/dynamicworld

Primary LanguageJupyter NotebookApache License 2.0Apache-2.0

Dynamic World Model Runner

DOI

Model files and example notebook for Dynamic World, see [PUBLICATION DOI PENDING REVIEW].

This is not an officially supported Google product.

  • TensorFlow SavedModels for the forward and backward path can be found in ./model/forward and ./model/backward respectively.
  • ./single_image_runner.ipynb gives a step-by-step guide to making predictions with the Dynamic World models.

See the image below for the model block diagram. For Dynamic World, m = 1.5 and b = 2.

Dynamic World neural network architecture