Ph.D. Student @ Visual AI Lab, Yonsei University AI Researcher | Computer Vision | Human-Centered AI

I am a Ph.D. student in the Visual AI Lab at Yonsei University, advised by Prof. Hae-Gon Jeon. My research focuses on building human-centered AI systems that understand and predict complex human movements and behaviors. My ultimate goal is to develop AI technologies that genuinely help people and contribute to society through meaningful, human-focused research.

• 🔭 Current Research: Trajectory Prediction, Crowd Behavior Modeling, Generative World Models.
• 🌱 Previous Focus: Facial Expression Recognition (FER), Face Recognition, and real-time vision-based applications.

• [2026.06] Our paper DTC-TRUS was accepted at MICCAI 2026 🎉
• [2026.03] Started my Ph.D. program at the Visual AI Lab, Yonsei University!
• [2025.03] Our paper V-NAW was accepted at the CVPR 2025 Workshop (ABAW Challenge) (Ranked 5th in EXPR Track) 🎉
• [2024.12] Our paper NLA was accepted at AAAI 2025 🎉

• V-NAW: Video-based Noise-aware Adaptive Weighting for Facial Expression Recognition CVPR Workshop (ABAW Challenge), 2025 An extension of NLA into the video domain. V-NAW leverages temporal dynamics and noise-aware sample emphasis for robust video-based FER.
  📄 Paper | 💻 Official Code

• Navigating Label Ambiguity for Facial Expression Recognition in the Wild AAAI, 2025 A novel method that addresses both label ambiguity and class imbalance in FER using adaptive weighting and consistency regularization.
  📄 Paper | 💻 Official Code

• Study on Facial Composite Feature Analysis for Determining Subject Anxiety Levels on Low-Power Computing Modules IEIE Summer Annual Conference, 2024 A study on estimating subject anxiety using eye blinking and facial expression recognition on embedded systems.
  📄 Paper | 💻 Code

• Expression Recognition App: A PyQt-based application for multi-face detection and expression classification from images, videos, or real-time webcams. Visualizes softmax probability distributions.
• Real-time Expression Recognition (Classification): Lightweight real-time facial expression recognition using softmax-based classification, including preprocessing and visualization pipelines.
• Real-time Expression Recognition (Valence-Arousal): An extension of expression recognition to continuous affect modeling via the valence–arousal space with visual affect mapping.

• Real-time Face Detector: A fast and lightweight face detector based on SSD and RetinaFace, optimized for single-person detection scenarios.
• Real-time Eye Blink Detection: Behavioral analysis tool tailored for human state monitoring. Optimized for lightweight deployment on embedded systems (useful for anxiety estimation).