Jongmin Lee

I am a research scientist / engineer in LG AI Research. I received my Ph.D. degree in Computer Science and Engineering (CSE) at POSTECH, South Korea, where I was a member of the Computer Vision Lab working with Prof. Minsu Cho.

Work Experience

LG AI Research, Seoul, South Korea.
- Research Scientist/Engineer, Oct 2024 ~ Now
- Deep document understanding with multi-modal large langauge models.
Meta Reality Labs., Burlingame, California, U.S.A.
- Research Scientist Intern, Jul 2023 ~ Oct 2023
- Manager: Survi Kyal, Mentor: Fengting Yang, in XR Eyes
Fast campus., Seoul, South Korea.
- Computer Vision Lecturer, Mar 2022 ~ Aug 2022
- Curriculum Sheet: Google Spreadsheet
Vuno Inc., Seoul, South Korea.
- Front-end Developer, Jun 2017 ~ Aug 2017

Education

Ph.D. in POSTECH, Pohang, South Korea. (Sep 2018 ~ Aug 2024)
- M.S./Ph.D. in Computer Science Engineering (CSE)
- Dissertation: Equivariant Image Matching and Pose Estimation
- Advisor: Prof. Minsu Cho
B.S. in POSTECH, Pohang, South Korea. (Mar 2013 ~ Aug 2018)
- B.S. in Industrial and Management Engineering (IME)
- B.S. in Computer Science Engineering (CSE)

Publications

3D Equivariant Pose Regression via Direct Wigner-D Harmonics Prediction [code]
Jongmin Lee, Minsu Cho, in NeurIPS 2024.

Learning Rotation-Equivariant Features for Visual Correspondence [code]
Jongmin Lee, Byungjin Kim, Seungwook Kim, Minsu Cho, in CVPR 2023.
Self-Supervised Equivariant Learning for Oriented Keypoint Detection [code]
Jongmin Lee, Byungjin Kim, Minsu Cho, in CVPR 2022.
Self-supervised Learning of Image Scale and Orientation Estimation [code]
Jongmin Lee, Yoonwoo Jeong, Minsu Cho, in BMVC 2021.
Learning to Distill Convolutional Features Into Compact Local Descriptors [code]
Jongmin Lee, Yoonwoo Jeong, Seungwook Kim, Juhong Min, Minsu Cho, in WACV 2021.
Learning to Compose Hypercolumns for Semantic Visual Correspondence [code]
Juhong Min, Jongmin Lee, Jean Ponce, Minsu Cho, in ECCV 2020.
SPair-71k: A Large-scale Benchmark for Semantic Correspondence [dataset]
Juhong Min, Jongmin Lee, Jean Ponce, Minsu Cho, in ArXiv 2019.
Hyperpixel Flow: Semantic Correspondence with Multi-layer Neural Features [code] [dataset]
Juhong Min, Jongmin Lee, Jean Ponce, Minsu Cho, in ICCV 2019.
Attentive Semantic Alignment with Offset-Aware Correlation Kernels [code]
Paul Hongsuck Seo, Jongmin Lee, Deunsol Jung, Bohyung Han, Minsu Cho, in ECCV 2018.

Research Projects

Kakao Brain, Nov. 2021 - Jul. 2023
Efficient equivariant representation learning in deep neural networks.
Samsung Advanced Institute of Technology (SAIT), Nov. 2022 - Jul. 2023
Non-uniformly exposed burst image restoration using robust base frame selector.
Samsung Advanced Institute of Technology (SAIT), Nov. 2021 - Oct. 2022
Burst image enhancement in an extremely degraded environment by noise, blur and shift.
Samsung Advanced Institute of Technology (SAIT), Nov. 2020 - Oct. 2021
Motion-aware burst image enhancement in the extremely low-light situation.

Professional Services

Reviewer of international conferences
Computer Vision and Pattern Recognition (CVPR) 2022, 2023, 2024, 2025
Neural Information Processing Systems (NeurIPS) 2023, 2024, 2025
International Conference on Computer Vision (ICCV) 2023, 2025
International Conference on Learning Representations (ICLR) 2024, 2025
International Conference on Machine Learning (ICML) 2024, 2025
European Conference on Computer Vision (ECCV) 2022, 2024
ACM International Conference on Multimedia (ACM MM) 2025
Asian Conference on Comoputer Vision (ACCV) 2024
International Conference on 3D Vision (3DV) 2022
British Machine Vision Conference (BMVC) 2021
Artificial Intelligence and Statistics Conference (AISTATS) 2025
Winter Conference on Applications of Computer Vision (WACV) 2021, 2022, 2023, 2024, 2025
International Conference on Machine Vision Applications (MVA) 2021, 2023
International Conference on Pattern Recognition (ICPR) 2020
Program committee of international conferences
Annual AAAI Conference on Artificial Intelligence (AAAI) 2025
Reviewer of international journals
IEEE Transactions on Pattern Analysis and Machine Intelligence (2023, 2024)
International Journal of Computer Vision (2023, 2024)
IEEE Transactions on Image Processing (2022, 2023)
Pattern Recognition (2022, 2023)
The Visual Computer (2022)

Honors and Awards

Qualcomm innovation fellowship Korea finalist, Qualcomm Technologies Inc. , 2023.
BK21 outstanding paper award, POSTECH CSE , 2022.
Global Ph.D fellowship, National Research Foundation of Korea (NRF), 2019 – Now.
POSTECH CSE research award (2nd place), Undergraduate Research Program, POSTECH Computer Science Engineering Department, 2018.
SK Hynix scholarship, SK Hynix Fellowship Program, POSTECH, 2015.

Invited Talks

3D Equivariance: Learning Equivariant Features for Visual Correspondence and Pose Estimation, Technion - Israel Institute of Technology, Haifa, Isreal, September 2024. (Host: Or Litany)
Hyperpixel Flow: Semantic Correspondence with Multi-layer Neural Features, ICCV 2019 Paper Day with Naver, Hotel Andaz Gangnam, Seoul, Korea, October 2019.
Where is semantic correspondence? - The general image matching problem in deep learning era, Hyundai Motors AI Lab seminar, Pohang, Korea, September 2019.
Semantic Alignment - Find Semantic Dense Correspondence, Naver corp., Pangyo, Korea, October 2018. [Youtube link (Korean)]

Research Interest

My research interests mainly focus on developing novel models and algorithms to address practical challenges in deploying artificial intelligence systems to various real-world application domains. I am currently interested and engaged in the following topics:

Product-oriented AI: Enterprise work agents with M-LLMs, advanced intelligence for augmenting human life, model compression for on-device AI deployment in resource-constrained environments.
Multi-modal large language models: Deep document understanding, video summarization, vision-language-action (VLA) models for robotics and real-world interaction, task-specific fine-tuning for complex multi-modal applications.
Spatial reasoning: Structural learning for visual geometry, image matching, pose estimation, semantic correspondence, optical flow for video understanding