def generate(mod, run, language, textgrid, overwrite=False):
name = os.path.basename(os.path.splitext(run)[0])
run_name = name.split('_')[-1] # extract the name of the run
save_all = None
# Defining paths
model_name = 'glove_model'
check_folder(
os.path.join(Paths().path2derivatives, 'fMRI/raw-features', language,
model_name))
path = os.path.join(
Paths().path2derivatives, 'fMRI/raw-features', language, model_name,
'raw-features_{}_{}_{}.csv'.format(language, model_name, run_name))
#### generating raw-features ####
if (os.path.exists(path)) & (not overwrite):
raw_features = pd.read_csv(path)
else:
raw_features = mod.generate(run, language, textgrid)
save_all = path
#### Retrieving data of interest ####
# columns2retrieve = [function.__name__ for function in mod.functions]
columns2retrieve = [
'embedding-{}'.format(index)
for index in range(mod.param['embedding-size'])
]
return raw_features[:textgrid.offsets.count()], columns2retrieve, save_all
def generate(mod, run, language, textgrid, overwrite=False):
from .WORDRATE import model
from .WORDRATE.utils import wordrate, function_words, content_words
name = os.path.basename(os.path.splitext(run)[0])
run_name = name.split('_')[-1] # extract the name of the run
save_all = None
mod = model.Wordrate([content_words, function_words, wordrate],
language) # all functions
model_name = 'wordrate_model'
check_folder(
os.path.join(Paths().path2derivatives, 'fMRI/raw-features', language,
model_name))
path = os.path.join(
Paths().path2derivatives, 'fMRI/raw-features', language, model_name,
'raw-features_{}_{}_{}.csv'.format(language, model_name, run_name))
#### parameters studied ####
parameters = sorted([wordrate])
#### generating raw-features ####
if (os.path.exists(path)) & (not overwrite):
raw_features = pd.read_csv(path)
else:
raw_features = mod.generate(run, language, textgrid)
save_all = path
#### Retrieving data of interest ####
columns2retrieve = [
function.__name__
for function in model.Wordrate(parameters, language).functions
]
return raw_features[:textgrid.offsets.count()], columns2retrieve, save_all
def generate(mod, run, language, textgrid, overwrite=False):
name = os.path.basename(os.path.splitext(run)[0])
run_name = name.split('_')[-1] # extract the name of the run
save_all = None
mod = model.EnergySpectrum([rms], language) # all functions for now
model_name = 'rms_model'
check_folder(
os.path.join(Paths().path2derivatives, 'fMRI/raw-features', language,
model_name))
path = os.path.join(
Paths().path2derivatives, 'fMRI/raw-features', language, model_name,
'raw-features_{}_{}_{}.csv'.format(language, model_name, run_name))
#### parameters studied ####
parameters = sorted([rms])
#### generating raw-features ####
if (os.path.exists(path)) & (not overwrite):
raw_features = pd.read_csv(path)
else:
raw_features = mod.generate(run,
language,
textgrid,
slice_period=10e-3)
save_all = path
#### Retrieving data of interest ####
columns2retrieve = [
function.__name__
for function in model.EnergySpectrum(parameters, language).functions
]
return raw_features[:textgrid.offsets.count()], columns2retrieve, save_all
def generate(mod, run, language, textgrid, overwrite=False):
name = os.path.basename(os.path.splitext(run)[0])
run_name = name.split('_')[-1] # extract the name of the run
save_all = None
model_name = 'bottomup_model'
check_folder(os.path.join(Paths().path2derivatives, 'fMRI/raw-features', language, model_name))
path = os.path.join(Paths().path2derivatives, 'fMRI/raw-features', language, model_name, 'raw-features_{}_{}_{}.csv'.format(language, model_name, run_name))
#### generating raw-features ####
if (os.path.exists(path)) & (not overwrite):
raw_features = pd.read_csv(path)
else:
raw_features = mod.generate(run, language, textgrid)
save_all = path
#### Retrieving data of interest ####
columns2retrieve = ['bottomup']
textgrid = pd.read_csv(os.path.join(paths.path2data, 'text', language, 'BOTTOMUP', 'onsets-offsets', '{}_{}_{}_onsets-offsets_{}'.format('text', language, 'BOTTOMUP', run_name)+'.csv')) # df with onsets-offsets-word
return raw_features[:textgrid.offsets.count()], columns2retrieve, save_all
def generate(mod, run, language, textgrid, overwrite=False):
name = os.path.basename(os.path.splitext(run)[0])
run_name = name.split('_')[-1] # extract the name of the run
save_all = None
model_name = 'mfcc_model'
check_folder(os.path.join(Paths().path2derivatives, 'fMRI/raw-features', language, model_name))
