def _main(wav_file,input_json,output_json,mode): root_path = os.path.join(os.path.dirname(__file__)) joint_cnn_model_path = os.path.join(root_path, 'cnnModels', 'joint') # load keras joint cnn model model_joint = load_model(os.path.join(joint_cnn_model_path, 'jan_joint0.h5')) # load log mel feature scaler scaler_joint = pickle.load(open(os.path.join(joint_cnn_model_path, 'scaler_joint.pkl'), 'rb')) data_wav, fs_wav = librosa.load(wav_file,sr=44100) mfshs = MFSHS(data_wav) mfshs.frame() pitches = mfshs.pitches zeroAmploc = mfshs.zeroAmploc #frequency = np.array(pitchResult['frequency']) log_mel_old = get_log_mel_madmom(wav_file, fs=fs_wav, hopsize_t=hopsize_t, channel=1) log_mel = scaler_joint.transform(log_mel_old) log_mel = feature_reshape(log_mel, nlen=7) log_mel = np.expand_dims(log_mel, axis=1) obs_syllable, obs_phoneme = model_joint.predict(log_mel, batch_size=128, verbose=2) obs_syllable = np.squeeze(obs_syllable) obs_syllable = smooth_obs(obs_syllable) obs_syllable[0] = 1.0 obs_syllable[-1] = 0.0 #print sf_onset_frame score_note,pauseLoc = parse_musescore(input_json) resultOnset = findPeak(obs_syllable,pitches,score_note) Note_and_onset = pitch_Note(pitches,resultOnset['onset_frame'],score_note) score_note = np.array(score_note) result_loc_info = sw_alignment(score_note,Note_and_onset['notes']) #result_info = saveJson(filename_json,pitches,resultOnset['onset_frame'],score_note,pauseLoc,mode) post_proprocess(output_json,pitches,resultOnset['onset_frame'],zeroAmploc,score_note,pauseLoc,result_loc_info,mode)
def _main(wav_file, input_json, output_json):
'''
first detect pitches
'''
mfshs = MFSHS(wav_file)
mfshs.pitch_detector()
pitches = mfshs.pitches
zero_amp_frame = mfshs.zeroAmploc
score_note, pauseLoc = parse_musescore(input_json)
'''
second detect onset
'''
onset_frame = detector_onset(wav_file, pitches, score_note)
'''
sw alignment
'''
match_loc_info = sw_alignment(pitches, onset_frame, score_note)
'''
post process and save result
'''
onset_offset_pitches = trans_onset_and_offset(match_loc_info, onset_frame,
pitches)
evaluator = Evaluator(output_json, onset_offset_pitches, zero_amp_frame,
score_note, pauseLoc)
def main(wav_file, score_file):
'''
first detect pitches
'''
mfshs = MFSHS(wav_file)
mfshs.pitch_detector()
pitches = mfshs.pitches
zero_amp_frame = mfshs.zeroAmploc
energes = mfshs.energes
score_note, note_type, pauseLoc = parse_musescore(
score_file) ## parse musescore
'''
second detect onset
'''
onset_frame = detector_onset(wav_file, pitches, score_note)
'''
sw alignment
'''
match_loc_info = sw_alignment(pitches, onset_frame, score_note)
'''
post process and save result
'''
onset_offset_pitches = trans_onset_and_offset(match_loc_info, onset_frame,
pitches)
filename_json = os.path.splitext(wav_file)[0] + ".json"
evaluator = Evaluator(filename_json, onset_offset_pitches, zero_amp_frame,
score_note, pauseLoc, note_type)
save_files(wav_file, onset_frame, pitches, evaluator.det_note, score_note)
'''
def cal_system_acc(wav_files):
def cal_acc(det_note, score_note):
det_note = np.array(det_note)
score_note = np.array(score_note)
det_note = is_octive(det_note, score_note)
diff_note = abs(det_note - score_note)
acc_index = np.where(diff_note <= 0.8)[0]
accuracy = len(acc_index) / len(diff_note)
return accuracy, diff_note
sys_acc_res = dict()
for index in range(len(wav_files)):
json_path = '/home/data/lj/onset_detect/MUS/evaluate/'
wav_file = wav_files[index]
fid = os.path.basename(wav_file)[0:-4]
#score_note = np.array(score_notes[fid]).astype(int)
json_path = os.path.join(json_path, fid)
score_file = [
os.path.join(json_path, x) for x in os.listdir(json_path)
if x.endswith('json')
][0]
mfshs = MFSHS(wav_file)
mfshs.pitch_detector()
pitches = mfshs.pitches
zero_amp_frame = mfshs.zeroAmploc
score_note, note_types, pauseLoc = parse_musescore(
score_file) # parse musescore
predictor = predictor_onset()
onset_time = predictor.predict(wav_file)
onset_frame = predictor.onset_frame
onset_frame = post_cnn_onset(pitches, onset_frame)
match_loc_info = sw_alignment(pitches, onset_frame, score_note)
onset_offset_pitches = trans_onset_and_offset(match_loc_info,
onset_frame, pitches)
filename_json = os.path.splitext(wav_file)[0] + ".json"
evaluator = Evaluator(filename_json, onset_offset_pitches,
zero_amp_frame, score_note, pauseLoc, note_types)
accuracy, diff_note = cal_acc(evaluator.det_note, score_note)
temp = dict()
temp['acc'] = accuracy
temp['dnote'] = evaluator.det_note
temp['snote'] = score_note
temp['diff_note'] = diff_note
sys_acc_res[fid] = temp
return sys_acc_res
def main(wav_file, score_file):
