Geometry-Aware Motion Retargeting with Adaptive Anchors
A new framework for retargeting motion across characters with different body shapes preserves interaction semantics like self-contact and near-body proximity. Unlike prior methods that rely on static correspondences and struggle with exaggerated proportions, this approach uses spatially adaptive anchors dynamically repositioned to reachable regions on the target character via a Transformer-based refinement strategy. The method constrains translated anchors to remain on the target geometry, improving robustness for diverse character shapes.
Key facts
- Retargeting motion across characters with varying body shapes while preserving interaction semantics is challenging.
- Prior geometry-aware approaches rely on static correspondences between predefined regions.
- Static correspondences struggle when the target character has exaggerated body proportions.
- The proposed framework uses spatially adaptive anchors for proximity matching.
- Anchors are dynamically repositioned to reachable regions on the target character.
- A Transformer-based anchor refinement strategy predicts anchor displacements.
- Translated anchors are constrained to remain on the target character geometry.
- The method is presented in a paper on arXiv (2605.19355).
Entities
Institutions
- arXiv