audtorch

Deep learning with PyTorch and audio.

Documentation: https://audtorch.readthedocs.io

If you are interested in PyTorch and audio you should also check out the efforts to integrate more audio directly into PyTorch:

• pytorch/audio
• keunwoochoi/torchaudio-contrib

Installation

audtorch is supported by Python 3.5 or higher. To install it run (preferable in a virtual environment):

pip install audtorch

Usage

audtorch automates the data iteration process for deep neural network training using PyTorch. It provides a set of feature extraction transforms that can be implemented on-the-fly on the CPU.

The following example creates a data set of speech samples that are cut to a fixed length of 10240 samples. In addition they are augmented on the fly during data loading by a transform that adds samples from another data set:

>>> import sounddevice as sd
>>> from audtorch import datasets, transforms
>>> noise = datasets.WhiteNoise(duration=10240, sampling_rate=16000)
>>> augment = transforms.Compose([transforms.RandomCrop(10240),
...                               transforms.RandomAdditiveMix(noise)])
>>> data = datasets.LibriSpeech(root='~/LibriSpeech', sets='dev-clean',
...                             download=True, transform=augment)
>>> signal, label = data[8]
>>> sd.play(signal.transpose(), data.sampling_rate)

Besides data sets and transforms the package provides standard evaluation metrics, samplers, and necessary collate functions for training deep neural networks for audio tasks.