Conditional motion in-betweening
作者：
Highlights：
• We propose a conditional motion in-betweening (CMIB) method, providing additional controllability in motion in-betweening (MIB). The proposed method can perform two different CMIB tasks, pose-conditioned and semantic-conditioned MIB.
• We propose a motion data augmentation method, which is effective in pose-conditioned MIB tasks by defining a probability distribution over smooth trajectories.
• Our proposed method achieves state-of-the-art performance on MIB tasks on multiple motion capture databases while providing additional controllability.
• We present an evaluation measure for semantic-conditioned MIB tasks.
摘要
•We propose a conditional motion in-betweening (CMIB) method, providing additional controllability in motion in-betweening (MIB). The proposed method can perform two different CMIB tasks, pose-conditioned and semantic-conditioned MIB.•We propose a motion data augmentation method, which is effective in pose-conditioned MIB tasks by defining a probability distribution over smooth trajectories.•Our proposed method achieves state-of-the-art performance on MIB tasks on multiple motion capture databases while providing additional controllability.•We present an evaluation measure for semantic-conditioned MIB tasks.
论文关键词：Motion in-betweening,Conditional motion generation,Generative model,Motion data augmentation
论文评审过程：Received 10 January 2022, Revised 28 May 2022, Accepted 9 July 2022, Available online 22 July 2022, Version of Record 29 July 2022.
论文官网地址：https://doi.org/10.1016/j.patcog.2022.108894