# mfshs = MFSHS(wav_file)
# mfshs.pitch_detector()
# pitches = mfshs.pitches #è¿â€Ã¥â€ºÅ¾Ã©Å¸Â³Ã©Â«Ë?
# print(type(pitches),pitches)
# zero_amp_frame = mfshs.zeroAmploc #音高�索�
# print(type(zero_amp_frame),zero_amp_frame)
# pitches_filepath = "/home/ywm/MUSIC/new_solfege_pYIN/data/1011_pitch.txt"
# pitches = []
# with open(pitches_filepath,'r') as f:
# a = f.readlines()
# for i in a:
# pitches.append(float(i.split()[0]))
# pitches = np.array(pitches)
pitches = demo.pYIN(wav_file)
pitches = np.array(pitches) - 20
pitches = np.where((pitches < 0.0), 0, pitches)
#print(type(pitches),pitches)
zero_amp_frame = np.where(pitches == 0)[0]
score_note, note_types, pauseLoc = parse_musescore( #解æžÂjsonä¹Âè°±,è¿â€Ã¥â€ºÅ¾Ã¤Â¹Âè°±ä¸Âçš„note值和休æ¢符ä½ÂçÂ?
score_file) # parse musescore
predictor = predictor_onset()
onset_time = predictor.predict(wav_file)
#draw_array(predictor.onset_pred)
onset_frame = predictor.onset_frame
onset_frame = post_cnn_onset(pitches, onset_frame)
#print("onset_frame:",onset_frame)
match_loc_info = sw_alignment(pitches, onset_frame, score_note)
#print(2)
onset_offset_pitches = trans_onset_and_offset(match_loc_info, onset_frame,
pitches)
#print("onset_offset_pitches:",onset_offset_pitches)
filename_json = os.path.splitext(wav_file)[0] + ".json"
evaluator = Evaluator(filename_json, onset_offset_pitches, zero_amp_frame,
score_note, pauseLoc, note_types)
#print(4)
save_files(wav_file, onset_frame, pitches, evaluator.det_note, score_note,
onset_offset_pitches['onset_frame'])
#print(5)
return evaluator.score
def print_save_acc(wav_files, ground_notes, score_notes):
fwt = open('log/acc.txt', 'w')
for index in range(len(wav_files)):
json_path = '/home/data/lj/onset_detect/MUS/evaluate/'
wav_file = wav_files[index]
fid = os.path.basename(wav_file)[0:-4]
# score_note = np.array(score_notes[fid]).astype(int)
json_path = os.path.join(json_path, fid)
score_file = [
os.path.join(json_path, x) for x in os.listdir(json_path)
if x.endswith('json')
][0]
mfshs = MFSHS(wav_file)
mfshs.pitch_detector()
pitches = mfshs.pitches
zero_amp_frame = mfshs.zeroAmploc
score_note, note_types, pauseLoc = parse_musescore(
score_file) # parse musescore
predictor = predictor_onset()
onset_time = predictor.predict(wav_file)
onset_frame = predictor.onset_frame
onset_frame = post_cnn_onset(pitches, onset_frame)
match_loc_info = sw_alignment(pitches, onset_frame, score_note)
onset_offset_pitches = trans_onset_and_offset(match_loc_info,
onset_frame, pitches)
filename_json = os.path.splitext(wav_file)[0] + ".json"
evaluator = Evaluator(filename_json, onset_offset_pitches,
zero_amp_frame, score_note, pauseLoc, note_types)
print(wav_file)
sys_acc, sys_diff_note = cal_acc(evaluator.det_note, score_note)
gnote = ground_notes[fid]
snote = score_note
ground_acc, ground_diff_note = cal_acc(gnote, snote)
ground_sys_acc, ground_sys_diff_note = cal_acc(evaluator.det_note,
gnote)
strs = '{}\t{}\t{}'.format('ground', 'sys', 'ground_sys')
print(strs)
strs = '{:.3f}\t{:.3f}\t{:.3f}\n'.format(ground_acc, sys_acc,
ground_sys_acc)
print(strs)
file_name = os.path.join('log', fid + '.txt')
with open(file_name, 'w') as fw:
fw.write('gnote\t\tsnote\t\tdnote\n')
for index in xrange(len(gnote)):
fw.write('{:.3f}\t\t{:.3f}\t\t{:.3f}\t\t'.format(
gnote[index], snote[index], evaluator.det_note[index]))
fw.write('{:.3f}\t\t{:.3f}\t\t{:.3f}\n'.format(
ground_diff_note[index], sys_diff_note[index],
ground_sys_diff_note[index]))
