Taeho Kang

I work on computer vision and graphics for interactive 3D and XR experiences. My research has two parts. One is full-body tracking from egocentric headset cameras, which recovers a user’s pose without external sensors or a heavy capture rig. The other is pose-driven 3D scene animation, where a reconstructed or generated 3D scene responds to the user’s movement. Along the way I have worked on egocentric 3D human pose estimation, dynamic scene reconstruction with Gaussian Splatting, and 3D content generation. I want to build XR worlds that recognize what a person does and respond to it realistically.