Skip to content

Commit

Permalink
Small refactoring and removing testARealCase() unittest
Browse files Browse the repository at this point in the history
  • Loading branch information
@xaviliz
xaviliz committed Jul 19, 2024
1 parent 382ad4f commit 044aa9a
Showing 1 changed file with 70 additions and 160 deletions.
230 changes: 70 additions & 160 deletions test/src/unittests/tonal/test_pitch2midi.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,7 @@

from essentia_test import *
from numpy import sin, pi, mean, random, array, float32, square
from pathlib import Path


class TestPitch2Midi(TestCase):
Expand Down Expand Up @@ -124,164 +125,37 @@ def runTest(
self.assertEqual(message_types[expected_idx], expected_value)
# TODO: think in an strategy similar to real cases that assess the number of notes, messages and timestamp tolerance

def testARealCase(self):

frameSize = 8192
sample_rate = 44100
hopSize = 128
loudness_threshold = -50
pitch_confidence_threshold = 0.25

filename = join(testdata.audio_dir, "recorded", "sax-test_min.wav")
if sys.platform == "darwin":
import soundfile as sf

audio, _ = sf.read(filename, dtype="float32")
audio = audio[:, 0]
else:
audio = MonoLoader(filename=filename, sampleRate=44100)()
frames = FrameGenerator(audio, frameSize=frameSize, hopSize=hopSize)

n_frames = (audio.shape[0] - (frameSize - hopSize)) / hopSize

# print(f"n_frames: {n_frames}")
# pitchDetect = Audio2Pitch(
# frameSize=frameSize,
# sampleRate=sample_rate,
# pitchConfidenceThreshold=0.15,
# loudnessThreshold=loudness_threshold,
# )

win = Windowing(normalized=False, zeroPhase=False)
spec = Spectrum()
pitchDetect = PitchYinFFT(
frameSize=frameSize,
sampleRate=sample_rate,
weighting="custom",
tolerance=1.0,
)
rmsDetect = RMS()

n_outputs = len(pitchDetect.outputNames())
pitch, confidence = ([] for _ in range(n_outputs))
loudness = []
voiced = []
n = 0
for frame in frames:
# print(n, end="\r")
# print(frame.shape)
# out = pitchDetect(frame)
_spec = spec(win(frame))
f, conf = pitchDetect(_spec)
pitch += [f]
confidence += [conf]
loudness += [rmsDetect(frame)]
if (
amp2db(loudness[n]) > loudness_threshold
and confidence[n] > pitch_confidence_threshold
):
voiced += [1]
else:
voiced += [0]
n += 1

expected_value = ["note_off"]
expected_idx = 2869
# TODO: create an annotation file for assessing the note toggle events
# TODO: apply audio2pitch module and assess with note metrics
self.runTest(sample_rate, hopSize, pitch, voiced, expected_value, expected_idx)

def testARealCaseWithEMajorScale(self):
frame_size = 8192
sample_rate = 48000
hop_size = 64
loudness_threshold = -40
pitch_confidence_threshold = 0.25

filename = join(
testdata.audio_dir, "recorded", "359500__mtg__sax-tenor-e-major.wav"
)
# TODO: save the below code in a function and reuse it for testARealCases

if sys.platform == "darwin":
import soundfile as sf

audio, _ = sf.read(filename, dtype="float32")
if audio.ndim > 1:
audio = audio[:, 0]
else:
audio = MonoLoader(filename=filename, sampleRate=sample_rate)()
frames = FrameGenerator(audio, frameSize=frame_size, hopSize=hop_size)

n_frames = (audio.shape[0] - (frame_size - hop_size)) / hop_size
step_time = hop_size / sample_rate

# initialize audio2pitch & pitch2midi instances
pitchDetect = Audio2Pitch(
frameSize=frame_size,
sampleRate=sample_rate,
pitchConfidenceThreshold=pitch_confidence_threshold,
loudnessThreshold=loudness_threshold,
)

p2m = Pitch2Midi(
sampleRate=sample_rate,
hopSize=hop_size,
midiBufferDuration=0.05,
)

# define estimate bin and some counters
estimated = []
n = 0
time_stamp = 0
n_notes = 0

# simulates real-time process
for frame in frames:
_pitch, _confidence, _loudness, _voiced = pitchDetect(frame)
message, midi_note, time_compensation = p2m(_pitch, _voiced)
time_stamp += step_time
if _voiced:
if message:
estimated.append(
[
n_notes,
time_stamp - time_compensation[1],
time_stamp - time_compensation[0],
int(midi_note[1]),
]
)
# print(
# f"[{n_notes}][{n}]:{(time_stamp-time_compensation[1]):.3f}, {midi2note(int(midi_note[1]))}({int(midi_note[1])})~{_pitch:.2f}Hz" # , {time_compensation}, {midi_note}, {message}
# )
if "note_on" in message:
n_notes += 1
n += 1

