torch_ecg.utils.generate_weight_mask#

torch_ecg.utils.generate_weight_mask(target_mask: ndarray[tuple[Any, ...], dtype[_ScalarT]], fg_weight: float | int, fs: float | int, reduction: float | int, radius: float | int, boundary_weight: float | int, plot: bool = False) → ndarray[tuple[Any, ...], dtype[_ScalarT]][source]#

Generate weight mask for a binary target mask, accounting the foreground weight and boundary weight.

Parameters:

target_mask (numpy.ndarray) – The target mask, assumed to be 1D and binary.
fg_weight (float or int) – Foreground (value 1) weight, usually > 1.
fs (float or int) – Sampling frequency of the signal.
reduction (float or int) – Reduction ratio of the mask w.r.t. the signal.
radius (float or int) – Radius of the boundary, with units in seconds.
boundary_weight (float or int) – Weight for the boundaries (positions where values change) of the target map.
plot (bool, default False) – If True, target_mask and the result weight_mask will be plotted.

Returns:

weight_mask – Weight mask generated from target_mask.

Return type:

numpy.ndarray

Examples

>>> target_mask = np.zeros(50000, dtype=int)
>>> target_mask[500:14000] = 1
>>> target_mask[35800:44600] = 1
>>> fg_weight = 2.0
>>> fs = 500
>>> reduction = 1
>>> radius = 0.8
>>> boundary_weight = 5.0
>>> weight_mask = generate_weight_mask(
...     target_mask, fg_weight, fs, reduction, radius, boundary_weight
... )
>>> weight_mask.shape
(50000,)
>>> reduction = 10
>>> weight_mask = generate_weight_mask(
...     target_mask, fg_weight, fs, reduction, radius, boundary_weight
... )
>>> weight_mask.shape
(5000,)