except:
pass
try:
os.mkdir('datasets')
except:
pass
#Downloading subjects
for subject in list_subjects:
utils.download_subject(subject,personal_access_key_id,secret_access_key)
#Reading data, extracting epochs and creating data-set
print("\n\n\n>>Creating a dataset with the subject: ",subject,"\n\n\n")
try:
#Creating epochs and preprocessing
X_train, Y_train, pick_chans = utils.create_dataset([subject])
except:
print("\n\nERROR! \nProbably your Amazon S3 credentials are not valid...\n\n")
sys.exit()
#Splitting
X_train, X_validate, Y_train, Y_validate = train_test_split(X_train, Y_train, test_size=0.2, random_state=1) #Second split the validation (20%)
#Convert labels to one-hot encodings (00010000000...)
Y_train = np_utils.to_categorical(Y_train)
Y_validate = np_utils.to_categorical(Y_validate)
#Number channels (Same for task and resting)
n_chans = len(pick_chans)
n_subjects_stage = 3 #Number of subjects load in memory in each stage of the training
#Spliting the testing subjects in different stages
subjects_stages_test = []
for i in range(0, len(list_subjects_test), n_subjects_stage):
subjects_stages_test.append(list_subjects_test[i:i + n_subjects_stage])
####################### TESTING LOOP #######################
for i in range(len(subjects_stages_test)):
####################### FIRST TEST - Whole recording #######################
#Reading data, extracting epochs and creating data-set
print("\n\n\n>>Creating a dataset with the subjects: ",
subjects_stages_test[i], "\n\n\n")
X_complete, Y_complete, pick_chans = utils.create_dataset(
subjects_stages_test[i])
#Convert labels to one-hot encodings (00010000000...)
Y_complete = np_utils.to_categorical(Y_complete)
#Number channels (Same for task and resting)
n_chans = len(pick_chans)
#Number of time points in each sample (Same for task and resting)
samples = X_complete.shape[-1]
#Kernels (To transform data to NCHW (batch_size, channel, height, width)
kernels = 1
#Convert data to NCHW (trials, kernels, channels, samples)
X_complete = X_complete.reshape(X_complete.shape[0], kernels, n_chans,
samples)