def test_tonas_load_label():
expected = [
Label(0.233333, 0.366666, 53),
Label(0.4, 0.583333, 58),
Label(0.6, 0.75, 58),
Label(0.766666, 0.95, 58)
]
gt_file_path = "./tests/resource/gt_files/tonas_gt_file.notes.Corrected"
labels = dset.TonasStructure.load_label(gt_file_path)
_assert_all_equal(expected, labels)
def test_cmedia_load_label():
expected = [
Label(0.12345, 0.3333, 50),
Label(1.112, 1.5, 66),
Label(1.6666, 1.78, 70),
Label(1.6666, 1.8, 73),
Label(1.94333, 2.3, 65)
]
gt_file_path = "./tests/resource/gt_files/cmedia_gt_file.csv"
labels = dset.CMediaStructure.load_label(gt_file_path)
_assert_all_equal(expected, labels)
def test_maps_load_label():
expected = [
Label(0.5, 1.74117, 69),
Label(0.843864, 1.13805, 66),
Label(0.843864, 1.13805, 50),
Label(0.990956, 1.74117, 73),
Label(1.13805, 1.74117, 74)
]
gt_file_path = "./tests/resource/gt_files/maps_gt_file.txt"
labels = dset.MapsStructure.load_label(gt_file_path)
_assert_all_equal(expected, labels)
def test_maestro_load_label():
expected = [
Label(31.979545, 32.340909, 60, 0, 80),
Label(32.4, 32.5, 65, 0, 80),
Label(32.520455, 33.079545, 67, 0, 80),
Label(33.059091, 33.479545, 65, 0, 80),
Label(33.579545, 33.940909, 63, 0, 80)
]
gt_file_path = "./tests/resource/gt_files/maestro_gt_file.mid"
labels = dset.MaestroStructure.load_label(gt_file_path)
_assert_all_equal(expected, labels)
def test_medleydb_load_label():
t_unit = 256 / 44100
expected = [
Label(2.461315, 2.46712, 74.312369),
Label(2.46712, 2.472925, 74.283337),
Label(2.472925, 2.47873, 74.256145),
Label(2.47873, 2.484535, 74.235337),
Label(2.484535, 2.4903, 74.209788)
]
gt_file_path = "./tests/resource/gt_files/medleydb_gt_file.csv"
labels = dset.MedleyDBStructure.load_label(gt_file_path)
_assert_all_equal(expected, labels)
def load_label(cls, label_path):
with open(label_path, "r") as lin:
lines = lin.readlines()
notes = np.array([round(float(note)) for note in lines])
note_diff = notes[1:] - notes[:-1]
change_idx = np.where(note_diff != 0)[0] + 1
change_idx = np.insert(change_idx, 0,
0) # Padding a single zero to the beginning.
labels = []
for idx, chi in enumerate(change_idx[:-1]):
note = notes[chi]
if note == 0:
continue
start_t = 0.01 * chi + 0.02 # The first frame starts from 20ms.
end_t = 0.01 * change_idx[idx + 1] + 0.02 # noqa: E226
if end_t - start_t < 0.05:
# Minimum duration should over 50ms.
continue
labels.append(
Label(start_time=float(start_t),
end_time=float(end_t),
note=note))
return labels
def load_label(cls, label_path):
labels = []
sample_rate = 44100
with open(label_path, "r") as label_file:
reader = csv.DictReader(label_file, delimiter=",")
for row in reader:
onset = float(row["start_time"]) / sample_rate
offset = float(row["end_time"]) / sample_rate
inst = int(row["instrument"]) - 1
note = int(row["note"])
# The statement used in the paper is 'measure', which is kind of ambiguous.
start_beat = float(row["start_beat"])
# It's actually beat length of 'end_beat' column, thus adding start beat position here to
# make it a 'real end_beat'.
end_beat = float(row["end_beat"]) + start_beat
note_value = row["note_value"]
label = Label(start_time=onset,
end_time=offset,
note=note,
instrument=inst,
start_beat=start_beat,
end_beat=end_beat,
note_value=note_value)
labels.append(label)
return labels
def load_label(cls, label_path):
labels = []
with open(label_path, "r") as label_file:
reader = csv.DictReader(label_file, delimiter=",")
for row in reader:
labels.append(
Label(start_time=float(row["onset"]),
end_time=float(row["offset"]),
note=int(row["note"])))
return labels
def load_label(cls, label_path):
lines = open(label_path, "r").readlines()[
1:] # Discard the first line which contains column names
labels = []
for line in lines:
if line.strip() == "":
continue
values = line.split("\t")
onset, offset, note = float(values[0]), float(values[1]), int(
values[2].strip())
labels.append(Label(start_time=onset, end_time=offset, note=note))
return labels
def load_label(cls, label_path):
with open(label_path, "r") as lin:
lines = lin.readlines()
labels = []
for line in lines[1:]:
onset, dura, note, _ = line.split(", ")
labels.append(
Label(start_time=float(onset),
end_time=float(onset) + float(dura),
note=round(float(note))))
return labels
def extract_feature_from_midi(midi_path, t_unit=0.01):
"""Extract feature for beat module from MIDI file.
