Causal Spatiotemporal State Evolution for Monocular 4D Reconstruction

Key points

• To address this, we propose a feed-forward framework for causal 4D visual geometry learning from monocular videos.1
• Through complementary Temporal-Spatial and Spatial-Temporal Scans, CSM continuously propagates geometry-aware states across frames to model spatiotemporal evolution.1
• To align this causal state evolution with observable dynamics, we introduce Unified Motion Supervision.1
• These results validate that the synergy between the directional state propagation of Causal Spatiotemporal Mamba and Unified Motion Supervision effectively captures continuous scene evolution, significantly reducing temporal drift and reinforcing spatial-temporal consistency in 4D representation learning.1

Sources (1)

• [1]Causal Spatiotemporal State Evolution for Monocular 4D Reconstruction
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 04:00 AM

  While feed-forward models excel in static 3D geometry estimation, extending them to dynamic 4D scenes requires modeling visual geometry as a causally evolving state.

  To address this, we propose a feed-forward framework for causal 4D visual geometry learning from monocular videos.

Extractive summary: sentences quoted from the sources.