Jun-Hee Kim
Ph.D. candidate, Department of Artificial Intelligence, Korea University.
I am an Integrated M.S.–Ph.D. student in the Department of Artificial Intelligence at Korea University, South Korea, and a member of the Pattern Recognition & Machine Learning Lab, advised by Prof. Seong-Whan Lee.
For more details, please see my CV.
My research aims to make 3D human pose estimation reliable under real-world conditions—robust to occlusion, rare poses, and domain shifts—by developing scalable and uncertainty-aware models.
I am also interested in extending 3D understanding to Vision-Language-Action (VLA) systems for physics-based animation and controllable human motion generation.
Research Interests
- 3D Human Pose Estimation
- Robust & Scalable Learning (Generalization, Uncertainty)
- Vision-Language Models & Generative Modeling (Text-to-Image / Text-to-Motion)
- Vision-Language-Action (VLA) for Physics-Based Animation
Education
Korea University, Seoul, South Korea
Integrated M.S.–Ph.D. in Artificial Intelligence (Mar. 2021 – Present)
Advisor: Prof. Seong-Whan Lee
Soongsil University, Seoul, South Korea
B.S. in Mathematics and Physics (Double Major) (Mar. 2015 – Aug. 2020)
News
| Sep 19, 2025 | 1 paper is accepted to NeurIPS 2025. |
|---|---|
| Jun 25, 2025 | 1 paper is accepted to ICCV 2025. |
| Dec 9, 2024 | 1 paper is accepted to AAAI 2025. |
| Sep 25, 2024 | 1 paper is accepted to NeurIPS 2024. |
Publications
2025
-
PoseAnchor: Robust Root Position Estimation for 3D Human Pose EstimationIn Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2025
-
Towards Generalizable 3D Human Pose Estimation via Ensembles on Flat Loss LandscapesIn The Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025
-
ProPose: Probabilistic 3D Human Pose Estimation with Instance-Level Distribution and Normalizing FlowIn Proceedings of the AAAI Conference on Artificial Intelligence, 2025
2024
-
Toward approaches to scalability in 3d human pose estimationAdvances in Neural Information Processing Systems, 2024
-
Geometry-driven self-supervision for 3D human pose estimationNeural Networks, 2024
2023
-
Egpose: Explicit and geometric self-supervision for 3d human pose estimationProcedia Computer Science, 2023