See Also
--------
omnizart.beat.features.extract_feature:
The main feature extraction function of beat module.
"""
midi = pretty_midi.PrettyMIDI(midi_path)
labels = []
for inst in midi.instruments:
for note in inst.notes:
labels.append(Label(start_time=note.start, end_time=note.end, note=note.pitch))
return extract_feature(labels, t_unit=t_unit)
def load_label(cls, label_path):
with open(label_path, "r") as lin:
lines = lin.readlines()
labels = []
for idx, line in enumerate(lines):
note = float(line)
if note < 0.1:
# No pitch
continue
start_t = 0.01 * idx + 0.02 # The first frame starts from 20ms.
end_t = start_t + 0.01
labels.append(Label(start_time=start_t, end_time=end_t, note=note))
return labels
def load_label(cls, label_path):
midi = pretty_midi.PrettyMIDI(label_path)
labels = []
for inst in midi.instruments:
if inst.id_drum:
continue
for note in inst.notes:
label = Label(start_time=note.start,
end_time=note.end,
note=note.pitch,
velocity=note.velocity,
instrument=inst.program)
if label.note == -1:
continue
labels.append(label)
return labels
def load_label(cls, label_path):
with open(label_path, "r") as fin:
lines = fin.readlines()
labels = []
t_unit = 256 / 44100 # ~= 0.0058 secs
for line in lines:
elems = line.strip().split(",")
sec, hz = float(elems[0]), float(elems[1]) # pylint: disable=invalid-name
if hz < 1e-10:
continue
note = float(
hz_to_midi(hz)) # Convert return type of np.float64 to float
end_t = sec + t_unit
labels.append(Label(start_time=sec, end_time=end_t, note=note))
return labels
def test_mir1k_load_label():
expected = [
Label(0.04, 0.06, 57.9108),
Label(0.06, 0.08, 57.4161),
Label(0.08, 0.1, 57.174),
Label(0.1, 0.12, 59.7627),
Label(0.12, 0.14, 60.0442),
Label(0.14, 0.16, 60.3304)
]
gt_file_path = "./tests/resource/gt_files/mir1k_gt_file.pv"
labels = dset.MIR1KStructure.load_label(gt_file_path)
_assert_all_equal(expected, labels)
def load_label(cls, label_path):
"""Load and parse labels for the given label file path.
Parses different format of label information to shared intermediate format,
encapslated with :class:`Label` instances. The default is parsing MIDI
file format.
"""
midi = pretty_midi.PrettyMIDI(label_path)
labels = []
for inst in midi.instruments:
if inst.is_drum:
continue
for note in inst.notes:
label = Label(start_time=note.start,
end_time=note.end,
note=note.pitch,
velocity=note.velocity,
instrument=inst.program)
if label.note == -1:
continue
labels.append(label)
return labels
def load_label(cls, label_path):
labels = []
sample_rate = 44100
with open(label_path, "r") as label_file:
reader = csv.DictReader(label_file, delimiter=",")
for row in reader:
onset = float(row["start_time"]) / sample_rate
offset = float(row["end_time"]) / sample_rate
inst = int(row["instrument"]) - 1
note = int(row["note"])
start_beat = float(row["start_beat"])
end_beat = float(row["end_beat"])
note_value = row["note_value"]
label = Label(start_time=onset,
end_time=offset,
note=note,
instrument=inst,
start_beat=start_beat,
end_beat=end_beat,
note_value=note_value)
labels.append(label)
return labels
def test_musicnet_load_label():
expected = [
Label(0.231428,
0.962857,
61,
instrument=41,
start_beat=0,
end_beat=1.489583,
note_value="Dotted Half"),
Label(0.231428,
0.568117,
65,
instrument=6,
start_beat=0,
end_beat=0.489583,
note_value="Quarter"),
Label(0.231428,
0.510068,
46,
instrument=0,
start_beat=0,
end_beat=0.333333,
note_value="Dotted Eighth"),
Label(0.579727,
0.660997,
63,
instrument=40,
start_beat=0.5,
end_beat=0.739583,
note_value="Eighth"),
Label(0.579727,
0.695827,
58,
instrument=70,
start_beat=0.5,
end_beat=0.833333,
note_value="Dotted Eighth"),
Label(0.672607,
0.777097,
65,
instrument=71,
start_beat=0.75,
end_beat=0.989583,
note_value="Eighth"),
Label(0.777097,
1.473696,
58,
instrument=73,
start_beat=1.0,
end_beat=2.989583,
note_value="Whole"),
Label(0.777097,
0.904807,
66,
instrument=68,
start_beat=1.0,
end_beat=1.333333,
note_value="Dotted Eighth"),
Label(0.962857,
1.055714,
63,
instrument=0,
start_beat=1.5,
end_beat=1.739583,
note_value="Eighth"),
Label(0.962857,
1.090566,
39,
instrument=43,
start_beat=1.5,
end_beat=1.833333,
note_value="Dotted Eighth")
]
gt_file_path = "./tests/resource/gt_files/musicnet_gt_file.csv"
labels = dset.MusicNetStructure.load_label(gt_file_path)
_assert_all_equal(expected, labels)
