Event-Based Vision: A High-Speed Alternative to Frame Cameras

National AI Research Lab (NAIRL) hosted an invited talk by Kuk-Jin Yoon, Professor in the Department of Mechanical Engineering at KAIST, on April 1, 2025, at the Seoul AI Hub.

Under the theme “Event-Based Vision: A High-Speed Alternative to Frame Cameras,” Professor Yoon introduced event cameras, next-generation vision sensors that asynchronously capture changes in a scene rather than recording full frames at fixed intervals. The talk covered how this sensing paradigm enables high-quality visual perception under conditions that challenge conventional frame cameras, such as extreme lighting and high-speed motion, and how computer vision algorithms can be designed to fully exploit these advantages.

Professor Yoon leads the Visual Intelligence Lab at KAIST, where his group studies deep learning-based computer vision spanning 3D depth estimation for autonomous driving, object detection and tracking, and semantic segmentation, alongside pioneering work on event camera-based vision, including algorithms for reconstructing high-quality, high-resolution images from event data without motion blur in low-light and high-light environments. NAIRL will continue to connect Korea’s research community with leading AI experts through its invited talk series.