def load_melody(f0_path):
"""Load cante100 f0 annotations
Args:
f0_path (str): path to audio file
Returns:
F0Data: predominant melody
"""
if not os.path.exists(f0_path):
raise IOError("f0_path {} does not exist".format(f0_path))
times = []
freqs = []
with open(f0_path, "r") as fhandle:
reader = csv.reader(fhandle, delimiter=",")
for line in reader:
times.append(float(line[0]))
freqs.append(float(line[1]))
times = np.array(times)
freqs = np.array(freqs)
confidence = (freqs > 0).astype(float)
return annotations.F0Data(times, freqs, confidence)
def load_melody(melody_path):
"""Load an Orchset melody annotation file
Args:
melody_path (str): path to melody annotation file
Raises:
IOError: if melody_path doesn't exist
Returns:
F0Data: melody annotation data
"""
if not os.path.exists(melody_path):
raise IOError("melody_path {} does not exist".format(melody_path))
times = []
freqs = []
confidence = []
with open(melody_path, "r") as fhandle:
reader = csv.reader(fhandle, delimiter="\t")
for line in reader:
times.append(float(line[0]))
freqs.append(float(line[1]))
confidence.append(0.0 if line[1] == "0" else 1.0)
melody_data = annotations.F0Data(np.array(times), np.array(freqs),
np.array(confidence))
return melody_data
def load_pitch(fhandle: TextIO) -> annotations.F0Data:
"""load a MedleyDB pitch annotation file
Args:
pitch_path (str): path to pitch annotation file
Raises:
IOError: if pitch_path doesn't exist
Returns:
F0Data: pitch annotation
"""
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)
freqs = np.array(freqs)
confidence = (cast(np.ndarray, freqs) > 0).astype(float)
pitch_data = annotations.F0Data(times, freqs, confidence)
return pitch_data
def load_pitch(pitch_path):
"""Load automatic extracted pitch or melody
Args:
pitch path (str): Local path where the pitch annotation is stored.
If `None`, returns None.
Returns:
F0Data: pitch annotation
"""
if pitch_path is None:
return None
if not os.path.exists(pitch_path):
raise IOError("pitch_path {} does not exist".format(pitch_path))
times = []
freqs = []
with open(pitch_path, "r") as fhandle:
reader = csv.reader(fhandle, delimiter="\t")
for line in reader:
times.append(float(line[0]))
freqs.append(float(line[1]))
if not times:
return None
times = np.array(times)
freqs = np.array(freqs)
confidence = (freqs > 0).astype(float)
return annotations.F0Data(times, freqs, confidence)
def load_pitch(pitch_path):
"""load a MedleyDB pitch annotation file
Args:
pitch_path (str): path to pitch annotation file
Raises:
IOError: if pitch_path doesn't exist
Returns:
F0Data: pitch annotation
"""
if not os.path.exists(pitch_path):
raise IOError("pitch_path {} does not exist".format(pitch_path))
times = []
freqs = []
with open(pitch_path, "r") as fhandle:
reader = csv.reader(fhandle, delimiter=",")
for line in reader:
times.append(float(line[0]))
freqs.append(float(line[1]))
times = np.array(times)
freqs = np.array(freqs)
confidence = (freqs > 0).astype(float)
pitch_data = annotations.F0Data(times, freqs, confidence)
return pitch_data
def load_melody(fhandle: TextIO) -> annotations.F0Data:
"""Load an Orchset melody annotation file
Args:
fhandle (str or file-like): File-like object or path to melody annotation file
Raises:
IOError: if melody_path doesn't exist
Returns:
F0Data: melody annotation data
"""
times = []
freqs = []
confidence = []
reader = csv.reader(fhandle, delimiter="\t")
for line in reader:
times.append(float(line[0]))
freqs.append(float(line[1]))
confidence.append(0.0 if line[1] == "0" else 1.0)
melody_data = annotations.F0Data(np.array(times), np.array(freqs),
np.array(confidence))
return melody_data
def load_melody(melody_path):
"""Load a MedleyDB melody1 or melody2 annotation file
Args:
melody_path (str): path to a melody annotation file
Raises:
IOError: if melody_path does not exist
Returns:
F0Data: melody data
"""
if not os.path.exists(melody_path):
raise IOError("melody_path {} does not exist".format(melody_path))
times = []
freqs = []
with open(melody_path, "r") as fhandle:
reader = csv.reader(fhandle, delimiter=",")
for line in reader:
times.append(float(line[0]))
freqs.append(float(line[1]))
times = np.array(times)
freqs = np.array(freqs)
confidence = (freqs > 0).astype(float)
melody_data = annotations.F0Data(times, freqs, confidence)
return melody_data
def test_f0_data():
times = np.array([1.0, 2.0, 3.0])
frequencies = np.array([100.0, 150.0, 120.0])
confidence = np.array([0.1, 0.4, 0.2])
f0_data = annotations.F0Data(times, frequencies, confidence)
assert np.allclose(f0_data.times, times)
assert np.allclose(f0_data.frequencies, frequencies)
assert np.allclose(f0_data.confidence, confidence)
def load_pitch(fhandle: TextIO) -> annotations.F0Data:
"""Load pitch
Args:
fhandle (str or file-like): path or file-like object pointing to a pitch annotation file
Returns:
F0Data: pitch annotation
"""
time_step = 128 / 44100 # hop-size / fs
reader = csv.reader(fhandle, delimiter=",")
freqs = np.array([float(line[0]) for line in reader])
times = np.array(np.arange(len(freqs)) * time_step)
confidence = np.array((freqs > 0.0).astype(float))
return annotations.F0Data(times, freqs, confidence)
def load_f0(fhandle: TextIO) -> annotations.F0Data:
"""Load an ikala f0 annotation
Args:
fhandle (str or file-like): File-like object or path to f0 annotation file
Raises:
IOError: If f0_path does not exist
Returns:
F0Data: the f0 annotation data
"""
lines = fhandle.readlines()
f0_midi = np.array([float(line) for line in lines])
f0_hz = librosa.midi_to_hz(f0_midi) * (f0_midi > 0)
confidence = (f0_hz > 0).astype(float)
times = (np.arange(len(f0_midi)) * TIME_STEP) + (TIME_STEP / 2.0)
f0_data = annotations.F0Data(times, f0_hz, confidence)
return f0_data
def load_f0(fhandle: TextIO) -> annotations.F0Data:
"""Load a Dagstuhl ChoirSet F0-trajectory.
