Inyoung Oh
Hi, I’m Inyoung Oh, a Postdoctoral Fellow in the Visual Intelligence Group, Center for Artificial Intelligence, AI & Robotics Institute at the Korea Institute of Science and Technology (KIST). I received my Ph.D. in Mechanical and Robotics Engineering from Gwangju Institute of Science and Technology (GIST), advised by Prof. Kwanghee Ko in the MODSIM Lab.
My research focuses on geometric deep learning for robust 3D perception and spatial intelligence. I develop geometry-aware learning frameworks that explicitly incorporate geometric structure—such as surface normals, curvatures, discontinuities, and spatial relationships—into neural networks, enabling reliable representation learning under sparse, noisy, and real-world sensing conditions.
My work spans sharp feature detection, surface normal estimation, normal-guided LiDAR semantic segmentation, and monocular 6DoF pose estimation, with applications in robotics, autonomous systems, and mixed reality. These systems are designed to remain stable under real-world variability and to support deployable perception in practical environments.
At KIST, I am expanding this research toward RGB-driven 3D perception and monocular spatial understanding, focusing on estimating metric attributes such as 3D position and human height from monocular RGB input, enabling physically grounded spatial understanding. My goal is to advance spatial AI systems that achieve robust, physically grounded scene understanding across sensing modalities and real-world conditions.
Ultimately, I aim to develop representation learning frameworks where geometric structure serves as a fundamental inductive bias, enabling scalable and deployable perception for intelligent systems operating in complex physical environments.
News
- 2025-11-06 — Our paper has been accepted for publication in Computers in Industry (IF = 9.1): “A Mixed Reality-based Remote Collaboration Framework Using Improved Pose Estimation.” Featured on: GIST official website, GIST Naver Blog, YTN Science.
- 2025-11-04 — Our paper has been submitted and is under review.
- 2024-11-26 — I have successfully defended my doctoral dissertation. I will graduate in February 2026.
- 2023-11-09 — Our paper has been accepted for publication in Journal of Computational Design and Engineering (IF = 4.8): “Improved Semantic Segmentation Network using Normal Vector Guidance for LiDAR Point Clouds.”
-
2021-12-22 — I had a great experience presenting my LiDAR research at a KEPCO research seminar: “Object-detection-based 3D LiDAR intensity calibration and real-time detection of clustered object candidates using deep learning.”
발표 주제: 객체 검출 기반의 3차원 라이다 반사강도 캘리브레이션과 딥러닝을 이용한 클러스터링 객체 후보군에 대한 실시간 객체 검출 - 2020-06-11 — Our paper has been accepted for publication in The Visual Computer (IF = 2.601): “Automated recognition of 3D pipelines from point clouds.”
- 2018-09-07 — Our paper has been accepted for presentation at EuroVR 2018: “Automatic detection of cylindrical objects from unorganized point cloud.”
- 2017-06-06 — Our paper has been accepted for presentation at Computer Graphics International 2017 (CGI 2017): “A RANSAC-based Method for Detection of Multiple Spheres from a Point Cloud.”
Recent Publications
