PhyMo: Learning Physical Dynamics with Accurate and Continuous Motion from Multi-View Videos

Key points

• Under rendering supervision alone, a model may fit the observed images by changing color or opacity to compensate for inaccurate deformation, rather than recovering the true evolving geometry and motion.1
• In this paper, we propose PhyMo, a framework for learning evolving dynamics from multi-view videos.1
• We introduce a Motion-Aware Trajectory Representation that drives the model to recover accurate motion from deformation, rather than relying on appearance shortcuts to explain observed motion.1
• On top of this, we propose a Continuous Motion Evolution module with bidirectional kinematic continuity constraint to promote temporally consistent evolution of displacement, velocity, and acceleration.1

Sources (1)

• [1]PhyMo: Learning Physical Dynamics with Accurate and Continuous Motion from Multi-View Videos
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 21, 02:49 AM

  Recovering dynamic 3D scene geometry, appearance, and the physical motion states that govern scene evolution from multi-view videos is important yet challenging.

  Under rendering supervision alone, a model may fit the observed images by changing color or opacity to compensate for inaccurate deformation, rather than recovering the true evolving geometry and motion.

Extractive summary: sentences quoted from the sources.