fw.write('{:.3f}\t\t{:.3f}\t\t{:.3f}\n'.format(
ground_acc, sys_acc, ground_sys_acc))
fwt.write(wav_file + '\n')
fwt.write(strs)
fwt.write('\n')
fwt.flush()
fwt.close()
def _main(wav_file,score_file,est_file=None):
print(wav_file)
data_wav, fs_wav = librosa.load(wav_file,sr=44100)
#start_time = time.time()
start_time = time.time()
mfshs = MFSHS(data_wav)
mfshs.frame()
pitches = mfshs.pitches
energes = mfshs.energes
zeroAmploc = mfshs.zeroAmploc
#print('pitch detection time:',time.time()-start_time)
root_path = os.path.join(os.path.dirname(__file__))
joint_cnn_model_path = os.path.join(root_path, 'cnnModels', 'joint')
# load keras joint cnn model
model_joint = load_model(os.path.join(joint_cnn_model_path, 'jan_joint0.h5'))
# load log mel feature scaler
scaler_joint = pickle.load(open(os.path.join(joint_cnn_model_path, 'scaler_joint.pkl'), 'rb'))
log_mel_old = get_log_mel_madmom(wav_file, fs=fs_wav, hopsize_t=hopsize_t, channel=1)
log_mel = scaler_joint.transform(log_mel_old)
log_mel = feature_reshape(log_mel, nlen=7)
log_mel = np.expand_dims(log_mel, axis=1)
#start_time = time.time()
obs_syllable, obs_phoneme = model_joint.predict(log_mel, batch_size=128, verbose=2)
#print('cnn detection time: ',time.time()-start_time)
obs_syllable = np.squeeze(obs_syllable)
obs_syllable = smooth_obs(obs_syllable)
obs_syllable[0] = 1.0
obs_syllable[-1] = 0.0
#start_time = time.time()
score_note,pauseLoc = parse_musescore(score_file)
resultOnset = findPeak(obs_syllable,pitches,score_note,est_file)
filename_json = os.path.splitext(wav_file)[0]+".json"
#print('post-processing time :' ,time.time()-start_time)
Note_and_onset = pitch_Note(pitches,resultOnset['onset_frame'],score_note)
#draw_energe(energes,resultOnset['onset_frame'],zeroAmploc)
score_note = np.array(score_note)
result_loc_info = sw_alignment(score_note,Note_and_onset['notes'])
#result_info,paddingzero_frame = saveJson(filename_json,pitches,resultOnset['onset_frame'],score_note,pauseLoc,0)
result_info,det_Note = post_proprocess(filename_json,pitches,resultOnset['onset_frame'],zeroAmploc,score_note,pauseLoc,result_loc_info,0)
#print("total time:",time.time()-start_time)
filename_pitch = os.path.splitext(wav_file)[0]+"_pitch.txt"
mfshs.saveArray(filename_pitch,pitches)
filename_onset = os.path.splitext(wav_file)[0]+"_onset.txt"
mfshs.saveArray(filename_onset,resultOnset['onset_time'])
filename_score = os.path.splitext(wav_file)[0]+"_score.txt"
mfshs.saveArray(filename_score,score_note)
filename_detnote = os.path.splitext(wav_file)[0]+"_detnote.txt"
mfshs.saveArray(filename_detnote,np.round(np.array(det_Note),2))
return result_info['score']
47, 47, 49, 49, 47, 47, 44, 42, 45, 44, 42, 40
]
onset_frame = [
11, 34, 74, 93, 128, 159, 192, 226, 261, 303, 343, 366, 439, 640, 642, 763,
797, 844, 873, 906, 943, 977, 1016, 1206, 1354, 1356, 1400, 1474, 1510,
1542, 1577, 1617, 1768, 1801, 1836, 1877, 1911, 1947, 2052, 2062, 2101,
2138, 2177, 2218, 2239
]
def load(f0_file):
f0_array = []
with open(f0_file, 'r+') as f:
f0_list = f.readlines()
for f0 in f0_list:
try:
f0 = float(f0.strip())
except BaseException as e:
print(e)
f0 = (69 + 12 * math.log(f0 / 440) / math.log(2)) if f0 > 0 else 0
f0_array.append(f0)
f.close()
pitches = np.array(f0_array)
return pitches
if __name__ == "__main__":
f0_file = os.path.join(dirpath, "1011_f0.txt")
pitches = load(f0_file)
match_loc_info = sw_alignment(pitches, onset_frame, score_note)
print(match_loc_info)