MS · Computer Vision & Image Processing Lab · Sogang University

Seunghun Oh

Graduate researcher advised by Prof. Unsang Park. I work on vision–language multimodal models and diffusion-based generation — from training-free control of diffusion attention to how we evaluate multimodal systems.

Vision–Language Multimodal / Diffusion Models / Multimodal Evaluation

01

About

I'm a master's student in the Computer Vision and Image Processing Lab (CVIP) at Sogang University, advised by Prof. Unsang Park. Before joining CVIP, I was an undergraduate researcher at the AI Accelerator Lab at Hallym University and a research intern at ETRI.

My current research treats cross-attention in diffusion models as a signal that can be analyzed and edited at inference time, and asks how vision–language models should be measured — designing diagnostics that separate what a model knows from how a benchmark is built.

Earlier work spanned sleep-stage classification, multi-modal physiological signals, and applied competitions in detection, segmentation, and OCR.

PositionMS Student, CVIP Lab
Sogang University · 2025 – present
AdvisorProf. Unsang Park
PriorResearch Intern, ETRI
Jan – Mar 2025
EducationB.S., Hallym University
2019 – 2025
Based inSeoul, South Korea
02

Publications

2026Under review

Steady-Forcing: Balancing Spatial Persistence and Motion Continuity in Long-Horizon Nature Video Diffusion

Matiur Rahman Minar, Seunghun Oh, Ganghyeon Jeong, Unsang Park

Under review (BMVC 2026). A memory-and-training framework for long-horizon, fixed-camera nature video generation that balances spatial persistence with motion continuity — sustaining plausible fluid dynamics such as water, fire, and smoke over multi-minute autoregressive rollouts.

2026First authorUnder review

Grounded Distractor Probing: A Diagnostic Protocol for Hard-Distractor Sensitivity in VLM Multiple-Choice Evaluation

Seunghun Oh, Ganghyeon Jeong, Unsang Park

Under review (CIKM 2026). A matched original / permutation / hard diagnostic protocol probing how vision–language models respond to same-type, visually plausible near-miss distractors — showing that evidence-first prompting is a model-dependent intervention rather than a universal fix.

2026First authorUnder review

Attention Frequency Modulation: Training-Free Spectral Modulation of Diffusion Cross-Attention

Seunghun Oh, Unsang Park

Under review (ECCV 2026). An inference-time method that modulates cross-attention logits in the Fourier domain to control generation without retraining, while largely preserving semantic alignment.

2026Journal

PSG-Free: Multi-View Facial Imaging and Attention-Based Fusion for OSA Severity Classification

Dongyoun Kim, Yunhee Woo, Jihoon Park, Jaemin Jeong, Seunghun Oh, Youngwoong Ko, Il-Hwan Lee, Dong-Kyu Kim, Jeong-Gun Lee

Expert Systems with Applications (ESWA), 2026

2024First author

PixleepFlow: A Pixel-Based Lifelog Framework for Predicting Sleep Quality and Stress Level

SeungHun Oh, SungJi Ko, YoungHoon Na, Hyunkyung Lee

Int'l Conf. on ICT Convergence (ICTC), 2024 · ETRI Human Understanding AI Paper Challenge

2024Preprint

MC2SleepNet: Multi-modal Cross-masking with Contrastive Learning for Sleep Stage Classification

Younghoon Na, Hyun Keun Ahn, Hyun-Kyung Lee, Yoongeol Lee, Seunghun Oh, Hongkwon Kim, Jeong-Gun Lee

arXiv preprint · SOTA on SleepEDF-78 and SHHS

2024First author

Improving Sleep Stage Classification with Autoencoders and Contrastive Learning

Seunghun Oh, Dongyoung Kim, Jeong-Gun Lee

Annual Symposium of KIPS (ASK), 2024 · in Korean

2023First author

Performance Evaluation of Single-Epoch Models for Real-Time Sleep Stage Classification

Seunghun Oh, Dongyoung Kim, Jeong-Gun Lee

Joint Conf. on Communications and Information (JCCI), 2023 · in Korean

03

Selected Projects

2024.10 – 2025.01cross-attention

AHI Classification from Facial Images

Transformer-based, PSG-free prediction of obstructive sleep apnea severity from multi-view facial images. Basis for the ESWA publication.

2024.10machine unlearning

Boundary Unlearning

Class-level forgetting by shifting the decision boundary of a trained network — motivated by deepfake and data-removal settings.

github ↗
2024.02 – 2024.06detection · ocr

Diet-Record App (Capstone)

Object detection + OCR pipeline that reads nutrition labels and logs personalized intake. Four-person team capstone project.

github ↗
2023.08machine learning

SeaGuard

Route-adjustment system that recommends safer ship paths from ocean-state conditions to help prevent maritime accidents.

github ↗
04

Awards

Minister's Award · 2024.10

3rd ETRI Human-Understanding AI Paper Challenge

Minister of Science and ICT Award · 1st place (public 1st / 41)
1st place
2025.02Distinguished Service AwardPresident's Award, Hallym University
2024.11SW Week Hallym AI Competition 2024Image restoration · SW Dean's Award2nd
2024.08Dacon Joint AI Competition — Fake Voice DetectionHancom Award · DANN-based domain alignment5th
2023.11SW Week Hallym AI Competition 2023Imbalanced classification · President's Award1st
2023.08Dacon Joint AI Competition — Satellite SegmentationSW Univ. Council Award · DINO pseudo-labels + U-Net refinement4th
2022.11Maicon Defense AI CompetitionHP Korea Award · all-weather image denoising4th
05

Contact

Let's talk research.
gnsgus190@gmail.com