Professor Lee Seong Won's research team from the Department of Electronic Engineering at Kookmin University (President Jeong Seung Ryul) has achieved outstanding research results by publishing two papers at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), the world's most prestigious international academic conference in the field of computer vision.

The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) is the most influential top-tier academic conference in the fields of artificial intelligence and computer vision, renowned for its low acceptance rate and rigorous review standards. It is regarded as a premier academic stage for gauging the research level in these fields, where key research achievements from leading universities and global research institutions worldwide are presented annually.

This research was jointly conducted by the research team of Professor Lee Seong Won from the Department of Electronic Engineering at Kookmin University and the research team of Professor Euntae Kim from Yonsei University. Based on close collaboration, the researchers from both institutions produced research results that combine theoretical sophistication with practical scalability.

The first paper, “Rethinking Pose Refinement in 3D Gaussian Splatting under Pose Prior and Geometric Uncertainty,” proposes a novel optimization framework that systematically analyzes and improves the core issues of camera pose error and geometric uncertainty in the 3D Gaussian Splatting technique, a next-generation 3D scene reconstruction technology gaining attention. By integrating pose prior information and geometric uncertainty, it enables more stable and precise 3D reconstruction, demonstrating scalability to diverse applications such as digital twins, robotics, and immersive content creation.

The second paper, “HypeVPR: Exploring Hyperbolic Space for Perspective to Equirectangular Visual Place Recognition,” proposes a novel visual place recognition method utilizing hyperbolic space to overcome the representational gap between perspective and equirectangular images. This framework is expected to serve as a core foundational technology for next-generation spatial recognition applications—including autonomous driving, drones, robotics, and extended reality (XR)—by ensuring robust recognition performance even across different camera formats.

Professor Lee Seong Won stated, stated, “CVPR is the academic conference where computer vision researchers worldwide most aspire to present, and it is the venue where research originality and completeness are evaluated most rigorously.” He added, “We will continuously strengthen our global competitiveness through research that combines theoretical sophistication with practical applicability.” He further remarked, “We will redouble our efforts in research and talent cultivation to ensure Kookmin University continues to produce leading research outcomes in the fields of artificial intelligence and computer vision.”