Audio Loader

This FiftyOne plugin is a Python plugin that allows for you to load in your audio datasets as spectograms in FiftyOne!

🎧 Explore a whole new modality with FiftyOne!

Installation

fiftyone plugins download https://github.com/danielgural/audio_loader

Operators

load_audio

Loads audio files in that are saved as ".wav" files based on classification directory tree. Such as:

<dataset_dir>/
    <classA>/
        <audio1>.wav
        <audio2>.wav
        ...
    <classB>/
        <audio1>.wav
        <audio2>.wav
        ...
    ...

When loading, spectograms will be created for each audio file and saved to the output directory. For example, if input directory = output directory:

<dataset_dir>/
    <classA>/
        <audio1>.wav
        <audio2>.wav
        ...
    <classsA_spectograms>/
        <spectogram1>.png
        <spectogram2>.png
        ...
    <classB>/
        <audio1>.wav
        <audio2>.wav
        ...
    <classsB_spectograms>/
        <spectogram1>.png
        <spectogram2>.png
        ...
    ...

Additionally, frame_rate, duration, channels, sample_width, and total_frame_count will be added as fields to the sample as well!