Research map / Video / Video segmentation and tracking
Disentangling Physical Dynamics From Unknown Factors for Unsupervised Video Prediction
Vincent Le Guen, Nicolas Thome
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 |
|---|---|---|---|
| Add a residual branch to the physical-only PhyCell (ablation) | PhyDNet · Moving MNIST | MSE: 50.8 -> 24.4 | Improved |
| Add a residual branch to the physical-only PhyCell (ablation) | PhyDNet · Traffic BJ | MSE × 100: 48.9 -> 41.9 | Improved |
| Add a residual branch to the physical-only PhyCell (ablation) | PhyDNet · Sea Surface Temperature | MSE × 10: 38.2 -> 31.9 | Improved |
| Add a residual branch to the physical-only PhyCell (ablation) | PhyDNet · Human 3.6 | MSE / 10: 42.5 -> 36.9 | Improved |
| Remove the moment loss from training (ablation) | PhyCell · Moving MNIST | MSE: 50.8 -> 43.4 | Improved |
| Add the moment loss to training (ablation) | PhyDNet · Moving MNIST | MSE: 29.0 -> 24.4 | Improved |
Closest papers by meaning
- Learning Physics From Video: Unsupervised Physical Parameter Estimation for Continuous Dynamical Systems (CVPR 2025 · similarity 0.87)
- PhyMPGN: Physics-encoded Message Passing Graph Network for spatiotemporal PDE systems (ICLR 2025 · similarity 0.85)
- Stable Long-Horizon PDE Forecasting via Latent Structured Spectral Propagators (NeurIPS 2026 · similarity 0.85)
- CALM-PDE: Continuous and Adaptive Convolutions for Latent Space Modeling of Time-dependent PDEs (NeurIPS 2025 · similarity 0.84)
- P$^2$C$^2$Net: PDE-Preserved Coarse Correction Network for efficient prediction of spatiotemporal dynamics (NeurIPS 2024 · similarity 0.84)
- ODEWorld: A Continuous Predictive Architecture via Physical-Time Flow (NeurIPS 2026 · similarity 0.84)
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.
