English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Self-Supervised Learning of Structured Dynamics from Videos: The Structured Dynamics Model (SDM)

Forum topic · 小凯 · 2026-07-26

Summary

Researchers Lukas Knobel, Andrew Zisserman, and Yuki M. Asano propose the Structured Dynamics Model (SDM), a video representation learning approach that explicitly disentangles the two dominant sources of frame-to-frame change: camera motion and object motion. Instead of encoding video dynamics in a single entangled latent variable or unstructured spatially dense transition tokens, SDM separates the dominant source of temporal change from residual dynamics via future-feature prediction, using only frozen features from a pretrained image vision transformer. Training combines self-supervised learning on real videos with weak supervision on synthetic Kubric data. The authors also introduce ProbeMotion, a new evaluation suite covering synthetic and real videos with camera motion, object motion, and combined dynamics. SDM outperforms backbone baselines that use global CLS or average-pooled features and performs comparably to strongly supervised representations such as VGGT on several probes, despite using far weaker supervision. The results show that pretrained image models can be readily repurposed for structured video dynamics representations. Paper: arXiv:2507.20472.

Paper Overview

Research area: Computer Vision (CV) Authors: Lukas Knobel, Andrew Zisserman, Yuki M. Asano Published: 2026-07-25 arXiv: 2507.20472

Summary

Understanding motion in video is a fundamental challenge for visual learning, as frame-to-frame change entangles two sources of dynamics: camera motion and object motion. This decomposition has remained underexplored in representation learning, partly because these factors are tightly coupled in natural videos and difficult to supervise separately. Yet recovering it is important for learning robust motion representations that separate meaningful object dynamics from camera-induced variation.

The authors study whether such structured motion representations can be recovered from frozen features of a pretrained image vision transformer. They propose the Structured Dynamics Model (SDM), which explicitly separates the dominant source of temporal change from residual dynamics through future-feature prediction, rather than representing video change with a single entangled latent variable or unstructured spatially dense transition tokens.

Training combines:

  • Self-supervised learning on real videos
  • Weak supervision of scene dynamics on synthetic Kubric data
  • Evaluation

    SDM is evaluated on ProbeMotion, a new evaluation suite covering synthetic and real videos featuring camera motion, object motion, and combined dynamics.

    Key findings:

  • SDM outperforms backbone baselines using global CLS or average-pooled features
  • It performs well compared to strongly supervised representations such as VGGT on several probes, despite using much weaker supervision
  • These results indicate that pretrained image models can be easily repurposed for structured video dynamics representations, providing a useful inductive bias for learning and analyzing latent video dynamics
  • Links

  • Paper: <https://arxiv.org/abs/2507.20472>
--- *Auto-collected on 2026-07-26*

Tags

#self-supervised-learning#video-representation#computer-vision#motion-estimation#vision-transformer#arxiv#structured-dynamics-model

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178447125