MV-SSM: Multi-View State Space Modeling
What it is
MV-SSM is a framework for 3D human pose estimation that uses State Space Modeling (SSM) to capture joint spatial sequences. It introduces a Projective State Space (PSS) block and Grid Token-guided Bidirectional Scanning (GTBS) to improve generalization across different camera configurations and occluded scenes.
The problem it solves
Overfitting to specific camera arrangements and difficulty modeling spatial arrangements of keypoints in occluded scenarios using standard attention-based transformers.
What you could use it for
- 3D human pose estimation in multi-view camera setups
- Robust pose estimation in scenarios with varying camera arrangements
- Cross-dataset 3D human pose estimation
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.