"""Four-Way Tabla Stroke Transcription and Classification Loader
.. admonition:: Dataset Info
:class: dropdown
The Four-Way Tabla Dataset includes audio recordings of tabla solo with onset annotations for particular
strokes types. This dataset was published in 2021 in the context of ISMIR2021 (Online), and may be used for
tasks related to tabla analysis, including problems such as onset detection and stroke classification.
Total audio samples: We do have a total of 226 samples for training and 10 for testing. Each audio has
an approximate duration of 1 minute.
Audio specifications:
* Sampling frequency: 44.1 kHz
* Bit-depth: 16 bit
* Audio format: .wav
Dataset usage: This dataset may be used for the data-driven research of tabla stroke transcription and
identification. In this dataset, four important tabla characteristic strokes are considered.
Dataset structure: The dataset is split in two subsets, containing training and testing samples. Within each
subset, there is a folder containing the audios, and another folder containing the onset annotations. The onset
annotations are organized in a folder per each stroke type: b, d, rb, rt. Therefore, the paths to onsets would
look like:
.. code-block:: bash
train/onsets/<StrokeType>/<ID>.onsets
The dataset is made available by CompMusic under a Creative Commons
Attribution 3.0 Unported (CC BY 3.0) License.
"""
import csv
import librosa
import numpy as np
from typing import BinaryIO, Optional, Tuple
from deprecated.sphinx import deprecated
from mirdata import annotations, core, download_utils, io, jams_utils
BIBTEX = """@article{RohitMA2021,
author = {M.A, Rohit and Bhattacharjee, Amitrajit and Rao, Preeti},
journal = {Proc. of the 22nd Int. Society for Music Information Retrieval Conf., Online, 2021},
title = {{Four-way Classification of Tabla Strokes with Models Adapted from Automatic Drum Transcription}},
year = {2021}
}"""
INDEXES = {
"default": "1.0",
"test": "1.0",
"1.0": core.Index(filename="four_way_tabla_index.json"),
}
REMOTES = {
"remote_data": download_utils.RemoteFileMetadata(
filename="4way-tabla-ismir21-dataset.zip",
url="https://zenodo.org/record/7110248/files/4way-tabla-ismir21-dataset.zip?download=1",
checksum="fcddb565f260d4170877f70a7c33d69d", # the md5 checksum
),
}
LICENSE_INFO = "Creative Commons Attribution 4.0 International License."
[docs]class Track(core.Track):
"""Four-Way Tabla track class
Args:
track_id (str): track id of the track
data_home (str): Local path where the dataset is stored.
Attributes:
track_id (str): track id
audio_path (str): audio path
onsets_b_path (str): path to B onsets
onsets_d_path (str): path to D onsets
onsets_rb_path (str): path to RB onsets
onsets_rt_path (str): path to RT onsets
"""
def __init__(
self,
track_id,
data_home,
dataset_name,
index,
metadata,
):
super().__init__(
track_id,
data_home,
dataset_name,
index,
metadata,
)
self.audio_path = self.get_path("audio")
self.onsets_b_path = self.get_path("onsets_b")
self.onsets_d_path = self.get_path("onsets_d")
self.onsets_rb_path = self.get_path("onsets_rb")
self.onsets_rt_path = self.get_path("onsets_rt")
self.train = True if "train" in self.audio_path else False
@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)
@property
def onsets_b(self) -> Optional[annotations.BeatData]:
"""Onsets for stroke B
Returns:
* annotations.BeatData - onsets annotation
"""
return load_onsets(self.onsets_b_path)
@property
def onsets_d(self) -> Optional[annotations.BeatData]:
"""Onsets for stroke D
Returns:
* annotations.BeatData - onsets annotation
"""
return load_onsets(self.onsets_d_path)
@property
def onsets_rb(self) -> Optional[annotations.BeatData]:
"""Onsets for stroke RB
Returns:
* annotations.BeatData - onsets annotation
"""
return load_onsets(self.onsets_rb_path)
@property
def onsets_rt(self) -> Optional[annotations.BeatData]:
"""Onsets for stroke RT
Returns:
* annotations.BeatData - onsets annotation
"""
return load_onsets(self.onsets_rt_path)
[docs] def to_jams(self):
"""Get the track's data in jams format
Returns:
jams.JAMS: the track's data in jams format
"""
return jams_utils.jams_converter(
audio_path=self.audio_path,
beat_data=[
(self.onsets_b, "onsets_b"),
(self.onsets_d, "onsets_d"),
(self.onsets_rb, "onsets_rb"),
(self.onsets_rt, "onsets_rt"),
],
metadata={"train": self.train},
)
[docs]@io.coerce_to_bytes_io
def load_audio(fhandle: BinaryIO) -> Tuple[np.ndarray, float]:
"""Load a Mridangam Stroke Dataset 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=44100, mono=True)
[docs]@io.coerce_to_string_io
def load_onsets(fhandle):
"""Load stroke onsets
Args:
fhandle (str or file-like): Local path where the pitch annotation is stored.
Returns:
EventData: onset annotations
"""
onsets = []
reader = csv.reader(fhandle, delimiter="\n")
for line in reader:
onsets.append(float(line[0]))
if not onsets:
return None
onsets = np.array(onsets)
# All strokes are out of an annotated metric
beat_position = np.zeros(onsets.shape)
return annotations.BeatData(onsets, "s", beat_position, "global_index")
[docs]@core.docstring_inherit(core.Dataset)
class Dataset(core.Dataset):
"""
The Four-Way Tabla dataset
"""
def __init__(self, data_home=None, version="default"):
super().__init__(
data_home,
version,
name="four_way_tabla",
track_class=Track,
bibtex=BIBTEX,
indexes=INDEXES,
remotes=REMOTES,
license_info=LICENSE_INFO,
)