Lee Hyoseok

Measurement-Consistent Langevin Corrector for Stabilizing Latent Diffusion Inverse Problem Solvers

Best Excellence Prize, Electronics Times ICT Paper Awards, 2025

Presented (Spotlight) in Workshop on Foundations of Deep Generative Models: Understanding Memorization, Generalization, and Reasoning, in conjunction with ICML 2026

MCLC, a plug-and-play module that stabilizes latent diffusion inverse solvers.

CLAY: Conditional Visual Similarity Modulation in Vision-Language Embedding Space

Presented in Workshop on 2nd Human-Inspired Computer Vision, in conjunction with ICCV 2025

CLAY, an adaptive similarity computation method that reframes the embedding space of pretrained Vision-Language Models (VLMs) as a text-conditional similarity space without additional training.

ELITE: Efficient Gaussian Head Avatar from a Monocular Video via Learned Initialization and Test-time Generative Adaptation

Presented in Workshop on 2nd 4D Vision Workshop Modeling the Dynamic World with CVPR 2026

ELITE synthesizes an animatable photorealistic Gaussian head avatar from a casual monocular video by synergistically exploiting 3D data priors and 2D generative priors.

JointDiT: Enhancing RGB-Depth Joint Modeling with Diffusion Transformers

JointDiT, a diffusion transformer that models the joint distribution of RGB and depth within a single unified model.

Dress-up: Generating Animatable Clothed 3D Humans via Latent Modeling of 3D Gaussian Texture Maps

Presented (Oral) in Workshop on Computer Vision for Fashion, Art, and Design with ICCV 2025

Dress-Up, a feed-forward, unconditional generative model for creating photorealistic, animatable, clothed 3D human avatars

FPGS: Feed-Forward Semantic-aware Photorealistic Style Transfer of Large-Scale Gaussian Splatting

Excellence Prize, Electronics Times ICT Paper Awards, 2024

FPGS performs feed-forward semantic-aware photorealistic style transfer of Gaussian Splatting.

Zero-shot Depth Completion via Test-time Alignment with Affine-invariant Depth Priors

Winner of the Qualcomm Innovation Fellowship Korea 2025

Zero-shot depth completion: Align sparse depth measurements with affine-invariant depth diffusion prior at test time.

Bootstrapping Multi-View Features via Bridging the Gap between Linearity of Rendering and Non-linearity of 2D Feature Space

Outstanding Poster Presentation Award

Constructing versatile 3D feature field by addressing feature rendering equation and bootstrapping multi-view features.