Research


3D Reconstruction and VR/AR


Performance Capture

  • Studies are conducted to extract 3D models and motions at real time with RGB- or RGB-D cameras. It can be used to develop AR systems for virtual concerts, teleconferences, and so on.

Deep learning-based 3D Reconstruction and Reverse Engineering

  • We study automatic 3D CAD model reconstruction from the scanned data of industrial sites using an RGB-D camera.



Physics-based Simulation for Computer Graphics


Deformable Body Simulation

  • Calculating the restoring forces determined in continuum mechanics, we mimic deformable bodies’ various motions such as deformation, collision, translation, and rotation.

Magnetic Object Simulation

  • The forces exerted on magnetic objects are calculated using Maxwell’s equations to mimic the realistic motions of magnetic objects.

Deep Learning-based Physical Simulation

  • We overcome the limitations of conventional physical simulation methods with deep neural networks.

Virtual Science Experiments

  • We develop a simulation/rendering engine and an authoring tool specialized for science experiments in virtual worlds, and we conduct researches demonstrating the effectiveness of virtual science experiments.