- Research team led by Professor Hongje Seong of the School of Electrical and Computer Engineering develops AI technologies for 3D spatial understanding and scene change analysis.

- Undergraduate seniors Moonyeon Jeong and Seunggi Min lead research on view-adaptive 3D Gaussian splatting technology for 3D spatial understanding.

  The University of Seoul announced that two papers from the Computer Vision Laboratory, led by Professor Hongje Seong of the School of Electrical and Computer Engineering, have been accepted to ECCV 2026, one of the world’s most prestigious international conferences in artificial intelligence (AI) and computer vision.

  The European Conference on Computer Vision (ECCV) is regarded as one of the world’s top three  international computer vision conferences alongside CVPR and ICCV. The upcoming ECCV 2026 is scheduled to take place in Malmö, Sweden, from September 8 to 12.

  One of the accepted research papers, titled “ViewSplat: View-Adaptive Splatting for Feed-Forward Synthesis,” proposes an AI technology that understands 3D space from multiple images without camera position information and synthesizes high-quality images from new viewpoints. The research team demonstrated that by dynamically adjusting the 3D representation according to the target viewpoint, they can achieve more precise 3D scene reconstruction and real-time rendering in real-world environments.

△Methodology proposed in the ViewSplat study

  This research is particularly noteworthy because it will be presented at ECCV 2026 by senior undergraduates Moonyeon Jeong (lead author) and Seunggi Min (co-author). This marks the first time that research led by undergraduate students has received international recognition at the world’s most prestigious computer vision conference, highlighting the University of Seoul’s excellent research and educational environment as well as its undergraduate research capabilities. The work is also significant as a university-industry collaborative research project conducted in partnership with Dr. Soo Hyun Lee of the Korea Electronics Technology Institute (KETI), with Professor Hongje Seong serving as a co-corresponding author.

  The second paper “Environmental Change Detection for Real-World Change Analysis,” is the result of an inter-university collaboration with Professor Euntai Kim and his research team at Yonsei University. This research focuses on environmental change detection technology designed to analyze changes between past and present scenes in real-world environments in a more practical manner, with Professor Hongje Seong serving as a co-corresponding author alongside Professor Euntai Kim. The acceptance of these two papers highlights the research competitiveness of Professor Hongje Seong’s team in the fields of 3D spatial understanding and real-world scene analysis.

△Methodology proposed in the “Environmental Change Detection” study

  Professor Hongje Seong stated, “These research results demonstrate the potential of AI technology to understand the 3D structure of the real world and changes over time.” He added, “It is especially significant that undergraduates participated as key authors in papers presented at the world’s most prestigious academic conferences and achieved outstanding results through collaborative research with KETI and the Yonsei University research team.”

△From left: Professor Hongje Seong (corresponding author), student Moonyeon Jeong (first author), and student Seunggi Min (second author)