box_embeddings.modules.regularization.regularizer

Module Contents

class BoxRegularizer(weight: float, log_scale: bool = True, reduction: str = 'sum', **kwargs: Any)

Bases: torch.nn.Module, box_embeddings.common.registrable.Registrable

Base box-regularizer class

forward(self, box_tensor: box_embeddings.parameterizations.box_tensor.BoxTensor) → Union[float, torch.Tensor]

Calls the _forward and multiplies the weight

Parameters

box_tensor – Input box tensor

Returns

scalar regularization loss