How birdnet-team/birdnet works
BirdNET is a Python library for acoustic bird identification, with a companion location-and-time species-presence capability. The primary workflow loads an acoustic model and backend, then predicts from audio files or arrays; users can also extract embeddings. Acoustic sessions coordinate processing and return prediction or encoding results. The README documents model-family, custom-classifier, multilingual, and export capabilities. Several implementation details are represented at subsystem level because their source was not sampled; edges are limited to README statements or relationships shown in indexed/read source.
Main entry points: acoustic/models Acoustic models · session.py Acoustic session · model_loader.py Model loader · geo/models Geo models
Components 20
Public API and loading
- Python API
src/birdnet/__init__.py - Model loader
src/birdnet/model_loader.py
Acoustic analysis
- Acoustic session
src/birdnet/acoustic/inference/session.py - Score prediction
src/birdnet/acoustic/inference/core/prediction/prediction_result.py - Embedding extraction
src/birdnet/acoustic/inference/core/encoding/encoding_result.py - Audio utilities
src/birdnet/utils/audio_files.py - Result export
src/birdnet/acoustic/inference/core/result_base.py
Inference runtime
- Process manager
src/birdnet/acoustic/inference/process_manager.py - Inference workers
src/birdnet/acoustic/inference/core/worker.py - Runtime backends
src/birdnet/core/backends.py
Models and metadata
- Acoustic models
src/birdnet/acoustic/models - Custom classifiers
src/birdnet/acoustic/models/custom_tflite.py - Species names
src/birdnet/utils/species_lists.py - Model downloads
src/birdnet/utils/downloads.py
Geographic prediction
- Geo models
src/birdnet/geo/models - Geo session
src/birdnet/geo/inference/session.py - Presence result
src/birdnet/geo/inference/prediction_result.py
Other components
- Library user
- ML runtimes
- Audio files
Connections 22
- Library user → Python API: calls
README.md - Python API → Model loader: loads models
src/birdnet_benchmark/cli.py - Model loader → Acoustic models: selects
src/birdnet/model_loader.py - Model loader → Geo models: selects
src/birdnet/model_loader.py - Model loader → Custom classifiers: loads custom
src/birdnet/model_loader.py - Acoustic models → Acoustic session: runs prediction
src/birdnet/acoustic/models/v3_0/model.py - Acoustic models → Acoustic session: extracts embeddings
src/birdnet/acoustic/models/v3_0/model.py - Acoustic session → Process manager: starts processing
src/birdnet/acoustic/inference/session.py - Process manager → Inference workers: coordinates workers
src/birdnet/acoustic/inference/process_manager.py - Inference workers → Runtime backends: uses backend
src/birdnet/acoustic/inference/core/worker.py - Runtime backends → ML runtimes: executes model
src/birdnet/core/backends.py - Audio files → Acoustic models: provides audio
README.md - Acoustic session → Score prediction: returns scores
src/birdnet/acoustic/inference/session.py - Acoustic session → Embedding extraction: returns embeddings
src/birdnet/acoustic/inference/session.py - Geo models → Geo session: predicts presence
src/birdnet/geo/models/base.py - Geo session → Presence result: returns result
src/birdnet/geo/inference/session.py - Acoustic models → Model downloads: gets official models
src/birdnet/acoustic/models/v3_0/model.py - Acoustic models → Species names: uses labels
src/birdnet/acoustic/models/v3_0/model.py - Geo models → Species names: uses species list
src/birdnet/geo/models/base.py - Score prediction → Result export: serializes result
src/birdnet/acoustic/inference/core/prediction/prediction_result.py - Embedding extraction → Result export: serializes result
src/birdnet/acoustic/inference/core/encoding/encoding_result.py - Audio files → Audio utilities: supports formats
README.md