madmom

The madmom.features module provides a collection of submodules for extracting various musical features from audio. Depending on your analysis needs, you can use specific submodules for:

• Beats: Beat tracking via features.beats, features.beats_crf, or features.beats_hmm.
• Chords: Chord recognition via features.chords.
• Downbeats: Downbeat detection via features.downbeats.
• Key: Key detection via features.key.
• Notes: Note detection via features.notes or features.notes_hmm.
• Onsets: Onset detection via features.onsets.
• Tempo: Tempo estimation via features.tempo.

The madmom.io module serves as the primary entry point for handling input and output operations within the madmom library. It is organized into specialized submodules for different media formats:

• madmom.io.audio: For audio-related I/O operations.
• madmom.io.midi: For MIDI-related I/O operations.

Use these submodules to load, save, or manipulate audio and MIDI data as required by your analysis pipeline.

Before installing, ensure you have Python 2.7 or Python 3.5+ and the following core packages installed:

• numpy
• scipy
• cython
• mido

Optional dependencies:

• For testing, improved FFT performance, or live audio: pytest, pyaudio, pyfftw.
• For audio file support beyond .wav (44.1kHz, 16-bit): ffmpeg (avoid avconv on Ubuntu due to decoding bugs).

To use madmom as a library or to run its bundled programs without modifying the source code, install it via pip. This method automatically installs dependencies and includes trained models.

To install globally (may require sudo):

pip install madmom

To install locally for the current user:

pip install --user madmom

Note on Executables: If you use --user, the executable programs are installed to a local directory (e.g., ~/.local/bin on Ubuntu or ~/Library/Python/X.X/bin on macOS). If these are not in your $PATH, you must add them manually:

export PATH='path/to/scripts':$PATH

To upgrade, use the same method used for the initial installation.

Upgrading via pip:

pip install --upgrade madmom [--user]

Note: If models or programs have changed, it is safer to uninstall and reinstall:

pip uninstall madmom
pip install madmom [--user]

Upgrading from source:

1. Pull latest code: git pull
2. Update models: git submodule update
3. If .pyx or .pxd files changed, recompile with Cython: python setup.py build_ext --inplace

If you intend to develop or modify madmom, clone the repository recursively to ensure the pre-trained model/data files (included as Git submodules) are downloaded.

Step 1: Clone the repository

git clone --recursive https://github.com/CPJKU/madmom.git

Alternatively, if you already cloned without submodules:

git clone https://github.com/CPJKU/madmom.git
cd madmom
git submodule update --init --remote

Step 2: Install in development mode

python setup.py develop --user

Step 3: Run tests

python setup.py pytest

The madmom package provides standalone reference implementations of its algorithms as executable programs located in the /bin folder. These programs support several execution modes depending on your needs:

• single: Processes a single audio file and writes the output to STDOUT or a specified output file.
• batch: Processes multiple audio files and writes outputs to files with a specified suffix. If no output directory is provided, outputs are saved in the same directory as the input files.
• online: Operates on live audio signals. This mode requires pyaudio to be installed.
• pickle: Used to store parameters to ensure exact reproduction of experiments.

You can access help messages for the main program or for specific modes using the -h flag.

The madmom tutorials are provided as a collection of Jupyter (IPython) notebooks. You can access them online at the dedicated tutorials repository to learn how to use the package through interactive examples.

https://github.com/CPJKU/madmom_tutorials

Before installing madmom, ensure you have Python 2.7 or Python 3.5 or newer installed.

Required dependencies:

• numpy
• scipy
• cython
• mido

Recommended for testing, live audio, or FFT performance:

• pytest
• pyaudio
• pyfftw

Audio format support: If you need to support audio files other than .wav (specifically 44.1kHz, 16-bit depth), you must install ffmpeg. Note: avconv on Ubuntu Linux is discouraged due to decoding bugs.

Use the same method (pip or source) that was used for the initial installation.

Upgrading via pip:

pip install --upgrade madmom [--user]

If programs or models have changed significantly, it is safer to uninstall and reinstall:

pip uninstall madmom
pip install madmom [--user]

Upgrading from source:

1. Pull latest code:

   git pull

2. Update models in the submodule:
 ```bash
git submodule update

3. If .pyx or .pxd files were changed, recompile the modules:

   python setup.py build_ext --inplace

To extend the library or access its full functionality, it is preferred to install madmom from source. Once installed, you can import it into your Python scripts alongside dependencies like numpy.

import madmom
import numpy as np