Object-Oriented Learning (OOL): Perception, Representation, and Reasoning
International Conference on Machine Learning (ICML)
Friday July 17, 2020, Virtual Workshop
We introduce Generative Structured World Models (G-SWM), a novel object-centric generative model for videos. G-SWM not only unifies the key properties of previous models in a principled framework but also achieves two crucial new abilities, multimodal uncertainty, and situated behavior. By investigating the generation ability in comparison to previous models, we demonstrate that G-SWM achieves the best or comparable performance for all experiment settings including a few complex settings that have not been tested before. Our project website can be found at https://sites.google.com/view/gswm.