[Oshita Lab.] [Japanese]


Research Overview


Oshita laboratory is mainly researching human motion data processing techniques for application to computer animation and computer games.
In order to realize virtual character animation, technology is needed to process information about the human body and its movements on a computer. Human motion data is high-dimensional time-series data, which is more complex than other media data such as video and audio. Therefore, various innovations are required to handle them.
In our laboratory, we conduct research to solve problems related to human motion processing by combining or developing basic techniques for handling human motion, such as kinematics, with various techniques in mathematics and computer science (geometry, statistics, biomechanics, machine learning, deep learning, etc.).
Specifically, we are working on the following research themes.
  • Motion control and synthesis
    Currently, the movements of virtual humans in computer games are basically achieved by replaying motion data created in advance. This causes a problem in that virtual humans can only repeat the same movements, which is unnatural. We are developing a technology to dynamically generate natural movements of virtual humans in response to interactions with users and the environment.

  • Motion control interface
    Computer games and animation production require interfaces that allow users to intuitively control various movements of virtual humans. In this laboratory, we are developing interfaces that use unique devices and novel methods for human motion control.

  • Motion deformation and editing
    Current animation production systems only support direct generation and deformation of motion data, which means that only creators with specialized knowledge of human motion can use them. We are developing methods that allow beginners without special knowledge to easily create animations.

  • Motion analysis and visualization
    We are developing technologies to analyze and visualize the movements of subjects acquired using motion capture for applications such as such as sports and dance training.

  • Simulation of the appearance of virtual humans
    In order to make the image of a virtual person look natural, not only the movements of the virtual person but also the clothes, hair, and skin of the virtual person must move naturally. We are developing a method to generate these movements at high speed by combining physical simulation with pseudo-shape deformation models of clothing, hair, and skin.


Research Projects


Motion Control and Synthesis

Agent Navigation Using Deep Learning for Crowd Simulation
(CASA 2019)
Generating Hand Motion from Body Motion
(CAVW 2016, MIG 2014)
Avoidance Motion Using Motion Graph
(MIG 2011)
Automatic Learning of Motion Rules from Control Logs
(CASA 2010)
Motion Control of Virtual Humans with Dynamics
(EUROGRAPHICS 2001)
Motion Control Interface

Multi-touch Interface for Character Motion Control
(WSCG 2012, VC 2010)
Automatic Leaning of Motion Recognition Model
(ISVC 2010)
Motion Control Using Data Gloves
(CGAT 2009)
Sketch-Based Interface for Crowd Animation
(SG 2009, Interaction 2009)
Motion-Capture-Based Avatar Control Framework
(ACE 2006)
Motion Control with Strokes
(CASA 2005)
Pen-based Interface for Motion Control
(SMB 2004, CAG 2005)
Motion Deformation and Editing

Lattice-Guided Human Motion Deformation for Collision Avoidance
(MIG 2017)
Expressive Human Pose Deformation Based on the Rules of Attractive Poses
(ArtsIT 2017, CGI 2017)
Motion Interpolation Using Adjectives
(NICOINT 2016)
Authoring System for Noh (Japanese Traditional) Dance Animation
(CW 2012)
Motion Database and Animation from Natural Language Texts
(VC 2010, CW 2009)
Smart Motion Synthesis
(PG 2008)
Motion Style Transformation
(CASA 2006)
Motion Analysis and Visualization

Self-Training System for Tennis Shots with Motion Feature Assessment and Visualization
(TVC 2019, CW 2018)
Simulation of the Appearance of Virtual Humans

Skin Deformations Using Dynamic Patterns
(ArtsIT 2009)
Real-time Hair Simulation on GPU
(CASA 2007)
Real-time Cloth Simulation and Virtual Try-On System
(CA 2001, VC 2003)


e-mail address