基于草图引导的人体动作编辑方法、装置、终端及介质
By combining motion autoencoders and diffusion generation models, two-dimensional sketch data is aligned with three-dimensional human motion sequences in the latent feature space, and text prompt data is used for editing. This solves the problem of unstable sketch correspondence and improves the controllability and robustness of human motion editing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- PEKING UNIV SHENZHEN GRADUATE SCHOOL
- Filing Date
- 2026-05-22
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, sketches cannot establish a stable correspondence with 3D human joint data, making it difficult to introduce sketch information into human motion editing models, which affects controllability and robustness.
A trained motion autoencoder is used to align the 2D sketch data and the initial 3D human motion sequence in the latent feature space to determine the latent features of the sketch. A trained diffusion generation model is then used to add noise and perform inverse denoising on the initial 3D human motion sequence based on guiding conditions. Combined with text prompt data, the sequence is decoded and reconstructed to determine the 3D human motion editing image.
It improves the controllability and robustness of human motion editing, achieves stable correspondence of sketch information in the 3D human motion editing model, and improves interaction efficiency and editing accuracy.
Smart Images
Figure CN122244401B_ABST