groover 0.0.3

A rhythm feature extractor and classifier for MIDI files

groover 0.0.3

Installation

groover is a beat-by-beat rhythm feature clustering and token generation tool for .mid files. You can download groover using pip:

pip install groover

To check if groover is successfully installed, type python in the terminal, and do the following:

>>> from groover import RhythmKMeans
>>> type(RhythmKMeans())
<class 'groover.classifier.RhythmKMeans'>

Documentation

data

get_heat_maps(midi_obj, beat_resolution=480, rid_melody=False, min_pitch=0, max_pitch=127)

Returns a numpy array of shape (n, 24), where n is the number of beats in midi_obj. Each row is the rhythmic heat map of a beat, taking into consideration the notes' velocity and pitch.

Parameters

• midi_obj: miditoolkit.midi.parser.MidiFile
  • the midi object to get heat maps from
• beat_resolution: int
  • the number of ticks per beat
• rid_melody: bool
  • whether to ignore melody notes when calculating rhythmic intensity
• min_pitch: int
  • the minimum pitch of the note to be considered when calculating rhythmic intensity
• max_pitch: int
  • the maximum pitch of the note to be considered when calculating rhythmic intensity

get_dataset(midi_objs, beat_resolution=480, rid_melody=False, min_pitch=0, max_pitch=127)

Returns a numpy array of shape (n, 24), where n is the total number of beats of midi objects in midi_objs. Each row is the rhythmic heat map of a beat, taking into consideration the notes' velocity and pitch.

Parameters

• midi_obj: list
  • the list containing midi objects to get heat maps from
• beat_resolution: int
  • the number of ticks per beat
• rid_melody: bool
  • whether to ignore melody notes when calculating rhythmic intensity
• min_pitch: int
  • the minimum pitch of the note to be considered when calculating rhythmic intensity
• max_pitch: int
  • the maximum pitch of the note to be considered when calculating rhythmic intensity

RhythmKMeans

RhythmKMeans classifies rhythmic heat maps and use them to predict and evaluate rhythmic tokens.

RhythmKMeans.__init__(self, cluster_centers=None)

Parameters

• cluster_centers: numpy.ndarray
  • the cluster centers in the shape of (k, 24), where k is the number of clusters and each row is a cluster.

RhythmKMeans.load_cluster_centers(self, cluster_centers)

Loads cluster_centers as the classifier's cluster centers.

Parameters

• cluster_centers: numpy.ndarray
  • the cluster centers in the shape of (k, 24), where k is the number of clusters and each row is a cluster.

RhythmKMeans.fit(self, dataset, k, max_iter=1000, epsilon=1e-6)

Makes the classifier's cluster centers align with the dataset.

Parameters

• dataset: numpy.ndarray
  • a numpy array of shape (n, 24), where n is the total number of beats in the dataset, with each row being the rhythmic heat map of a beat
• k: int
  • the number of clusters to be generated
• max_iter: int
  • the maximum number of iterations to perform
• epsilon: float
  • if the average distance of the cluster centers between iterations is lower than epsilon, clustering ends early

RhythmKMeans.k(self)

Returns the number of clusters of the classifier.

RhythmKMeans.is_empty(self)

Returns True if the classifier is not fitted to any data yet, False otherwise.

RhythmKMeans.add_beat_clusters(self, midi_obj, beat_resolution=480, preprocessing='default', min_pitch=0, max_pitch=127)

Adds markers with rhythm types to midi_obj.

Parameters

• midi_obj: miditoolkit.midi.parser.MidiFile
  • the midi object to add beat-by-beat rhythm markers to
• beat_resolution: int
  • the number of ticks per beat
• preprocessing: str
  • can be either 'default', 'binary', or 'quantized', which will then change the rhythmic heat maps' values accordingly
• min_pitch: int
  • the minimum pitch of the note to be considered when calculating rhythmic intensity
• max_pitch: int
  • the maximum pitch of the note to be considered when calculating rhythmic intensity

RhythmKMeans.get_rhythm_scores(self, midi_obj, beat_resolution=480, min_pitch=0, max_pitch=127)

Returns a tuple of numpy arrays. The first is the rhythm types in shape (n,) that is specified by the markers in the midi object, and the second array is the alignment score between the notes and the rhythm type in shape (n,). n is the number of beats in the midi object.

Parameters

• midi_obj: miditoolkit.midi.parser.MidiFile
  • the midi object to be evaluated
• beat_resolution: int
  • the number of ticks per beat
• preprocessing: str
  • can be either 'default', 'binary', or 'quantized', which will then change the rhythmic heat maps' values accordingly
• min_pitch: int
  • the minimum pitch of the note to be considered when calculating rhythmic intensity
• max_pitch: int
  • the maximum pitch of the note to be considered when calculating rhythmic intensity