Song Data #
The songs directory contains audio recordings, along with metadata for each
recording. The dataset is split into four zip files:
songs
├── song-files.zip.part1
├── song-files.zip.part2
├── song-files.zip.part3
└── song-files.zip.part4
These files can be stitched together before unzipping using the command cat
song-files.zip.part* > sf.zip && zip -FF sf.zip --out song-files.zip, or your
preferred method if you are not on a Unix-like system.
When you unzip the file, you will see the following structure:
song-files
├── JSON
│ ├── file.JSON
│ └── ...
└── WAV
├── 1_20190501_000000_0.wav
└── ...
Dataset Size #
There are 109,963 files in each of the WAV and JSON folders. The total size of the dataset is 11.4 GB.
File Format #
The dataset is provided in ZIP format, with two folders: WAV and JSON. The
WAV folder contains the audio recordings in WAV format, and the JSON folder
contains the metadata for each recording in JSON format. The filenames match
between the two folders-only the extension changes.
The WAV files are mono 16-bit PCM audio files with a sample rate of 22,050 Hz.
The files are named according to the following convention:
<ID>_<YYYYMMDD_HHMMSS>_<start frame>.wav
Where ID is the unique identifier for the recording, YYYYMMDD_HHMMSS is the
datetime of the recording, and start frame is the
start frame of the recording in the original audio file.
The JSON files contain metadata for each recording. The files are named
according to the same convention, <ID>_<YYYYMMDD_HHMMSS>_<start frame>.JSON, and
each file contains the following fields:
| Key | Description |
|---|---|
| sample_rate | The sample rate of the audio |
| bit_rate | The bit rate of the audio |
| length_s | The duration of the audio segment |
| ID | The unique identifier for the audio segment |
| label | The label associated with the audio segment |
| start | The start position of the audio segment |
| end | The end position of the audio segment |
| lower_freq | The lower frequency bound of the audio segment |
| upper_freq | The upper frequency bound of the audio segment |
| max_amplitude | The maximum amplitude of the audio segment |
| min_amplitude | The minimum amplitude of the audio segment |
| source_wav | The file path to the original source WAV file |
| annotation_file | The file path to the XML annotation file |
| wav_file | The file path to the segmented WAV file |
| class_id | The class identifier for the audio segment |
| datetime | The date and time of the recording |
| onsets | A list of onset times in seconds |
| offsets | A list of offset times in seconds |
| silence_durations | A list of silence durations in seconds |
| unit_durations | A list of unit durations in seconds |