Seonghak Lee

I am a Ph.D. candidate at Computer vision and Machine Learning Lab (CVML) at Chung-Ang University, under the supervision of Prof. Junseok Kwon.

My research focuses on photorealistic digital humans and speech-driven human modeling, with the broader goal of better understanding and representing the human head. I am particularly interested in modeling facial geometry, appearance, motion, and speech-related dynamics, as well as 3D reconstruction, neural rendering, and scene representation for realistic and controllable digital humans.

Email / GitHub / Google Scholar / LinkedIn / CV

Seonghak Lee

Publications - Conference

* denotes equal contribution.  † denotes corresponding author.

SENSE teaser

SENSE: Semantic Neural Speech Synthesis from Brain Dynamics via Spatial Graph Encoding

Jisoo Park, Seonghak Lee, Hyojin Park, Junseok Kwon†

The Conference on Neural Information Processing Systems (NeurIPS), Sydney, Australia, 2026

Project Page Code
URHead teaser

URHead: A Unified UV-Space Representation for Joint Mesh–3DGS Optimization in Head Avatars

Seonghak Lee, Junhee Cho, Jisoo Park, Min-Gyu Park, Jongmin Lee, Ju Hong Yoon, Junseok Kwon†

The European Conference on Computer Vision (ECCV), Malmö, Sweden, 2026

Project Page PDF Code
UniVoiceLite teaser

Lightweight Wasserstein Audio-Visual Model for Unified Speech Enhancement and Separation

Seonghak Lee*, Jisoo Park*, Junseok Kwon†

IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), Honolulu, Hawaii, USA, 2025

PDF Code
ESM teaser

Enhancing Sidewalk Maintenance through Accurate Joint Deflection Measurement

Seonghak Lee*, Jisoo Park*, Paul Hyunbin Cho*, Zongkai Xia†, Loucif Hebbache†

IEEE 43rd International Conference on Consumer Electronics (ICCE), Las Vegas, USA, 2025 — in collaboration with IRIS R&D Group Inc.

PDF
WildTalker teaser

WildTalker: Talking Portrait Synthesis in the Wild

Seonghak Lee*, Jisoo Park*, Junseok Kwon†

The 18th European Conference on Computer Vision Workshop (ECCV Workshop) — 3D Modeling, Reconstruction, and Generation in the Wild, Milano, Italy, 2024

Project Page PDF Code
POTF teaser

POTF: Prompt-based Object-centric Tensorial Field

Seonghak Lee*, Jisoo Park*, Junseok Kwon†

International Conference on ICT Convergence (ICTC) — Oral, Jeju, Korea, 2024

Project Page PDF Code

Publications - Journal

WildTalker-infinity teaser

WildTalker∞: Pushing the Limits of 3D Talking Portrait Synthesis in Unconstrained Environments

Seonghak Lee*, Jisoo Park*, Junseok Kwon†

IEEE Transactions on Audio, Speech and Language Processing (T-ASLP), vol. 34, pp. 1259–1271, 2026

PDF
VT-Surf teaser

VT-Surf: Visual Tracking with Switching Dynamics under Time-Series Forecasting

Seonghak Lee, Jisoo Park, Radu Timofte, Junseok Kwon†

IET Electronic Letters, Vol. 61, 2025

PDF
SOTA teaser

SOTA: Sequential Optimal Transport Approximation for Visual Tracking in Wild Scenario

Seonghak Lee, Jisoo Park, Radu Timofte, Junseok Kwon†

IEEE Access (ACCESS), Vol. 12, no. 1, pp. 177028–177037, 2024

PDF

Awards