Disentangling Physical Dynamics from Unknown Factors for Unsupervised Video Prediction
This work addresses video prediction for applications like forecasting and missing data handling, but it is incremental as it builds on existing physics-informed and disentanglement approaches.
The paper tackles unsupervised video prediction by disentangling physical dynamics from unknown factors using a two-branch architecture and PDE-constrained prediction, resulting in outperforming state-of-the-art methods on four datasets.
Leveraging physical knowledge described by partial differential equations (PDEs) is an appealing way to improve unsupervised video prediction methods. Since physics is too restrictive for describing the full visual content of generic videos, we introduce PhyDNet, a two-branch deep architecture, which explicitly disentangles PDE dynamics from unknown complementary information. A second contribution is to propose a new recurrent physical cell (PhyCell), inspired from data assimilation techniques, for performing PDE-constrained prediction in latent space. Extensive experiments conducted on four various datasets show the ability of PhyDNet to outperform state-of-the-art methods. Ablation studies also highlight the important gain brought out by both disentanglement and PDE-constrained prediction. Finally, we show that PhyDNet presents interesting features for dealing with missing data and long-term forecasting.