Disentangled Sequential Autoencoder
This work addresses the challenge of controlling content and dynamics in sequential data generation, which is incremental in improving generative models for applications like video compression.
The authors tackled the problem of modeling high-dimensional sequential data by proposing a VAE architecture that disentangles static and dynamic latent features, enabling content swapping in video and audio sequences, such as converting a male speaker to a female speaker.
We present a VAE architecture for encoding and generating high dimensional sequential data, such as video or audio. Our deep generative model learns a latent representation of the data which is split into a static and dynamic part, allowing us to approximately disentangle latent time-dependent features (dynamics) from features which are preserved over time (content). This architecture gives us partial control over generating content and dynamics by conditioning on either one of these sets of features. In our experiments on artificially generated cartoon video clips and voice recordings, we show that we can convert the content of a given sequence into another one by such content swapping. For audio, this allows us to convert a male speaker into a female speaker and vice versa, while for video we can separately manipulate shapes and dynamics. Furthermore, we give empirical evidence for the hypothesis that stochastic RNNs as latent state models are more efficient at compressing and generating long sequences than deterministic ones, which may be relevant for applications in video compression.