WOOJEONG JIN
I am a M.S. student in CVLAB@KAIST AI under the advisement of Prof. Seungryong Kim.
My research interest lies in understanding the real world through visual information, with the goal of solving practical problems. In particular, my current research focuses on Multimodal Large Language Models (MLLMs), Agentic AI, and Referring Image/Video Object Segmentation.
Before joining KAIST, I received my B.S. in Biomedical Engineering from Korea University.
If you have any questions or would like to connect, feel free to reach out via email (wooj0216@kaist.ac.kr)!
Research Interests
Multimodal Large Language Models (MLLMs)
Agentic AI
Image/Video Understanding
Referring Image/Video Object Segmentation
Education
KAIST, M.S. in Artificial Intelligence, Sep. 2024 - Present
Supervisor: Prof. Seungryong Kim
Supervisor: Prof. Seungryong Kim
Korea University, B.S. in Biomedical Engineering, Mar. 2020 - Aug. 2024
Publications
International Conference
CVPR 2026
InterRVOS: Interaction-aware Referring Video Object Segmentation
IEEE Conf. Computer Vision and Pattern Recognition (CVPR), 2026
CVPR 2026 Findings
Pose-dIVE: Pose-Diversified Augmentation with Diffusion Model for Person Re-Identification
IEEE Conf. Computer Vision and Pattern Recognition (CVPR) Findings, 2026
Under Review
Under Review
AgentRVOS: Reasoning Over Object Tracks for Zero-shot Referring Video Object Segmentation
Under Review at ECCV 2026
Under Review
Referring Video Object Segmentation via Language-aligned Track Selection
Under Review at Pattern Recognition (PR)
Research Experiences
Unified Framework for Semantic Segmentation and Captioning
KAIST
Sep. 2024 - Sep. 2025 · Samsung Electro-Mechanics (SEMCO)
Developed an MLLM-based model that jointly performs anomaly detection, segmentation, and anomaly-type captioning for defect samples arising in manufacturing processes.
Anomaly Detection
VLM & MLLM
Segmentation and Captioning
Development of Deepfake Detection and OCR Text Extraction
KAIST
Sep. 2024 - Jun. 2025 · PlantyNet
Developed a high-performance, lightweight deepfake detection model leveraging pre-trained models, suitable for mobile deployment.
On-Device
Deepfake Detection
Development of Person Re-Identification for Security
Korea University
Jan. 2024 - Aug. 2024 · SK Telecom (SKT)
Developed a pose-diversified data augmentation approach to improve generalization in person re-identification across diverse scenarios.
Person Re-Identification
Honors & Scholarships
Semester High Honors, Fall 2022, Korea University · Award for Academic Excellence
Academic Scholarship, Fall 2021 - Fall 2023, Korea University · Received for Academic Excellence