def verify_working_directory(directory):
train_files = prsg.ordering_files(directory)
if len(train_files) == 0:
print("ERROR : found no file nor directory")
print("This may be the wrong folder")
return 0
else:
for filename in train_files:
path = directory + filename
# print(path)
if (os.path.isdir(path)):
if (filename == CSV_FOLDER_NAME[:-1]
or filename == SPLITTED_FOLDER_NAME[:-1]):
continue
else:
print("WARNING : found an non initial folder")
print("This could be the wrong folder, be carefull !")
# return 0
continue
elif (os.path.isfile(path)):
if (filename.endswith('.txt') or filename.endswith('wav')):
# print(filename+" looks like an expected file")
continue
else:
print('ERROR : found a non \'txt\' nor \'wav\' file')
print("This should be the wrong folder")
return 0
else:
print("ERROR : found no file nor directory")
print("This is the wrong folder")
return 0
return 1
def split_record_in_cycle(dir,file_csv,output_dir) :
pbar = ProgressBar()
lines = prsg.nb_lines(file_csv)
with open(file_csv, newline='') as csvfile:
data = list(csv.reader(csvfile))
# print(data)
input_dir = prsg.ordering_files(dir)
# print(input_dir)
i=0
for filename in pbar(input_dir):
if(filename.endswith('.wav')):
cpt=1
save_file_name = filename[:-4]
print()
filename = data[i][0]
while data[i][0] == save_file_name:
print("Processed record = "+data[i][0]+" nb cycle = "+str(cpt))
myaudio = pydub.AudioSegment.from_wav(dir+data[i][0]+".wav")
chunk_data = myaudio[int(float(data[i][1])*1000):int(float(data[i][2])*1000)]
saved_file = (output_dir+save_file_name+"_"+"{0:0=2d}".format(cpt)+".wav")
# print("saved cycle name = "+saved_file)
chunk_data.export(saved_file, format="wav")
i+=1
cpt+=1
if i == lines:
break
return i
input_train_dir = arguments[1] + TRAIN_FOLDER
input_test_dir = arguments[1] + TEST_FOLDER
out_train_csv = open(arguments[1] + 'ft_train.csv', 'w')
out_test_csv = open(arguments[1] + 'ft_test.csv', 'w')
#########
# TRAIN #
#########
### STEP 3 : Compute features representing audio
################################################
# features_train_dir,status = prsg.verify_folder(input_train_dir,FEATURES_FOLDER_NAME) # Verify if folder already exists and if it is empty or not
list_features = prsg.ordering_files(input_train_dir)
# if(status == 0):
# print("ERROR : Can not find nor create the asked folder")
# sys.exit()
# elif(status == 1):
print("TRAIN")
for f in pbar(list_features):
filename = f[:-4]
ft_path = input_train_dir + filename
print(input_train_dir + f)
if (os.path.isfile(input_train_dir + f)):
# print(f)
ft = essentia_lowlevel_features_computation(input_train_dir, f)
# print(ft)
out_train_csv.write(filename)
for val in ft:
csv_train_file = csv_train_dir + CSV_FILE_NAME # define the name of the folder according to the previous step
cycles_train_dir, status = prsg.verify_folder(
input_train_dir, SPLITTED_FOLDER_NAME
) # Verify if folder already exists and if it is empty or not
if (status == 0):
print("ERROR : Can not find nor create the asked folder")
sys.exit()
elif (status == 1):
split_record_in_cycle(input_train_dir, csv_train_file, cycles_train_dir)
### STEP 3 : Compute features representing audio
################################################
features_train_dir, status = prsg.verify_folder(
input_train_dir, FEATURES_FOLDER_NAME
) # Verify if folder already exists and if it is empty or not
list_features = prsg.ordering_files(cycles_train_dir)
if (status == 0):
print("ERROR : Can not find nor create the asked folder")
sys.exit()
elif (status == 1):
for f in list_features:
ft_path = features_train_dir + f[:-4]
ft = essentia_lowlevel_features_computation(cycles_train_dir, f)
pickle.dump(ft, open(ft_path, 'wb'))
########
# TEST #
########
### STEP 1 : Preparing CSV file
################################################
def parsing_data_to_csv(path_to_data_folder, path_to_diagnostic_file,
path_to_csv_folder, csv_filename):
input_data_dir = path_to_data_folder
input_info = path_to_diagnostic_file
output_dir = path_to_csv_folder
# os.makedirs(output_dir, exist_ok=True)
output = output_dir + csv_filename
out_file = open(output, "w")
cpt = 0
# nb_files = (len(os.listdir(input_data_dir)))/2
nb_files = prsg.nb_files(input_data_dir) / 2
with open(input_info) as fp:
diagnostics = fp.read().splitlines()
# ordered_files = sorted(os.listdir(input_data_dir))
ordered_files = prsg.ordering_files(input_data_dir)
for filename in ordered_files:
if (filename.endswith('txt')):
cpt += 1
input = input_data_dir + filename
input_file = open(input, 'r')
content = input_file.readline()
file_id = filename[:-4]
patient_number, record_index, body_area, channel, record_tool = file_id.split(
"_")
pathology = 0
for i in range(0, len(diagnostics) - 1):
tmp_patient, tmp_pathology = diagnostics[i].split("\t")
if patient_number == tmp_patient:
if (tmp_pathology == "Asthma"):
pathology = 0
if (tmp_pathology == "LRTI"):
pathology = 1
if (tmp_pathology == "Pneumonia"):
pathology = 2
if (tmp_pathology == "Bronchiectasis"):
pathology = 3
if (tmp_pathology == "Bronchiolitis"):
pathology = 4
if (tmp_pathology == "URTI"):
pathology = 5
if (tmp_pathology == "COPD"):
pathology = 6
if (tmp_pathology == "Healthy"):
pathology = 7
while content:
start_time, end_time, crackle, wheeze = content.split('\t')
# out_file.write(patient_number+","+record_index+","+body_area+","+channel+","+record_tool+","+start_time+","+end_time+","+str(pathology)+","+crackle+","+wheeze)
out_file.write(file_id + "," + start_time + "," + end_time +
"," + str(pathology) + "," + crackle + "," +
wheeze)
content = input_file.readline()