urbanlc.model.pipeline_transforms.MSSTransformer.transform#

MSSTransformer.transform(img: Tensor, mask: Union[None, Tensor], is_training: Optional[bool] = True, p_hflip: Optional[float] = 0.5, p_vflip: Optional[float] = 0.5, p_mix_patch: Optional[float] = 1.0, repeat: Optional[int] = 1) → Tuple[Tensor, Union[None, Tensor]]#

Apply transformations to the input Landsat image and land cover.

Parameters:

  • img (torch.Tensor) – Input image tensor.

  • mask (Optional[torch.Tensor]) – Land cover tensor.

  • is_training (Optional[bool]) – Flag indicating whether the transformation is applied during training.

  • p_hflip (Optional[float]) – Probability of horizontal flip.

  • p_vflip (Optional[float]) – Probability of vertical flip.

  • p_mix_patch (Optional[float]) – Probability of mixing patches.

  • repeat (Optional[int]) – Number of times to repeat the transformations.

Returns:

Transformed image and land cover tensors.

Return type:

Tuple[torch.Tensor, Union[None, torch.Tensor]]