selfclean_audio.datasets.utils

Members

fast_scandir	Quickly scan a directory for files with specified extensions.
resize_waveform	Resize a waveform tensor to the target length by trimming or padding.

selfclean_audio.datasets.utils.resize_waveform(waveform: Tensor, target_length: int)

Resize a waveform tensor to the target length by trimming or padding.

Parameters:

• waveform (torch.Tensor) – Input waveform tensor of shape (channels, length).

• target_length (int) – Desired length of the output waveform.

Returns:

Resized waveform tensor with the specified length.

Return type:

torch.Tensor

selfclean_audio.datasets.utils.fast_scandir(path: str, exts: list[str], recursive: bool = False)

Quickly scan a directory for files with specified extensions. From github.com/drscotthawley/aeiou/blob/main/aeiou/core.py

Parameters:

• path (str) – Directory path to scan.

• exts (list[str]) – List of file extensions to filter (e.g., [‘.wav’, ‘.mp3’]).

• recursive (bool) – If True, scan subdirectories recursively. Defaults to False.

Returns:

A tuple containing a list of subfolder paths and a list of matched file paths.

Return type:

tuple[list[str], list[str]]