BirdNET-Go v0.4.4

• BirdNET meta model support for location based species filtering
• BirdWeather support
• Species specific threshold support
• Lots of small fixes

Archive contains pre-compiled libtensorflow_c.so library, copy this file on Linux systems to /usr/local/lib and run ldconfig.
For Windows version libtensorflowlite_c.so needs to be in a directory included in PATH.

Please see detailed documentation at https://github.com/tphakala/birdnet-go/wiki

gobirdnet_v0.4.4_linux_arm64 is tested to run on Raspberry Pi 3B+, 4 with Bulleseye and 5 with Bookworm
gobirdnet_v0.4.4_linux_amd64 is tested to run on Debian 11 and Ubuntu 23.10 Intel
gobirdnet_v0.4.4_windows_amd64 is tested to run on Windows 11 Intel

Go-BirdNET v0.3.0

• Supports 48khz WAV file processing
• Supports realtime audio processing from soundcard source
• BirdNET tflite model and labels are now embedded in binary reducing runtime dependencies
• Two new output types, Raven table which is default and csv, output is stored in directory specified by --output or -o flag
• Directory processing mode, "birdnet directory my-sound-files" processes all wav files in that directory, supports recursing with --recursive or -r flag
• Set prediction confidence threshold with --threshold flag, value range 0.0 - 1.0

Archive contains pre-compiled libtensorflow_c.so library, copy this file on Linux systems to /usr/local/lib and run ldconfig.
For Windows version libtensorflowlite_c.so needs to be in a directory included in PATH.
For macOS version copy libtensorflowlite_c.dylib to /usr/local/lib.

gobirdnet_v0.3.0_linux_arm64 is tested to run on Raspberry Pi 4 with 64-bit Raspberry Pi OS Bullseye
gobirdnet_v0.3.0_linux_amd64 is tested to run on Debian 11 and Ubuntu 23.10 Intel
gobirdnet_v0.3.0_windows_amd64 is tested to run on Windows 11 Intel
gobirdnet_v0.3.0_darwin_amd64 is tested to run on macOS 14.1 Intel

Go-BirdNET v0.2.1

Supports WAV file analysis and real time audio capture analysis.

Real time capture detections are logged in ./log/detections.log file which can be used for example as chatlog overlay for birdfeeder streams in OBS. Real time mode also saves detections exceeding set confidence threshold as WAV files to ./clips directory.

File analysis supports only 48000 Hz input files, bit depth can be 16, 24 or 32 bit integer type.

Archive contains pre-compiled libtensorflow_c.so library which should be copied on Linux systems to system LDPATH, on Windows it works from directory you extracted birdnet.exe to.