Real2Sim & Sim2Real — Genesis-based Physical Learning
We develop a unified Real2Sim / Sim2Real pipeline built on the Genesis physics engine. At its core are neural inverse-dynamics and trajectory-to-controls (ST...
Our Research Field
Engineering the imaging stack. Graphics × Vision × Generative AI — from in-house rendering engines and inverse graphics to Gaussian Splatting, sim-to-real learning, and domain-tuned perception.
Projects
DiT-based Controllable Hair Style Editing
We build controllable hair-editing systems on top of diffusion-transformer (DiT) backbones. Building on our prior GAN-inversion work for hairstyle manipulati...
Modern In-house Real-time Graphics Engine
We develop a modern in-house real-time graphics engine (VizMotive) built on a contemporary architecture, integrating production-quality rendering features ad...
3D Gaussian Splatting — Medium Models and Real-time Reconstruction
We push 3D Gaussian Splatting (3DGS) as a unifying scene representation for both physically-grounded reconstruction and real-time digital twins. Our research...
Domain-Optimized Segmentation via SAM3 Attention Tuning
We investigate domain-optimized segmentation by tuning the attention mechanisms inside SAM 3-class foundation models so that promptable concept segmentation ...
Dental Realistic Inverse Graphics
We develop a photorealistic inverse-graphics pipeline for the dental domain that combines diffusion-transformer (DiT) priors with 3D Gaussian Splatting (3DGS...
AR Navigation-based Surgical Guidance System
We have developed an AR navigation-based surgical guidance system that augments a surgeon’s view with intra-operative imaging and tool-tracking information. ...
Auto- Vehicle Asistantance Imaging
We have been interested in technology development that combines computer graphics and computer vision, and recently, we have been conducting the development ...
Immersive Extended Reality Experience
Building on our experience in computer graphics and computer vision, we develop immersive extended reality (XR) experiences that combine the latest Unreal En...
GAN-Inversion with StyleGAN2
We have explored GAN-inversion with StyleGAN2 as a powerful tool for high-fidelity hairstyle manipulation. Building on the latent-space structure of StyleGAN...
Rendering and Visualization
We have been developing our own rendering engine for over 10 years since 2010, and have extensive experience in successfully applying it to commercial visual...
Sensor Imaging
Our Expertise in Industrial Vision Solutions