BUSTER
Adaptive video sampling for efficient, VLM-guided unsupervised anomaly detection.
BMVC 2026 · Accepted
Reliable and efficient video understanding for social good.
Research for social good has always been my first priority, and it will continue to guide my work. I am a Ph.D. candidate in Brain & Cognitive Engineering at Korea University and a student researcher at KIST AI·Robotics.
My early work on sensing applications, including an ultrasonic SLAM helmet for firefighters and indoor drone tracking, shaped my commitment to dependable systems for real-world needs. I now study reliable and efficient video understanding: identifying anomalous behavior, understanding its temporal context, and explaining it with visual evidence.
My work connects video anomaly detection and understanding, adaptive video sampling, action recognition, and vision-language models (VLMs). Through public-safety research at KIST, I work on contextual anomaly detection and behavior analysis for real-world surveillance video.
Recent work includes BUSTER for VLM-guided anomaly detection with adaptive sampling, Where and What for contextual anomaly detection, and Look Closer for action-guided proficiency estimation.
Current focus: why vision-language models hallucinate in video anomaly understanding, and how selective visual verification can make their explanations more faithful.
Research keywords
- Video understanding
- Video anomaly detection & understanding
- Adaptive video sampling
- Efficient vision-language models
- VLM hallucination & reliability
- Temporal & contextual reasoning
- Action recognition & human motion analysis
- Video surveillance & public safety
Publications
Visit my Google Scholar page for a comprehensive listing.
BUSTER: Adaptive Sampling for VLM-Guided Unsupervised Video Anomaly Detection
BMVC 2026 · Accepted
Learning Robust Representations for Few-Shot Action Recognition with Frame-Level Ambiguities
IEEE Access · Submitted · 2026
Where and What: Contextual Dynamics-Aware Anomaly Detection in Surveillance Videos
IEEE Transactions on Image Processing · 2025
🏆 Impact Factor 13.7 · Top 2.2% in Electrical & Electronic Engineering
A Foundational Research Framework for Real-World Abandoned Object Detection
Expert Systems with Applications · 2025
No publications match this search.
Experience
Research Intern → Student Researcher
Korea Institute of Science and Technology (KIST) · AI·Robotics Institute, Vision Intelligence Group
- Contribute to public-safety video systems and national R&D projects for city-scale CCTV control centers.
- Study action recognition and contextual anomaly detection in real-world surveillance video, focusing on when and where anomalous behavior occurs.
- Research VLM-based explanations of anomalous human behavior, adaptive video sampling, and the reliability of video understanding.
Ph.D. Candidate in Brain & Cognitive Engineering
Korea University
- Study whether ideas inspired by human cognitive processes can lead to better deep neural networks and new insights into video understanding.
- Develop systems that use context to reason about interactions, temporal change, and anomalous behavior rather than relying on appearance alone.
Advised by Prof. Jae-Ho Han.
Researcher
The Korea Transport Institute
- Contributed to autonomous bus development through SLAM, systems for control centers, policy research, and project coordination.
B.Eng. & M.Eng. in Electronic Engineering
Kwangwoon University
- Worked on SLAM and IoT systems, including ultrasonic SLAM for a wearable helmet that helps firefighters navigate indoor environments where GPS is unavailable. This work sparked my interest in research for social good.
- Proposed a passive target tracking method based on DNNs for indoor drone tracking in my master's thesis while studying RF signals, systems, and electronic engineering.
Advised by Prof. Youngok Kim.
Awards
2025 AIR Outstanding Research Award · KIST AI·Robotics Institute
2025 Excellent Paper Award · Korea University Graduate School
2024 Outstanding Oral Presentation Award · KIST Academia and Research Convergence Conference