Researchers propose BioModule, a lightweight plug-in temporal transformer that attaches to any 3D pose estimator to predict biomechanical attributes from standard 17-joint skeletons. The team constructs a large-scale aligned dataset pairing Human3.6M video and keypoints with the biomechanical label space of Human3.6Mplus to enable frame-accurate cross-modal supervision.
- BioModule is estimator-agnostic and requires no modification of the upstream pose model.
- It predicts biomechanical quantities describing how the body moves, loads, and activates.
- The system is benchmarked across seven state-of-the-art 3D pose estimators.
- This provides the first systematic analysis of how upstream pose estimation quality propagates to downstream biomechanical prediction fidelity.
BioModule serves as a compact, modular bridge between vision-based pose estimation and physically interpretable motion analysis for applications in rehabilitation and sports science.