This repository provides tools for automated bird species detection from audio recordings and for generating quality control reports. It supports local processing and API-based workflows and is built on an AI model developed at the University of Jyväskylä as part of the Muuttolintujen kevät -project.
shared/— Shared code: classifier and functionsapi/— HTTP API (FastAPI)cli/— Command-line scripts for local analysis and report generationapp/— Desktop application
- Analyzes audio recordings (WAV, MP3, FLAC) to detect bird species, either locally or via API
- Files are divided into smaller chunks based on chunk_size parameter, because the model has a limit on the input size. There chunks are then divided into segments of clip_dur seconds (currently fixed to 3 seconds), and overlap parameter defines how much of each segment is overlapped with the next segment.
- Uses species distribution and temporal modeling to improve detection accuracy
- Handles batch processing of multiple audio files
- Generates reports with species statistics and sample audio clips to help verifying the results
- Desktop application for users who prefer a graphical interface
- Docker
- Git
- The following AI models:
- BirdNET, included in this repository
- BSG – Finnish Birds Model in keras format, available at https://github.com/plauha/BSG_classifier_builder/tree/main/Run%20BSG%20models/models/Finland
git clonecd bird-identification- Place models to the
/modelsfolder: BirdNET and Muuttolintujen kevät docker compose up --build # start both cli and apidocker compose up cli --build # start only clidocker compose up api --build # start only api- Desktop app has another setup, see app/README.md
From the host machine:
docker compose run --rm test
- For command line tool, see cli/README.md
- For API, see api/README.md
- For desktop app, see app/README.md