pretty_name: Animal Sounds Collection
description: >
A collection of audio recordings for various animal vocalizations, including
birds, dogs, Egyptian fruit bats, giant otters, macaques, orcas, and zebra
finches. Useful for training and evaluating models in bioacoustics, species
classification, and sound event detection.
license: cc-by-4.0
configs:
- config_name: birds
data_files: birds/*
- config_name: dogs
data_files: dogs/*
- config_name: egyptian_fruit_bat
data_files: egyptian_fruit_bat/*
- config_name: egyptian_fruit_bat_extended
data_files: egyptian_fruit_bat/*
- config_name: giant_otter
data_files: giant_otter/*
- config_name: macaque
data_files: macaque/*
- config_name: orca
data_files: orca/*
- config_name: zebra_finch
data_files: zebra_finch/*
Animal Sounds Collection
This dataset contains audio recordings of various animal vocalizations from a range of species, curated to support research in bioacoustics, species classification, and sound event detection. It includes clean and annotated audio samples from the following animals:
- Birds
- Dogs
- Egyptian fruit bats
- Giant otters
- Macaques
- Orcas
- Zebra finches
The dataset is designed to be lightweight and modular, making it easy to explore cross-species vocal behaviors or train models for species-specific tasks.
This work is inspired by the excellent Earth Species Project Library of Datasets, which also offers some additonal datasets.
Birds
This dataset consists of songs and calls from 87 bird species and was originally released as part of the NIPS4B challenge.
- Recordings: 686
- Labels: Multi-label per recording
- Sample rate: 44,100 Hz
- Duration: ~1s to ~5s
- Download: Internet Archive
Dogs
This dataset consists of 693 recordings from 10 adult dogs (ages 3–13), recorded in three contexts: disturbance (e.g., doorbell ringing), isolation, and play.
- Device: DCR VX2000 video recorder with built-in omni-directional mic
- Sample rate: 48,000 Hz (16-bit)
- Download: Preprocessed | Raw
- Citation: Yin, S. & McCowan, B. (2004). Animal Behaviour, 68, 343–355.
Egyptian Fruit Bats
This dataset includes 10,000 recording. They are divided between the egyptian_fruit_bat and egyptian_fruit_bat_extended splits. Another larger sample (~90,000 samples) is available at the Internet Archive.
In this repo, there are 10,000 recordings (1,000 per class), 4.9 GB.
- Sample rate: 250,000 Hz
- Duration: 0.02 to 34 seconds
- Labels: Emitter, addressee, context, and behavior annotations
- Download: Internet Archive – 10k
Giant Otters
This dataset includes 441 recordings across 22 vocalization types, collected between 2009–2011 by Christina A. S. Mumm and Mirjam Knörnschild.
- Microphone: Sennheiser MKH 416-P48U3
- Recorders: Zoom H2 Handy (48 or 96 kHz), with hydrophone (RESON TC4013) for underwater calls
- Bit depth: 24-bit
- Download: Internet Archive
- Citation: Mumm, C.A.S. & Knörnschild, M. (2016). PLOS ONE. DOI
Macaques
This dataset contains 7,285 macaque coo calls from 8 individuals (4 males, 4 females), annotated at the individual level.
- Sample rate: 5,756 files at 24,414 Hz, 1,529 at 44,100 Hz
- Duration: 0.04s to 1.69s
- Download: Internet Archive
- Citation: PMC Article
Orcas
This dataset comes from the Orcasound Project and contains two versions:
- Classification version: 4-second clips, 398 positive and 196 negative examples
- Sample rate: 44,100 Hz
- Download: Internet Archive
Zebra Finches
Collected between 2011–2014 by Julie E. Elie at UC Berkeley using zebra finches from the Theunissen Lab colony.
- Recordings: 3,405 across 11 call types
- Subjects: Adults and ~30-day-old juveniles
- Labels: Call type, individual, age
- Sample rate: Not specified
- Download: Figshare
- Citations:
Citation
If you use this collection, please cite the original sources provided in each section above.