Research map / Video / Video segmentation and tracking
Structured Object-Aware Physics Prediction for Video Modeling and Planning
Jannik Kossen, Karl Stelzner, Marcel Hussing, Claas Voelcker, Kristian Kersting
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use a separate inference dynamics network instead of sharing the generative dynamics model (ablation) | STOVE · Billiards | pixel prediction error: 0.240 -> 0.262 | Hurt |
| use a separate inference dynamics network instead of sharing the generative dynamics model (ablation) | STOVE · Gravity | pixel prediction error: 0.040 -> 0.042 | Hurt |
| remove explicit velocity variables from the state (ablation) | STOVE · Billiards | pixel prediction error: 0.240 -> 0.272 | Hurt |
| remove explicit velocity variables from the state (ablation) | STOVE · Gravity | pixel prediction error: 0.040 -> 0.053 | Hurt |
| remove latent state variables (ablation) | STOVE · Billiards | pixel prediction error: 0.240 -> 0.338 | Hurt |
| remove latent state variables (ablation) | STOVE · Gravity | pixel prediction error: 0.040 -> 0.089 | Hurt |
| use STOVE rather than VRNN | video prediction · Billiards | pixel prediction error: 0.526 +- 0.014 -> 0.240 \pm 0.014 | Improved |
| use STOVE rather than DDPAE | video prediction · Gravity | position prediction error: 0.298 -> 0.142 | Improved |
| train STOVE on sequences with varying rather than constant total energies | STOVE · billiards sequences with varying energies | energy prediction quality | Improved |
Closest papers by meaning
- Unsupervised Learning for Physical Interaction through Video Prediction (NeurIPS 2016 · similarity 0.89)
- OCK: Unsupervised Dynamic Video Prediction with Object-Centric Kinematics (ICCV 2025 · similarity 0.88)
- Learning to Generate Rigid Body Interactions with Video Diffusion Models (ECCV 2026 · similarity 0.87)
- PlaySlot: Learning Inverse Latent Dynamics for Controllable Object-Centric Video Prediction and Planning (ICML 2025 · similarity 0.87)
- Physics-as-Inverse-Graphics: Unsupervised Physical Parameter Estimation from Video (ICLR 2020 · similarity 0.87)
- Dynamics: Language-Based Representation for Inferring Rigid-Body Dynamics From Videos (CVPR 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
