RandomMasking#

class torch_ecg.augmenters.RandomMasking(fs: int, mask_value: Real = 0.0, mask_width: Sequence[Real] = [0.08, 0.18], prob: Sequence[Real] | Real = [0.3, 0.15], inplace: bool = True, **kwargs: Any)[source]#

Bases: Augmenter

Randomly mask ECGs with a probability.

Parameters:

  • fs (int) – Sampling frequency of the ECGs to be augmented.

  • mask_value (numbers.Real, default 0.0) – Value to mask with.

  • mask_width (Sequence[numbers.Real], default [0.08, 0.18]) – Width range of the masking window, with units in seconds

  • prob (numbers.Real or Sequence[numbers.Real], default [0.3, 0.15]) – Probabilities of masking ECG signals, the first probality is for the batch dimension, the second probability is for the lead dimension. Note that 0.15 is approximately the proportion of QRS complexes in ECGs.

  • inplace (bool, default True) – Whether to mask inplace or not.

  • kwargs (dict, optional) – Additional keyword arguments.

Examples

rm = RandomMasking(fs=500, prob=0.7)
sig = torch.randn(32, 12, 5000)
critical_points = [np.arange(250, 5000 - 250, step=400) for _ in range(32)]
sig, _ = rm(sig, None, critical_points=critical_points)
extra_repr_keys() → List[str][source]#

Extra keys for __repr__() and __str__().

forward(sig: Tensor, label: Tensor | None, *extra_tensors: Sequence[Tensor], critical_points: Sequence[Sequence[int]] | None = None, **kwargs: Any) → Tuple[Tensor, ...][source]#

Forward method of the RandomMasking augmenter.

Parameters:

  • sig (torch.Tensor) – Batched ECGs to be augmented, of shape (batch, lead, siglen).

  • label (torch.Tensor) – Label tensor of the ECGs. Not used, but kept for compatibility with other augmenters.

  • extra_tensors (Sequence[torch.Tensor], optional) – Not used, but kept for consistency with other augmenters.

  • critical_points (Sequence[Sequence[int]], optional) – If given, random masking will be performed in windows centered at these points. This is useful for example when one wants to randomly mask QRS complexes.

  • kwargs (dict, optional) – Not used, but kept for consistency with other augmenters.

Returns:

  • sig (torch.Tensor) – The augmented ECGs, of shape (batch, lead, siglen).

  • label (torch.Tensor) – Label tensor of the augmented ECGs, unchanged.

  • extra_tensors (Sequence[torch.Tensor], optional) – Unchanged extra tensors.