Source code for mirdata.datasets.medleydb_melody

"""MedleyDB melody Dataset Loader

.. admonition:: Dataset Info
    :class: dropdown

    MedleyDB melody is a subset of the MedleyDB dataset containing only
    the mixtures and melody annotations.

    MedleyDB is a dataset of annotated, royalty-free multitrack recordings.
    MedleyDB was curated primarily to support research on melody extraction,
    addressing important shortcomings of existing collections. For each song
    we provide melody f0 annotations as well as instrument activations for
    evaluating automatic instrument recognition.

    For more details, please visit:


import csv
import json
import logging
import os
from typing import BinaryIO, cast, Optional, TextIO, Tuple

import librosa
import numpy as np

from mirdata import download_utils
from mirdata import jams_utils
from mirdata import core
from mirdata import annotations
from mirdata import io

BIBTEX = """@inproceedings{bittner2014medleydb,
    Author = {Bittner, Rachel M and Salamon, Justin and Tierney, Mike and Mauch, Matthias and Cannam, Chris and Bello, Juan P},
    Booktitle = {International Society of Music Information Retrieval (ISMIR)},
    Month = {October},
    Title = {Medley{DB}: A Multitrack Dataset for Annotation-Intensive {MIR} Research},
    Year = {2014}
    To download this dataset, visit:
    and request access.

    Once downloaded, unzip the file
    and copy the result to:

    "Creative Commons Attribution Non-Commercial Share-Alike 4.0 (CC BY-NC-SA 4.0)."

[docs]class Track(core.Track): """medleydb_melody Track class Args: track_id (str): track id of the track Attributes: artist (str): artist audio_path (str): path to the audio file genre (str): genre is_excerpt (bool): True if the track is an excerpt is_instrumental (bool): True of the track does not contain vocals melody1_path (str): path to the melody1 annotation file melody2_path (str): path to the melody2 annotation file melody3_path (str): path to the melody3 annotation file n_sources (int): Number of instruments in the track title (str): title track_id (str): track id Cached Properties: melody1 (F0Data): the pitch of the single most predominant source (often the voice) melody2 (F0Data): the pitch of the predominant source for each point in time melody3 (MultiF0Data): the pitch of any melodic source. Allows for more than one f0 value at a time """ def __init__( self, track_id, data_home, dataset_name, index, metadata, ): super().__init__( track_id, data_home, dataset_name, index, metadata, ) self.melody1_path = self.get_path("melody1") self.melody2_path = self.get_path("melody2") self.melody3_path = self.get_path("melody3") self.audio_path = self.get_path("audio") @property def artist(self): return self._track_metadata.get("artist") @property def title(self): return self._track_metadata.get("title") @property def genre(self): return self._track_metadata.get("genre") @property def is_excerpt(self): return self._track_metadata.get("is_excerpt") @property def is_instrumental(self): return self._track_metadata.get("is_instrumental") @property def n_sources(self): return self._track_metadata.get("n_sources") @core.cached_property def melody1(self) -> Optional[annotations.F0Data]: return load_melody(self.melody1_path) @core.cached_property def melody2(self) -> Optional[annotations.F0Data]: return load_melody(self.melody2_path) @core.cached_property def melody3(self) -> Optional[annotations.MultiF0Data]: return load_melody3(self.melody3_path) @property def audio(self) -> Optional[Tuple[np.ndarray, float]]: """The track's audio Returns: * np.ndarray - audio signal * float - sample rate """ return load_audio(self.audio_path)
[docs] def to_jams(self): """Get the track's data in jams format Returns: jams.JAMS: the track's data in jams format """ # jams does not support multiF0, so we skip melody3 return jams_utils.jams_converter( audio_path=self.audio_path, f0_data=[(self.melody1, "melody1"), (self.melody2, "melody2")], metadata=self._track_metadata, )
[docs]@io.coerce_to_bytes_io def load_audio(fhandle: BinaryIO) -> Tuple[np.ndarray, float]: """Load a MedleyDB audio file. Args: fhandle (str or file-like): File-like object or path to audio file Returns: * np.ndarray - the mono audio signal * float - The sample rate of the audio file """ return librosa.load(fhandle, sr=None, mono=True)
[docs]@io.coerce_to_string_io def load_melody(fhandle: TextIO) -> annotations.F0Data: """Load a MedleyDB melody1 or melody2 annotation file Args: fhandle (str or file-like): File-like object or path to a melody annotation file Raises: IOError: if melody_path does not exist Returns: F0Data: melody data """ times = [] freqs = [] reader = csv.reader(fhandle, delimiter=",") for line in reader: times.append(float(line[0])) freqs.append(float(line[1])) times = np.array(times) # type: ignore freqs = np.array(freqs) # type: ignore confidence = (cast(np.ndarray, freqs) > 0).astype(float) return annotations.F0Data(times, freqs, confidence)
[docs]@io.coerce_to_string_io def load_melody3(fhandle: TextIO) -> annotations.MultiF0Data: """Load a MedleyDB melody3 annotation file Args: fhandle (str or file-like): File-like object or melody 3 melody annotation path Raises: IOError: if melody_path does not exist Returns: MultiF0Data: melody 3 annotation data """ times = [] freqs_list = [] conf_list = [] reader = csv.reader(fhandle, delimiter=",") for line in reader: times.append(float(line[0])) freqs_list.append([float(v) for v in line[1:]]) conf_list.append([float(float(v) > 0) for v in line[1:]]) times = np.array(times) # type: ignore melody_data = annotations.MultiF0Data(times, freqs_list, conf_list) return melody_data
[docs]@core.docstring_inherit(core.Dataset) class Dataset(core.Dataset): """ The medleydb_melody dataset """ def __init__(self, data_home=None): super().__init__( data_home, name="medleydb_melody", track_class=Track, bibtex=BIBTEX, download_info=DOWNLOAD_INFO, license_info=LICENSE_INFO, ) @core.cached_property def _metadata(self): metadata_path = os.path.join(self.data_home, "medleydb_melody_metadata.json") if not os.path.exists(metadata_path): raise FileNotFoundError("Metadata not found. Did you run .download()?") with open(metadata_path, "r") as fhandle: metadata = json.load(fhandle) return metadata
[docs] @core.copy_docs(load_audio) def load_audio(self, *args, **kwargs): return load_audio(*args, **kwargs)
[docs] @core.copy_docs(load_melody) def load_melody(self, *args, **kwargs): return load_melody(*args, **kwargs)
[docs] @core.copy_docs(load_melody3) def load_melody3(self, *args, **kwargs): return load_melody3(*args, **kwargs)