`projects.pointnav_baselines.models.point_nav_models`#

`PointNavActorCritic`#

class PointNavActorCritic(VisualNavActorCritic)

Use raw image as observation to the agent

class ResnetTensorPointNavActorCritic(VisualNavActorCritic)

Use resnet_preprocessor to generate observations to the agent

 | @property
 | is_blind() -> bool

True if the model is blind (e.g. neither 'depth' or 'rgb' is an input observation type).

class ResnetTensorGoalEncoder(nn.Module)

 | get_object_type_encoding(observations: Dict[str, torch.FloatTensor]) -> torch.FloatTensor

Get the object type encoding from input batched observations.

class ResnetDualTensorGoalEncoder(nn.Module)

 | get_object_type_encoding(observations: Dict[str, torch.FloatTensor]) -> torch.FloatTensor

Get the object type encoding from input batched observations.