Junseok Lee
I am a Robotics Researcher at the Humanoid Robot System Research Department, ETRI (Electronics and Telecommunications Research Institute) 
, where I
focus on developing the intelligence of humanoid robots.
I received my Ph.D. in AI Convergence (Intelligent Robotics Program) from GIST (Gwangju Institute
of Science and Technology) in 2025, where I was a member of the AI Lab under the
supervision of Prof. Kyoobin Lee. Prior to joining ETRI, I conducted humanoid robotics research at Advanced Robotics Lab, LG Electronics 
until 2026.
Research
My research lies at the intersection of computer vision and robot learning, with a particular focus on making complex AI models practical for real-world robotic systems. My primary research interests include:
- Robot Vision & Vision-Language-Action Models (VLA)
- Neural Architecture Design
- Model Compression & Efficiency (e.g., Knowledge Distillation)
When Low-Bit VLAs Fail: Diagnostics and Quantization-Conditioned Reinforcement Learning
Conference on Robot Learning (CoRL 2026) (Under review)
CD-FKD: Cross-Domain Feature Knowledge Distillation for Robust Single-Domain Generalization in Object Detection
IEEE International Conference on Robotics & Automation (ICRA 2026)
Rethinking Deformable Convolution as an Adapter with Cross-layer Weight Sharing for Robust Semantic Segmentation in the Wild
IEEE/CVF Computer Vision and Pattern Recognition (CVPR 2026) (Under Review)
Robust Maritime Object Detection under Adverse Conditions via Joint Semantic Learning without Extra Computational Overhead
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2025)
MV2: A Large-Scale 360-degree Multi-View Maritime Vision Dataset for Object Detection and Segmentation
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2025)
Automated Diagnosis for Extraction Difficulty of Maxillary and Mandibular Third Molars and Post-Extraction Complications using Deep Learning
Scientific Reports
IF 4.3(Top 18.5%, Q1)
3rd Workshop on Maritime Computer Vision (MaCVi) 2025: Challenge Results
Winter Conference on Applications of Computer Vision (WACV 2025) Workshop
🥇 1st in Semantic Segmentation Track , 🥈 2nd in Panoptic Segmentation Track
MART: MultiscAle Relational Transformer Networks for Multi-agent Trajectory Prediction
European Conference on Computer Vision (ECCV 2024)
- 27.9% Acceptance Rate
Teaching Where to Look: Attention Similarity Knowledge Distillation for Low Resolution Face Recognition
European Conference on Computer Vision (ECCV 2022)
- Best Paper Award, 2019, Korea Computer Congress
- 28.0% Acceptance Rate
Buoy Light Detection and Pattern Classification for Unmanned Surface Vehicle Navigation
International Conference on Ubiquitous Robots (UR 2024)
- Best Paper Award Finalist
Automatic Detection of Injection and Press Mold Parts on 2D Drawing using Deep Neural Network
International Conference on Control, Automation and Systems (ICCAS 2021)
Deep learning based food instance segmentation using synthetic data
International Conference on Ubiquitous Robots (UR 2021)