Models

V2.4, June 2023

  • more than 6,000 species worldwide

  • covers frequencies from 0 Hz to 15 kHz with two-channel spectrogram (one for low and one for high frequencies)

  • 0.826 GFLOPs, 50.5 MB as FP32

  • enhanced and optimized metadata model

  • global selection of species (birds and non-birds) with 6,522 classes (incl. 11 non-event classes)

Technical Details

  • 48 kHz sampling rate (we up- and downsample automatically and can deal with artifacts from lower sampling rates)

  • we compute 2 mel spectrograms as input for the convolutional neural network:

    • first one has fmin = 0 Hz and fmax = 3000; nfft = 2048; hop size = 278; 96 mel bins

    • second one has fmin = 500 Hz and fmax = 15 kHz; nfft = 1024; hop size = 280; 96 mel bins

  • both spectrograms have a final resolution of 96x511 pixels

  • raw audio will be normalized between -1 and 1 before spectrogram conversion

  • we use non-linear magnitude scaling as mentioned in Schlüter 2018

  • V2.4 uses an EfficienNetB0-like backbone with a final embedding size of 1024

  • See this comment for more details

Species range model V2.4 - V2, Jan 2024

  • updated species range model based on eBird data

  • more accurate (spatial) species range prediction

  • slightly increased long-tail distribution in the temporal resolution

  • see this discussion post for more details

Using older models

Older models can also be used as custom classifiers in the GUI or using the –classifier argument in the birdnet_analyzer.analyze command line interface.

Just download your desired model version and unzip.

  • GUI: Select the *_Model_FP32.tflite file under Species selection / Custom classifier

  • CLI: python -m birdnet_analyzer ... --classifier 'path_to_Model_FP32.tflite'

Model Version History

Note

All models listed here are licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0).

V2.4

  • more than 6,000 species worldwide

  • covers frequencies from 0 Hz to 15 kHz with two-channel spectrogram (one for low and one for high frequencies)

  • 0.826 GFLOPs, 50.5 MB as FP32

  • enhanced and optimized metadata model

  • global selection of species (birds and non-birds) with 6,522 classes (incl. 11 non-event classes)

  • Download here: BirdNET-Analyzer-V2.4.zip

V2.3

  • slightly larger (36.4 MB vs. 21.3 MB as FP32) but smaller computational footprint (0.698 vs. 1.31 GFLOPs) than V2.2

  • larger embedding size (1024 vs 320) than V2.2 (hence the bigger model)

  • enhanced and optimized metadata model

  • global selection of species (birds and non-birds) with 3,337 classes (incl. 11 non-event classes)

  • Download here: BirdNET-Analyzer-V2.3.zip

V2.2

  • smaller (21.3 MB vs. 29.5 MB as FP32) and faster (1.31 vs 2.03 GFLOPs) than V2.1

  • maintains same accuracy as V2.1 despite more classes

  • global selection of species (birds and non-birds) with 3,337 classes (incl. 11 non-event classes)

  • Download here: BirdNET-Analyzer-V2.2.zip

V2.1

  • same model architecture as V2.0

  • extended 2022 training data

  • global selection of species (birds and non-birds) with 2,434 classes (incl. 11 non-event classes)

  • Download here: BirdNET-Analyzer-V2.1.zip

V2.0

  • same model design as 1.4 but a bit wider

  • extended 2022 training data

  • global selection of species (birds and non-birds) with 1,328 classes (incl. 11 non-event classes)

  • Download here: BirdNET-Analyzer-V2.0.zip

V1.4

  • smaller, deeper, faster

  • only 30% of the size of V1.3

  • still linear spectrogram and EfficientNet blocks

  • extended 2021 training data

  • 1,133 birds and non-birds for North America and Europe

  • Download here: BirdNET-Analyzer-V1.4.zip

V1.3

  • Model uses linear frequency scale for spectrograms

  • uses V2 fusion blocks and V1 efficient blocks

  • extended 2021 training data

  • 1,133 birds and non-birds for North America and Europe

  • Download here: BirdNET-Analyzer-V1.3.zip

V1.2

  • Model based on EfficientNet V2 blocks

  • uses V2 fusion blocks and V1 efficient blocks

  • extended 2021 training data

  • 1,133 birds and non-birds for North America and Europe

  • Download here: BirdNET-Analyzer-V1.2.zip

V1.1

  • Model based on Wide-ResNet (aka “App model”)

  • extended 2021 training data

  • 1,133 birds and non-birds for North America and Europe

  • Download here: BirdNET-Analyzer-V1.1.zip

App Model