def audio_training(root_data_folder, dry=False, max_sessions=False):
from ibllib.io.extractors import training_audio as audio
audio_flags = Path(root_data_folder).rglob('audio_training.flag')
c = 0
for flag in audio_flags:
c += 1
if max_sessions and c > max_sessions:
return
_logger.info(flag)
if dry:
continue
session_path = flag.parent
try:
settings = raw_data_loaders.load_settings(session_path)
typ = extract_session.get_task_extractor_type(
settings.get('PYBPOD_PROTOCOL'))
except json.decoder.JSONDecodeError:
typ = 'unknown'
# this extractor is only for biased and training sessions
if typ not in ['biased', 'training', 'habituation']:
flag.unlink()
continue
audio.extract_sound(session_path, save=True, delete=True)
flag.unlink()
session_path.joinpath('register_me.flag').touch()
def test_qc_extract(self):
# extract audio
audio.extract_sound(self.ses_path, save=True)
D = alf.io.load_object(self.ses_path / 'alf', '_ibl_audioSpectrogram')
cues = alf.io.load_object(self.ses_path / 'alf',
'_ibl_audioOnsetGoCue.times_microphone')
self.assertEqual(cues['times_microphone'].size, 5)
self.assertEqual(D['power'].shape[0], D['times_microphone'].shape[0])
self.assertEqual(D['frequencies'].shape[1], D['power'].shape[1])
def test_qc_extract(self):
# extract audio
audio.extract_sound(self.ses_path, save=True)
D = alf.io.load_object(self.ses_path / 'raw_behavior_data',
'audioSpectrogram')
cues = alf.io.load_object(self.ses_path / 'raw_behavior_data',
'audioOnsetGoCue')
self.assertEqual(cues['times_mic'].size, 4)
self.assertEqual(D['power'].shape[0], D['times_mic'].shape[0])
self.assertEqual(D['frequencies'].shape[1], D['power'].shape[1])
def _run(self, overwrite=False):
return training_audio.extract_sound(self.session_path,
save=True,
delete=True)
from pathlib import Path
import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
import alf.io
from ibllib.io.extractors import training_audio as audio
main_path = '/mnt/s0/Data/Subjects'
# step 1 is to launch the extraction on audio
for wfile in Path(main_path).rglob('*.wav'):
print(wfile)
ses_path = wfile.parents[1]
audio.extract_sound(ses_path, save=True)
# step 2 plot the result - here for the last session only
D = alf.io.load_object(ses_path / 'alf', 'audioSpectrogram')
cues = alf.io.load_object(ses_path / 'alf',
'audioOnsetGoCue',
attribute='times',
timescale='microphone')
tlims = D['times_microphone'][[0, -1]].flatten()
flims = D['frequencies'][0, [0, -1]].flatten()
fig = plt.figure(figsize=[16, 7])
ax = plt.axes()
im = ax.imshow(20 * np.log10(D['power'].T),