Datasets:

Modalities:
Audio
Text
Formats:
parquet
Libraries:
Datasets
Dask
License:
audio
audioduration (s)
5
5
label
stringclasses
20 values
rewbla
redcro
bewwre
rubpep1
houspa
bewwre
houspa
houspa
rubpep1
rubpep1
gbwwre1
carwre
bewwre
cubthr
norcar
houspa
swathr
cubthr
houspa
eursta
spotow
rucspa1
cubthr
eursta
eursta
sonspa
bncfly
spotow
comrav
redcro
gbwwre1
spotow
houwre
comrav
spotow
redcro
mallar3
redcro
eursta
houwre
cubthr
norcar
swathr
redcro
carwre
comrav
spotow
bewwre
houspa
amerob
carwre
rubpep1
carwre
rubpep1
houspa
bewwre
bewwre
rubpep1
swathr
carwre
houspa
gbwwre1
houwre
rewbla
carwre
swathr
bewwre
swathr
comrav
houspa
eursta
rewbla
swathr
redcro
cubthr
sonspa
barswa
houwre
spotow
redcro
barswa
cubthr
gbwwre1
mallar3
carwre
amerob
spotow
spotow
redcro
carwre
eursta
sonspa
mallar3
mallar3
houspa
houspa
redcro
houspa
barswa
swathr

This dataset contains 5 second clips of birdcalls for audio generation tests.

There are 20 species represented, with ~500 recordings each. Recordings are from xeno-canto.

These clips were taken from longer samples by identifying calls within the recordings using the approach shown here: https://www.kaggle.com/code/johnowhitaker/peak-identification

The audio is represented at 32kHz (mono)

Downloads last month
68
Edit dataset card