🏷️ Jongmin Lee (이종민)
Hello, I’m Jongmin Lee, a Postdoctoral Researcher at KAIST (Kim Jaechul Graduate School of AI) in Seoul, working with Prof. Seungryong Kim.
I received my Integrated M.S./Ph.D. in Computer Science and Engineering from Seoul National University under Prof. Sungjoo Yoo, with a dissertation on integrating dense matching into Structure-from-Motion.
As a 3D vision researcher, my work spans image matching, visual localization, 3D reconstruction, and depth estimation. I am currently expanding my research toward 3D/4D generation, including 3D scene generation and world models.
Contact: ✉️ jongmin.cv@gmail.com
Site: 🌐 icetea-cv.github.io
Language: 🇰🇷 Korean / 🇺🇸 English
💡 Research Interests
- Image Matching / Structure-from-Motion
- 3D/4D Reconstruction
- Visual Localization / SLAM
- Depth Estimation / Completion
- 3D/4D Generation (3D scene generation, world models)
🎓 Education
[Mar. 2021 - Feb. 2026] Integrated MS / Ph.D. in Computer Science and Engineering
- Seoul National University in Seoul, Korea
- Advisor: Prof. Sungjoo Yoo
- GPA 3.74 / 4.30
- Dissertation: Integrating Dense Matching into Structure-from-Motion
[Mar. 2016 - Feb. 2021] BS in Computer Science and Engineering
- Seoul National University in Seoul, Korea
- GPA 3.82 / 4.30
🏬 Research Experience
[Mar. 2026 - Present] KAIST (Kim Jaechul Graduate School of AI)
- Postdoctoral Researcher, CVLAB
- under Prof. Seungryong Kim
- Research on 3D/4D reconstruction and generation: autoregressive 4D generation, geometry-supervised video generation, 3D scene generation, world models
[Jul. 2025 - Dec. 2025] Naver Labs Korea
- Collaborating Researcher, Spatial AI Team
- 3D reconstruction and depth completion for panoramic images
[Mar. 2021 - Feb. 2026] Seoul National University, CMALAB
- Ph.D. Researcher (Prof. Sungjoo Yoo)
- Integrated dense matching into Structure-from-Motion for more accurate correspondences, camera poses, and 3D points: Dense-SfM (CVPR 2025) and MV-RoMa (CVPR 2026 Highlight)
- Research on keypoint description, visual localization, and 3D reconstruction (CVPRW 2023)
[Jan. 2023 - Aug. 2023] Naver Labs Korea
- Research Intern, Spatial AI Team
- Supervised by Jerome Revaud (Naver Labs Europe)
- Model-free, one-shot object pose estimation (MFOS, AAAI 2024, first author)
- Scene-agnostic coordinate regression for visual localization (SACReg, CVPRW 2024)
[Jun. 2020 - Feb. 2021] Seoul National University, CMALAB
- Undergraduate Research Intern
- Monocular depth estimation
📖 Publications
- L3G-World: Real-Time Infinite World Generation with Latent 3D Gaussian Memory
- Jinhyeok Choi*, JoungBin Lee*, Jongmin Lee, Minkyeong Jeon, Jiyoung Kim, Kaeun Han, Yoonjin Yoon, Seungryong Kim
- Keywords: World model, 3D Gaussian splatting, Long-horizon video generation, Real-time generation
- Tetris3D: 3D Scene Generation With Objects That Fit Together
- Jaeyeong Kim, Jinhyuk Jang, Jongmin Lee, Kyehong Park, Seungryong Kim
- Keywords: 3D Scene generation, Single-image 3D reconstruction, Physical plausibility
- MVTrack4Gen: Multi-View Point Tracking as Geometric Supervision for 4D Video Generation
- JoungBin Lee, Jaewoo Jung, Jongmin Lee, Tongmin Kim, Hyunsung Kim, Takuya Narihira, Kazumi Fukuda, Jahyeok Koo, Jisang Han, Yuki Mitsufuji, Seungryong Kim
- Keywords: 4D Video generation, Multi-view point tracking
- MORPHOS: Autoregressive 4D Generation with Temporal Structured Latents
- Minkyung Kwon*, Jinhyeok Choi*, Youngjin Shin, Jaeyeong Kim, Jongmin Lee, Seungryong Kim
- Keywords: 4D Generation, Autoregressive generation
- MV-RoMa: From Pairwise Matching into Multi-View Track Reconstruction
- Jongmin Lee*, Seungyeop Kang*, Sungjoo Yoo
- Keywords: Multi-view dense matching, Track reconstruction, Structure-from-Motion
- Dense-SfM: Structure from Motion with Dense Consistent Matching
- Jongmin Lee, Sungjoo Yoo
- Keywords: Structure-from-Motion, Dense matching, Gaussian Splatting, Track refinement
- MFOS: Model-Free & One-Shot Object Pose Estimation
- Jongmin Lee, Yohann Cabon, Romain Brégier, Sungjoo Yoo, Jerome Revaud
- AAAI 2024 [arXiv]
- Keywords: 6D object pose estimation, Visual localization
- SACReg: Scene-Agnostic Coordinate Regression for Visual Localization
- Jerome Revaud, Yohann Cabon, Romain Brégier, Jongmin Lee, Philippe Weinzaepfel
- CVPRW 2024 [arXiv]
- Keywords: Visual localization, Scene coordinate regression
- Multi-Scale Local Implicit Keypoint Descriptor for Keypoint Matching
- Jongmin Lee, Eunhyeok Park, Sungjoo Yoo
- CVPRW 2023 [Paper]
- Keywords: Keypoint matching, Visual localization
* Equal contribution
🛠️ Skills
- Research areas: Image matching, Structure-from-Motion, Visual localization, SLAM, Depth estimation, Object pose estimation, 3D/4D Reconstruction, 3D/4D Generation
- Programming: Python (advanced), C/C++
- Frameworks & tools: PyTorch (advanced), Docker, Git, LaTeX
- Languages: Korean (native), English (fluent)
📖 Service & Other Experience
- Reviewer: AAAI 2027, IJCV 2026, IROS 2025
- Teaching Assistant: Hardware System Design, Seoul National University (2021 - 2023)
- System Administrator: Bacchus (CSE department sysadmin group), Seoul National University (2018 - 2021)