This is an experimental and unofficial module add-on for the new fastai v1 library. It adds the capability of audio classification to fastai by loading raw audio files and generating spectrograms on the fly. Please check out the notebooks directory for usage examples.
The accompanying article can be found here: Audio Classification using FastAI and On-the-Fly Frequency Transforms
The fastai library is currently being developed rapidly, so this repo may quickly go out of date with new versions.
- python 3.6
- fastai 1.0.39
- librosa 0.6.2
This repo was also heavily inspired by torchaudio, especially transforms.py.