연구실

지능형계산영상연구실(ICL)은 계산영상에 뿌리를 두고 있습니다. 계산영상에서는 수학적으로 엄밀하면서도 효율적인 최적화가 모든 영상을 그것이 측정된 물리 과정에 단단히 연결합니다. 저희는 이 그라운딩(grounding)의 원리를 멀티모달, 멀티스케일 표현과 파운데이션 모델로 확장하며, 엄밀성을 잃지 않으면서도 대규모로 확장 가능한 알고리즘을 설계합니다. 나아가 물리 세계에서 행동하고 사람과 협력하는 에이전트로 그 고리를 완성하며, 측정이든 사람의 지시든 모든 입력을 가치 있지만 불확실한 것으로 다룹니다.

계산 및 의료영상;
AI 신호처리;
딥러닝

H G Seo, B H Lee, M J Kim, D H Mah, J H Lee, S Y Chun, "Strong Helps Weak: Directional Cross-Modal Alignment Transfer in Multi-modal LLMs," Accepted at NeurIPS, 2026 (Spotlight)
J S Bang, D W Choi, D J Mun, S Y Chun, "FLINT: Coupling Proximal Initialization and Bounded Stochasticity for Flow-Matching Inverse Problems," Accepted at NeurIPS 2026
J H Jang, H Y Kim, C W Lee, J J Kim, S Y Chun, "HyFL-CLIP: Hyperbolic Fine-Tuning of CLIP for Robust Long-Context Understanding," ECCV 2026
J S Bang, D J Mun, H G Seo, S M Hong, S Y Chun, "Triadic Dynamics Aware Diffusion Posterior Sampling for Inverse Problems: Optimizing Guidance and Stochasticity Schedules," ICML 2026
G H Kim, J J Kim, S Y Jeon, J S Park, S Y Chun, "Human Interaction-Aware 3D Reconstruction from a Single Image," CVPR 2026 (Highlight)
H Y Kim, J H Jang, J J Kim, S Y Chun, "Uncertainty-guided Compositional Alignment with Part-to-Whole Semantic Representativeness in Hyperbolic Vision-Language Models," CVPR 2026 (Highlight)
H G Seo, B H Lee, J H Cho, S J Lim, S Y Chun, "Erasing Thousands of Concepts: Towards Scalable and Practical Concept Erasure for Text-to-Image Diffusion Models," CVPR 2026
W G Jeong, H G Seo, S Y Chun, "Training-free, Perceptually Consistent Low-Resolution Previews with High-Resolution Image for Efficient Workflows of Diffusion Models," CVPR 2026
W G Jeong, K R Lee, H G Seo, S Y Chun, "Training-free Mixed-Resolution Latent Upsampling for Spatially Accelerated Diffusion Transformers," Accepted at CVPR 2026 (Highlight)
S M Hong*, J J Kim*, D H Kim, I S Chung, S Y Chun, "DiffBMP: Differentiable Rendering with Bitmap Primitives," CVPR 2026