Kanai Lab

AboutAbout the lab

Making computer graphics and physics simulation faster

Kanai Lab conducts research at the intersection of computer graphics, physics simulation, and machine learning.
Drawing on machine learning techniques, the lab works on making computer graphics and physics simulation faster and more efficient.
The lab keeps applications to games and VR in view.

ResearchResearch themes

The lab addresses topics such as liquid simulation, learning the shapes of broken objects, and character animation. Drawing on machine learning techniques as well, it explores ways to make computation faster and more efficient, with applications to games and VR in view.

*Illustrative image

Large-Scale Liquid Simulation↗

A tall cell grid reduces the degrees of freedom in pressure computation by keeping high resolution only near the liquid surface. In addition to the horizontal resolution, the lab introduced the thickness of the layer near the surface as a degree of freedom, reducing computational cost while preserving visual quality, even in scenes with complex water surface motion.

*Illustrative image

Learning the Shapes of Broken Fragments↗

Implicit representations such as signed distance fields are widely used for learning 3D shapes, but learning becomes difficult when there are many small pieces, as with fragments from brittle fracture. The lab proposed a "boundary far field," which rebuilds the distance field for each individual fragment, improving shape reconstruction accuracy and the consistency of internal boundaries.

*Illustrative image

Generating Diverse Walking Styles↗

Real-time character animation requires natural motion that preserves diverse walking styles under user control. The lab proposed a two-stage learning approach that first learns the local periodic features of the limbs and then integrates the periodicity of the whole body.

*Illustrative image

Varying the Thickness of the Layer Near the Liquid Surface↗

In a tall cell grid, the region near the liquid surface is represented with a fine grid, and this region is called the optical layer. Its thickness was previously constant, but the lab proposed a structure that varies the thickness according to the motion of the liquid, reducing the number of cells needed in the projection step.

VoicesStudent voices

Ratings and reviews reflect the experiences of individual current or former students, not the lab's official views. Whether a lab has reviews, or how many, says nothing about the lab itself.

Students who take part in an interview may receive a thank-you gift worth ¥1,000, regardless of what they say.