Args:
fhandle (str or file-like): File-like object or path to F0 file
Returns:
F0Data Object - the F0-trajectory
"""
times = []
freqs = []
confs = []
reader = csv.reader(fhandle, delimiter=",")
for line in reader:
times.append(float(line[0]))
freqs.append(float(line[1]))
if len(line) == 3:
confs.append(float(line[2]))
else:
confs.append(float(1.0))
return annotations.F0Data(np.array(times), np.array(freqs), np.array(confs))
def load_melody(fhandle: TextIO) -> Optional[annotations.F0Data]:
"""Load cante100 f0 annotations
Args:
fhandle (str or file-like): path or file-like object pointing to melody annotation file
Returns:
F0Data: predominant melody
"""
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)
def load_pitch_contour(jams_path, string_num):
"""Load a guitarset pitch contour annotation for a given string
Args:
jams_path (str): Path of the jams annotation file
string_num (int), in range(6): Which string to load.
0 is the Low E string, 5 is the high e string.
Returns:
F0Data: Pitch contour data for the given string
"""
if not os.path.exists(jams_path):
raise IOError("jams_path {} does not exist".format(jams_path))
jam = jams.load(jams_path)
anno_arr = jam.search(namespace="pitch_contour")
anno = anno_arr.search(data_source=str(string_num))[0]
times, values = anno.to_event_values()
if len(times) == 0:
return None
frequencies = [v["frequency"] for v in values]
return annotations.F0Data(times, np.array(frequencies))
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)
def test_f0s():
f0_data_1 = [
(
annotations.F0Data(
np.array([0.016, 0.048]), np.array([0.0, 260.9]), np.array([0.0, 1.0])
),
None,
)
]
f0_data_2 = [
(
annotations.F0Data(
np.array([0.016, 0.048]), np.array([0.0, 260.9]), np.array([0.0, 1.0])
),
"f0s_1",
)
]
f0_data_3 = [
(
annotations.F0Data(
np.array([0.016, 0.048]), np.array([0.0, 260.9]), np.array([0.0, 1.0])
),
"f0s_1",
),
(
annotations.F0Data(
np.array([0.003, 0.012]), np.array([0.0, 230.5]), np.array([0.0, 1.0])
),
"f0s_2",
),
]
f0_data_4 = (
annotations.F0Data(
np.array([0.016, 0.048]), np.array([0.0, 260.9]), np.array([0.0, 1.0])
),
"f0s_1",
)
f0_data_5 = [
[
annotations.F0Data(
np.array([0.016, 0.048]), np.array([0.0, 260.9]), np.array([0.0, 1.0])
),
"f0s_1",
],
(
annotations.F0Data(
np.array([0.003, 0.012]), np.array([0.0, 230.5]), np.array([0.0, 1.0])
),
"f0s_2",
),
]
f0_data_6 = [(None, None)]
f0_data_7 = [
(
annotations.EventData(
np.array([[0.2, 0.3], [0.3, 0.4]]).T,
["event A", "event B"],
),
None,
)
]
jam_1 = jams_utils.jams_converter(f0_data=f0_data_1)
jam_2 = jams_utils.jams_converter(f0_data=f0_data_2)
jam_3 = jams_utils.jams_converter(f0_data=f0_data_3)
jam_6 = jams_utils.jams_converter(f0_data=f0_data_6)
time, duration, value, confidence = get_jam_data(jam_1, "pitch_contour", 0)
assert time == [0.016, 0.048]
assert duration == [0.0, 0.0]
assert value == [
{"frequency": 0.0, "index": 0, "voiced": False},
{"frequency": 260.9, "index": 0, "voiced": True},
]
assert confidence == [0.0, 1.0]
assert jam_2.annotations[0]["sandbox"]["name"] == "f0s_1"
time, duration, value, confidence = get_jam_data(jam_3, "pitch_contour", 0)
assert time == [0.016, 0.048]
assert duration == [0.0, 0.0]
assert value == [
{"frequency": 0.0, "index": 0, "voiced": False},
{"frequency": 260.9, "index": 0, "voiced": True},
]
assert confidence == [0.0, 1.0]
time, duration, value, confidence = get_jam_data(jam_3, "pitch_contour", 1)
assert time == [0.003, 0.012]
assert duration == [0.0, 0.0]
assert value == [
{"frequency": 0.0, "index": 0, "voiced": False},
{"frequency": 230.5, "index": 0, "voiced": True},
]
assert confidence == [0.0, 1.0]
time, duration, value, confidence = get_jam_data(jam_6, "pitch_contour", 0)
assert time == []
assert duration == []
assert value == []
assert confidence == []
assert type(jam_1) == jams.JAMS
with pytest.raises(TypeError):
jams_utils.jams_converter(f0_data=f0_data_4)
with pytest.raises(TypeError):
jams_utils.jams_converter(f0_data=f0_data_5)
with pytest.raises(TypeError):
jams_utils.jams_converter(f0_data=f0_data_7)
