def process_feature_song_list(dataset_name,
song_list,
harmonic=False,
num_harmonic=0):
fs = 44100
if harmonic:
freq_range = [1.0, fs / 2]
else:
freq_range = [27.5, 4487.0]
hdf_out = h5py.File(dataset_name + ".hdf", "w")
for idx, song in enumerate(song_list):
print("Extracting({}/{}): {}".format(idx + 1, len(song_list), song))
out = feature_extraction(song,
fc=freq_range[0],
tc=(1 / freq_range[1]),
Down_fs=fs)
cenf = out[5]
#z, spec, gcos, ceps, cenf = out[0:5]
piece = np.transpose(np.array(out[0:4]), axes=(2, 1, 0))
if harmonic:
# Harmonic spectrum
har = []
for i in range(num_harmonic + 1):
har.append(fetch_harmonic(out[1], cenf, i))
har_s = np.transpose(np.array(har), axes=(2, 1, 0))
# Harmonic GCoS
har = []
for i in range(num_harmonic + 1):
har.append(fetch_harmonic(out[2], cenf, i))
har_g = np.transpose(np.array(har), axes=(2, 1, 0))
# Harmonic cepstrum
har = []
for i in range(num_harmonic + 1):
har.append(fetch_harmonic(out[3], cenf, i, is_reverse=True))
har_c = np.transpose(np.array(har), axes=(2, 1, 0))
piece = np.dstack((har_s, har_g, har_c))
key = os.path.basename(song)
key = key.replace(".wav", "")
hdf_out.create_dataset(key,
data=piece,
compression="gzip",
compression_opts=5)
hdf_out.close()
def main(args):
# Pre-process features
assert (
os.path.isfile(args.input_audio)
), "The given path is not a file!. Please check your input again. Given input: {}".format(
audio.input_audio)
print("Processing features of input audio: {}".format(args.input_audio))
Z, tfrL0, tfrLF, tfrLQ, t, cenf, f = feature_extraction(args.input_audio)
# Post-process feature according to the configuration of model
feature_type, channels, out_class, timesteps = model_info(args.model_path)
if feature_type == "HCFP":
assert (len(channels) == (args.num_harmonics * 2 + 2))
spec = []
ceps = []
for i in range(args.num_harmonics):
spec.append(fetch_harmonic(tfrL0, cenf, i))
ceps.append(fetch_harmonic(tfrLQ, cenf, i))
spec = np.transpose(np.array(spec), axes=(2, 1, 0))
ceps = np.transpose(np.array(ceps), axes=(2, 1, 0))
feature = np.dstack((spec, ceps))
else:
assert (len(channels) <= 4)
feature = np.array([Z, tfrL0, tfrLF, tfrLQ])
feature = np.transpose(feature, axes=(2, 1, 0))
model = load_model(args.model_path)
print("Predicting...")
#pred = predict(feature[:,:,channels], model, timesteps, out_class, batch_size=4, overlap_ratio=2/4)
pred = predict_v1(feature[:, :, channels], model, timesteps, batch_size=4)
#p_out = h5py.File("pred.hdf", "w")
#p_out.create_dataset("0", data=pred)
#p_out.close()
midi = MultiPostProcess(pred,
mode="note",
onset_th=args.onset_th,
dura_th=0.5,
frm_th=3,
inst_th=1.1,
t_unit=0.02)
if args.to_midi is not None:
midi.write(args.to_midi)
print("Midi written as {}".format(args.to_midi))
def main():
parser = create_parser()
args = parser.parse_args()
# Pre-process features
assert (
os.path.isfile(args.input_audio)
), f"The given path is not a file!. Please check your input again. Given input: {args.input_audio}"
print("Processing features of input audio: {}".format(args.input_audio))
Z, tfrL0, tfrLF, tfrLQ, t, cenf, f = feature_extraction(args.input_audio)
# Load pre-trained model
minfo = ModelInfo()
model = minfo.load_model(args.model_path)
__import__('ipdb').set_trace()
minfo.onset_th = minfo.onset_th if args.onset_th is None else args.onset_th
print(minfo)
# Post-process feature according to the configuration of model
if minfo.feature_type == "HCFP":
assert (len(minfo.input_channels) == (HarmonicNum * 2 + 2))
spec = []
ceps = []
for i in range(HarmonicNum + 1):
spec.append(fetch_harmonic(tfrL0, cenf, i))
ceps.append(fetch_harmonic(tfrLQ, cenf, i))
spec = np.transpose(np.array(spec), axes=(2, 1, 0))
ceps = np.transpose(np.array(ceps), axes=(2, 1, 0))
feature = np.dstack((spec, ceps))
else:
assert (len(minfo.input_channels) <= 4)
feature = np.array([Z, tfrL0, tfrLF, tfrLQ])
feature = np.transpose(feature, axes=(2, 1, 0))
print("Predicting...")
pred = predict_v1(feature[:, :, minfo.input_channels],
model,
minfo.timesteps,
batch_size=4)
mode_mapping = {
"frame": "true_frame",
"frame_onset": "note",
"multi_instrument_frame": "true_frame",
"multi_instrument_note": "note"
}
midi = MultiPostProcess(pred,
mode=mode_mapping[minfo.label_type],
onset_th=minfo.onset_th,
dura_th=minfo.dura_th,
frm_th=minfo.frm_th,
inst_th=minfo.inst_th,
t_unit=0.02)
if args.to_midi is not None:
midi.write(args.to_midi)
print("Midi written as {}".format(args.to_midi))