Associate Professor at Kwangwoon University. (EDUCATION) Ph.D. & M.S. Computer Science Engineering at Seoul National University. B.S. Electrical Engineering at Seoul National University. (KEYWORDS): [Computer Graphics] [Computer Vision] [Differentiable Graphics Engine] [Neural Physics] [World Model] [In-house Graphics Engine] [eXtended Reality] [Dental/Medical Imaging] [Real2Sim/Sim2Real].
As an Associate Professor and the head of the DIG Lab, I am dedicated to mentoring students to become skilled researchers and collaborators. Allow me to introduce our team.
I work on consistent 3D texture generation with geometry-aware relighting, focusing on multi-view texture optimization and dental imaging applications using the Mitsuba3 renderer.
I am interested in medium-aware 3D Gaussian Splatting reformulations for active digital twins, focusing on forward rendering optimizations and fragment-level gradient accumulation in overlapped gaussian scenarios.
I work on real-time graphics engine optimization and physically-based rendering systems, with expertise in DirectX12 pipeline enhancement and voxel-based global illumination (VXGI) for production-quality rendering engines.
I work on controllable hair editing using sketch-guided diffusion models, focusing on context-robust generation and matte-gated ControlNet architectures in DiT frameworks.
I lead our Genesis-based Real2Sim project — spanning state and trajectory-to-state mappers, inverse-dynamics calibration, residual reinforcement learning (BC + RL), and large-scale multi-agent combat simulation.
I focus on terrain-aware inverse dynamics calibration between Blender and Genesis simulators, exploring 3D terrain generalization for sim-to-sim transfer.
I am researching residual reinforcement learning approaches for path-following control, with an emphasis on covariate shift mitigation and sim2real transfer.
I work on rig-constrained on-the-fly 3D Gaussian Splatting for real-time novel view synthesis with multi-camera systems, specializing in rotation-only rig optimization and incremental pose estimation.
