Remember, Revisit, Revise: Scaffold Memory for Long-Horizon 3D Geometry

Key points

• Geometry foundation models estimate camera poses and 3D scene structure from video sequences, but maintaining consistent reconstructions over long horizons remains challenging.1
• We evaluate SMORE with eight competitive geometry models on five datasets spanning camera pose estimation, multi-view depth estimation, 3D point-cloud reconstruction, simultaneous localization and mapping (SLAM), and novel view synthesis.1
• By retaining corrected poses and observations, SMORE further improves the fidelity and spatial consistency of large-scale 3D scene reconstruction.1
• These results establish revisable external memory as a flexible interface for extending pretrained geometry models beyond fixed-context prediction toward iterative, error-correcting inference over long visual sequences.1

Sources (1)

• [1]Remember, Revisit, Revise: Scaffold Memory for Long-Horizon 3D Geometry
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 14, 06:50 AM

  To address this, we introduce Scaffold MemOry for Revision (SMORE), a framework equipping frozen geometry foundation models with a revisable external memory for long-horizon inference.

  Geometry foundation models estimate camera poses and 3D scene structure from video sequences, but maintaining consistent reconstructions over long horizons remains challenging.

Extractive summary: sentences quoted from the sources.