modulation.data.audio.speech.seperation.dsd100.DSD100

class modulation.data.audio.speech.seperation.dsd100.DSD100(root: str, transforms, crop_size=None, use_cuda=False)

Bases: Dataset

DSD100 dataset for audio source separation (2-stem) :param root: Path to the directory where the dataset is found. :param transforms: audiomentations transform object for waveform transform. Must end with Spectrogram transform :param use_cuda: If true, will move tensor to cuda before applying transforms. (default: False)

__init__(root: str, transforms, crop_size=None, use_cuda=False) → None

Methods

__init__(root, transforms[, crop_size, use_cuda])