## convert note toggle messages to note features
expected_notes = numpy.load(
join(filedir(), "pitch2midi/359500__mtg__sax-tenor-e-major.npy")
)

# estimate the number of notes for expected and detected
n_detected_notes = len(estimated)
n_expected_notes = len(expected_notes)

# estimate the onset error for each note and estimate the mean
onset_mse = mean(
[square(note[1] - estimated[int(note[0])][1]) for note in expected_notes]
)

# estimate the midi note error for each note and estimate the mean
midi_note_mse = mean(
[square(note[-1] - estimated[int(note[0])][-1]) for note in expected_notes]
min_frequency = 103.83
max_frequency = 659.26
n_notes_tolerance = 0
onset_tolerance = 0.01
midi_note_tolerance = 0

stem = "359500__mtg__sax-tenor-e-major"
audio_path = Path("recorded") / f"{stem}.wav"
reference_path = Path("pitch2midi") / f"{stem}.npy"

self.runARealCase(
audio_path=audio_path,
reference_path=reference_path,
sample_rate=sample_rate,
frame_size=frame_size,
hop_size=hop_size,
pitch_confidence_threshold=pitch_confidence_threshold,
loudness_threshold=loudness_threshold,
max_frequency=max_frequency,
min_frequency=min_frequency,
n_notes_tolerance=n_notes_tolerance,
onset_tolerance=onset_tolerance,
midi_note_tolerance=midi_note_tolerance,
)

# assert outputs
self.assertAlmostEqual(onset_mse, 0, 0.01)
self.assertAlmostEqual(n_detected_notes, n_expected_notes, 0)
self.assertAlmostEqual(midi_note_mse, midi_note_mse, 0)

def testARealCaseWithDMinorScale(self):
frame_size = 8192
sample_rate = 48000
Expand All @@ -290,11 +164,45 @@ def testARealCaseWithDMinorScale(self):
pitch_confidence_threshold = 0.25
min_frequency = 103.83
max_frequency = 659.26

filename = join(
testdata.audio_dir, "recorded", "359628__mtg__sax-tenor-d-minor.wav"
n_notes_tolerance = 0
onset_tolerance = 0.01
midi_note_tolerance = 0

stem = "359628__mtg__sax-tenor-d-minor"
audio_path = Path("recorded") / f"{stem}.wav"
reference_path = Path("pitch2midi") / f"{stem}.npy"

self.runARealCase(
audio_path=audio_path,
reference_path=reference_path,
sample_rate=sample_rate,
frame_size=frame_size,
hop_size=hop_size,
pitch_confidence_threshold=pitch_confidence_threshold,
loudness_threshold=loudness_threshold,
max_frequency=max_frequency,
min_frequency=min_frequency,
n_notes_tolerance=n_notes_tolerance,
onset_tolerance=onset_tolerance,
midi_note_tolerance=midi_note_tolerance,
)

def runARealCase(
self,
audio_path: str,
reference_path: str,
sample_rate: int,
frame_size: int,
hop_size: int,
pitch_confidence_threshold: float,
loudness_threshold: float,
max_frequency: float,
min_frequency: float,
n_notes_tolerance: int = 0,
onset_tolerance: float = 0.01,
midi_note_tolerance: int = 0,
):
filename = join(testdata.audio_dir, audio_path)
if sys.platform == "darwin":
import soundfile as sf

Expand Down Expand Up @@ -353,13 +261,13 @@ def testARealCaseWithDMinorScale(self):
n_notes += 1
n += 1

## convert note toggle messages to note features
expected_notes = numpy.load(
join(filedir(), f"pitch2midi/359628__mtg__sax-tenor-d-minor.npy")
)
# read the expected notes file manually annotated
expected_notes = numpy.load(join(filedir(), reference_path))
# print("Expected notes:")
# print(expected_notes)

## convert note toggle messages to note features

# estimate the number of notes for expected and detected
n_detected_notes = len(estimated)
n_expected_notes = len(expected_notes)
Expand All @@ -375,11 +283,13 @@ def testARealCaseWithDMinorScale(self):
)

# assert outputs
onset_tolerance = 0.01
self.assertAlmostEqual(n_detected_notes, n_expected_notes, 0)

self.assertAlmostEqual(n_detected_notes, n_expected_notes, n_notes_tolerance)
self.assertAlmostEqual(onset_mse, 0, onset_tolerance)
self.assertAlmostEqual(midi_note_mse, midi_note_mse, 0)
self.assertAlmostEqual(midi_note_mse, midi_note_mse, midi_note_tolerance)


# TODO: create a new unittest for separated notes to assess offset in a REALCASE!

suite = allTests(TestPitch2Midi)

Expand Down

0 comments on commit 044aa9a

Please sign in to comment.