vak.transforms.transforms.FramesStandardizer

class vak.transforms.transforms.FramesStandardizer(mean_freqs=None, std_freqs=None, non_zero_std=None)

Bases: object

transform that standardizes spectrograms so they are all on the same scale, by subtracting off the mean and dividing by the standard deviation from a ‘fit’ set of spectrograms.

mean_freqs

mean values for each row across the fit set of spectrograms

Type:

numpy.ndarray

std_freqs

standard deviation for each row across the fit set of spectrograms

Type:

numpy.ndarray

non_zero_std

boolean, indicates where std_freqs has non-zero values. Used to avoid divide-by-zero errors.

Type:

numpy.ndarray

__init__(mean_freqs=None, std_freqs=None, non_zero_std=None)

initialize a new StandardizeSpect instance

Parameters:

• mean_freqs (numpy.ndarray) – vector of mean values for each row across the fit set of spectrograms

• std_freqs (numpy.ndarray) – vector of standard deviations for each row across the fit set of spectrograms

• non_zero_std (numpy.ndarray) – boolean, indicates where std_freqs has non-zero values. Used to avoid divide-by-zero errors.

Methods

__init__([mean_freqs, std_freqs, non_zero_std])	initialize a new StandardizeSpect instance
fit(spect)	Fit a StandardizeSpect instance.
fit_dataset_path(dataset_path[, split, subset])	Returns a StandardizeSpect instance that is fit to a split from a dataset, given the path to that dataset and the name of the split.
fit_inputs_targets_csv_path(...[, split, ...])

classmethod fit(spect)

Fit a StandardizeSpect instance.

Parameters:

spect (numpy.ndarray) – with dimensions (frequency bins, time bins)

Notes

Input should be spectrogram. Fit function finds the mean and standard deviation of each row, which are used by transform method to scale other spectrograms.

classmethod fit_dataset_path(dataset_path, split='train', subset: str | None = None)

Returns a StandardizeSpect instance that is fit to a split from a dataset, given the path to that dataset and the name of the split.

Parameters:

• dataset_path (str or pathlib.Path) – Path to a dataset.

• split (str) – Name of split from dataset to fit.

Returns:

standardize_spect – Instance that has been fit to input data from split.

Return type:

StandardizeSpect