path = os.path.join(Paths().path2derivatives, 'fMRI/raw-features', language, model_name, 'raw-features_{}_{}_{}.csv'.format(language, model_name, run_name))
#### generating raw-features ####
if (os.path.exists(path)) & (not overwrite):
raw_features = pd.read_csv(path)
else:
raw_features = mod.generate(run, language)
save_all = path
#### Retrieving data of interest ####
columns2retrieve = ["mfcc #{}".format((i)//3) if i%3==0 else ("mfcc' #{}".format((i)//3) if i%3==1 else "mfcc'' #{}".format((i)//3)) for i in range(mod.num_cepstral*3)]
textgrid = pd.read_csv(os.path.join(paths.path2data, 'wave', language, 'MFCC', 'onsets-offsets', '{}_{}_{}_onsets-offsets_{}'.format('wave', language, 'MFCC', run_name)+'.csv')) # df with onsets-offsets-word
return raw_features[:textgrid.offsets.count()], columns2retrieve, save_all
def generate(mod, run, language, textgrid, overwrite=False):
from .OTHER import model
from .OTHER.utils import sentence_onset
name = os.path.basename(os.path.splitext(run)[0])
run_name = name.split('_')[-1] # extract the name of the run
save_all = None
mod = model.Other([sentence_onset], language) # all functions
model_name = 'other_sentence_onset'
check_folder(
os.path.join(Paths().path2derivatives, 'fMRI/raw-features', language,
model_name))
path = os.path.join(
Paths().path2derivatives, 'fMRI/raw-features', language, model_name,
'raw-features_{}_{}_{}.csv'.format(language, model_name, run_name))
#### generating raw-features ####
if (os.path.exists(path)) & (not overwrite):
raw_features = pd.read_csv(path)
else:
raw_features = mod.generate(run, language, textgrid)
save_all = path
#### Retrieving data of interest ####
columns2retrieve = [function.__name__ for function in mod.functions]
return raw_features[:textgrid.offsets.count()], columns2retrieve, save_all
#plt.xlim(0,0.2)
#plt.ylim(0,0.2)
plt.plot([
max([np.min(x), np.min(y)]),
min([np.max(x), np.max(y)])
], [
max([np.min(x), np.min(y)]),
min([np.max(x), np.max(y)])
],
c='blue')
plt.axhline(y=0., color='blue', linestyle='-')
plt.legend()
save_folder = os.path.join(paths.path2derivatives, source,
'analysis', language,
'scatter_plots', analysis_name)
check_folder(save_folder)
plt.savefig(
os.path.join(
save_folder, analysis['title'] + ' - ' +
labels[index_mask + 1] + ' - ' + subject + '.png'))
plt.close()
i += 1
##########################################################################
############################## Check models ##############################
##########################################################################
if 'check_model' in args.analysis[0]:
# retrieve default atlas (= set of ROI)
atlas = datasets.fetch_atlas_harvard_oxford(params.atlas)
labels = atlas['labels']
# $ cd $LePetitPrince/code/
# $ python create_data_architecture.py
################################################################################
import os
from utilities.settings import Subjects, Paths, Params
from utilities.utils import check_folder
subjects = Subjects()
paths = Paths()
params = Params()
languages = params.languages
##################### ROOT #####################
os.chdir(paths.path2root)
check_folder('data')
check_folder('derivatives')
check_folder('paradigm')
check_folder('oldstuff')
##################### DATA #####################
os.chdir(paths.path2data)
## fMRI ##
check_folder('fMRI')
for language in languages:
check_folder('fMRI/{}'.format(language))
for subject in subjects.get_all(language):
check_folder('fMRI/{}/{}'.format(language, subject))
check_folder('fMRI/{}/{}/anat'.format(language, subject))
check_folder('fMRI/{}/{}/func'.format(language, subject))
default=False,
action='store_true',
help="Precise if we want to run code in parallel")
args = parser.parse_args()
input_data_type = 'raw-features'
output_data_type = 'features'
extension = '.csv'
source = 'fMRI'
model = args.model
output_parent_folder = get_output_parent_folder(source, output_data_type,
args.language, model)
check_folder(
output_parent_folder
) # check if the output_parent_folder exists and create it if not
raw_features = get_data(args.language,
input_data_type,
model=model,
source='fMRI')
if not args.parallel:
for i, run in enumerate(raw_features):
compute_features(run, output_parent_folder, output_data_type,
args.language, model, extension, args.tr,
args.overwrite)
else:
Parallel(n_jobs=-2)(delayed(compute_features) \
(run, output_parent_folder, output_data_type, args.language, model, extension, args.tr, args.overwrite) for run in raw_features)
