Stochastic Latent Residual Video Prediction

TL;DR

This paper introduces a novel stochastic latent temporal model for video prediction that uses a residual update rule in latent space, outperforming previous methods on challenging datasets.

Contribution

It proposes the first fully latent stochastic model for video prediction using a residual update scheme inspired by differential equations.

Findings

Outperforms prior state-of-the-art methods on challenging datasets.

Introduces a residual update rule in latent space for stochastic video prediction.

Provides a more interpretable and simpler model for dynamic video prediction.

Abstract

Designing video prediction models that account for the inherent uncertainty of the future is challenging. Most works in the literature are based on stochastic image-autoregressive recurrent networks, which raises several performance and applicability issues. An alternative is to use fully latent temporal models which untie frame synthesis and temporal dynamics. However, no such model for stochastic video prediction has been proposed in the literature yet, due to design and training difficulties. In this paper, we overcome these difficulties by introducing a novel stochastic temporal model whose dynamics are governed in a latent space by a residual update rule. This first-order scheme is motivated by discretization schemes of differential equations. It naturally models video dynamics as it allows our simpler, more interpretable, latent model to outperform prior state-of-the-art methods on challenging datasets.