Scholarlink

In computer animation, generating motion is difficult. There is a need for a technology to generate actions by reusing motion data in animation and game production. In this study, we develop a system that allows users to generate motions interactively by specifying the types of motions and detailed features based on existing motion data. Because there is no need for new motion capture, this system is useful for quickly creating animation prototypes. The idea of the system is to map key postures in motion to a latent space such that the user can control the posture used for motion generation. Furthermore, detailed posture features can be specified interactively if necessary. The set of original motions and key postures in the motion are input in advance. At runtime, the user specifies the types and characteristics of the motions. A motion graph is used to generate and output the motions. We created a prototype of the system and confirmed that it can generate motions by specifying key postures.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Motion Generation Using Combination of Motion Graphs and Key Pose Selection with Statistical Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Shuntaro Kono,
  • Kunio Yamamoto,
  • Masaki Oshita

摘要

In computer animation, generating motion is difficult. There is a need for a technology to generate actions by reusing motion data in animation and game production. In this study, we develop a system that allows users to generate motions interactively by specifying the types of motions and detailed features based on existing motion data. Because there is no need for new motion capture, this system is useful for quickly creating animation prototypes. The idea of the system is to map key postures in motion to a latent space such that the user can control the posture used for motion generation. Furthermore, detailed posture features can be specified interactively if necessary. The set of original motions and key postures in the motion are input in advance. At runtime, the user specifies the types and characteristics of the motions. A motion graph is used to generate and output the motions. We created a prototype of the system and confirmed that it can generate motions by specifying key postures.