light. Unlike general-purpose cameras, LIDAR is able to scan a space in 360-degrees, and is successfully able to recognize people or obstacles in front of the vehicle.
These three professors at the Computer Vision Lab publish noteworthy research findings annually at the Computer Vision and Pattern Recognition (CVPR) conference, the International Conference on Computer Vision (ICCV) and the European Conference on Computer vision, the global ‘top three’ computer vision conferences recognized as authorities in the field. Professor Cho is serving as the Area Chair at both the CVPR conference and the ICCV. He commented, “Computer vision relates to a variety of fields and is well-balanced as a discipline that combines both theory and practice”.
Head of Lab
Science Building Ⅱ 303