Skip to content

FAQ#

What is acoupi-birdnet?#

acoupi-birdnet is the implementation of the latest AI Bioacoustics Classifier BirdNET-Analyzer using acoupi Python toolkit.

What is BirdNET-Analyzer?#

BirdNET-Analyzer is a deep-learning model to detect and classify birds vocalizations. The model has been developed by the K. Lisa Yang Center for Conservation Bioacoustics at the Cornell Lab of Ornithology in collaboration with Chemnitz University of Technology.

What is birdnetlib?#

acoupi-birdnet uses the python API provided by the python package birdnetlib to access the BirdNET-Analyzer model.

Can the BirNET model classify any bird species?#

The BirdNET model is updated regularly. The V2.4 can recognized more than 6,000 species worldwide. The BirdNET model is used widely but as with any deep-learning model, the model can predict false-positives and false-negatives.

For who is acoupi-birdnet?#

acoupi-birdnet is itended for researchers, practioners, and individuals interested in recording and classifying birds species on devices.

Can I configure acoupi-birdnet?#

Yes. Users can customised the configuration parameters of acoupi-birdnet to suit their own needs. See tutorials/configuration to learn more about the configuration options.

What are the requirements to use acoupi-birdnet?#

To use acoupi-birdnet you will need the following hardware:

  • a Raspberry Pi 4
  • an SD Card (32GB or 64GB) with RaspbiOS-Arm64-Lite installed.
  • a USB microphone such as the AudioMoth or a Lavalier microhpon.

Where can I found more information about BirdNET?#

  1. The BirdNET-Analyzer GitHub repository contains a lot of information about the model.

  2. The publication "BirdNET: A deep learning solution for avian diversity monitoring" (Kahl S., et al., 2021) is also a great resource to learn more about the architecture of the model.

  3. The research article "Guidelines for appropriate use of BirdNET scores and other detector outputs" (Wood C.M and Kahl S., 2024) provides insightful information about how to understand and treat the detections scores from the model.