Patch-based Object-centric Transformers for Efficient Video Prediction

Paper Code

Abstract

In this work, we present Patch-based Object-centric Video Transformer (POVT), a novel region-based video generation architecture that leverages object-centric information to efficiently model temporal dynamics in videos. We build upon prior work in video prediction via an autoregressive transformer over the discrete latent space of compressed videos, with an added modification to model object-centric information via bounding boxes. Due to better compressibility of object-centric representations, we can improve training efficiency by allowing the model to only access object information for longer horizon temporal information. When evaluated on various difficult object-centric datasets, our method achieves better or equal performance to other video generation models, while remaining computationally more efficient and scalable. In addition, we show that our method is able to perform object-centric controllability through bounding box manipulation, which may aid downstream tasks such as video editing, or visual planning.

Select the corresponding tabs above to view video samples for each dataset

For MoCoGAN-HD, we include unconditional video samples with latent vector for frame 1 repeated for 5 randomly sampled futures

For all other methods, we include single-frame conditioned samples each with 5 randomly sampled futures.

Samples for our method POVT also include examples of samples with generated bounding boxes as well