def run(self):
if (self.RUN_READER):
read = Reader()
read.execute(self.BASE_PATH, self.READER_TYPES, self.BASE_SUBJECTS)
if (self.RUN_EXTRACTOR):
extract = Extractor()
extract.execute(self.BASE_PATH, self.BASE_WINDOW,
self.WINDOW_OVERLAP, self.BASE_SUBJECTS)
if (self.RUN_SELECTOR):
selection_results = {}
select = Selector()
for st in self.SELECTOR_SELECTION_TYPE:
for sig in self.SELECTOR_SIGNALS:
selection_results[sig] = []
selection_results[sig] = select.execute(
self.BASE_PATH, sig, self.BASE_SUBJECTS,
self.BASE_WINDOW, self.WINDOW_OVERLAP, st,
self.SELECTOR_ALL_SIGNS)
if (self.SELECTOR_ALL_SIGNS):
break
print('Results = ', st)
print('RATIO, STD')
print(selection_results)
if (self.RUN_CLASSIFIER):
# self.one_classifier_and_decision('lda', 8)
# self.one_classifier_and_decision('', 9)
self.individual('ecg', 'pca', 1)
# self.individual('ecg', 'lda', 2)
# self.individual('ecg', '', 3)
self.individual('eda', 'pca', 1)
# self.individual('eda', 'lda', 2)
# self.individual('eda', '', 3)
# self.individual('emg', 'pca', 1)
# self.individual('emg', 'lda', 2)
# self.individual('emg', '', 3)
self.individual('resp', 'pca', 1)
# self.individual('resp', 'lda', 2)
# self.individual('resp', '', 3)
self.todos('pca', 4)
# self.todos('lda', 5)
# self.todos('', 6)
self.one_classifier_and_decision('pca', 7)
def execute(self):
""" To make feature map correctly, add self.padding_size_in_model."""
padded_image_patch_size = self.image_patch_size + self.padding_size_in_model * 2
dummy = sitk.Image(self.image.GetSize(), sitk.sitkUInt8)
""" Make patch. """
etr = Extractor(image=self.image,
label=dummy,
image_patch_size=padded_image_patch_size,
label_patch_size=self.label_patch_size,
mask=self.mask)
etr.execute()
image_array_list, _ = etr.output(kind="Array")
is_cuda = torch.cuda.is_available() and True
device = torch.device("cuda" if is_cuda else "cpu")
""" Make feature maps. """
self.feature_map_list = []
with tqdm(total=len(image_array_list),
desc="Making feature maps..",
ncols=60) as pbar:
for image_array in image_array_list:
image_array = torch.from_numpy(image_array)[
None, None, ...].to(device, dtype=torch.float)
feature_map = self.model.forwardWithoutSegmentation(
image_array)
feature_map = feature_map.to("cpu").detach().numpy().astype(
np.float)
feature_map = np.squeeze(feature_map)
lower_crop_size = [0] + self.padding_size_in_model.tolist()
upper_crop_size = [0] + self.padding_size_in_model.tolist()
feature_map = croppingForNumpy(feature_map, lower_crop_size,
upper_crop_size)
self.feature_map_list.append(feature_map)
pbar.update(1)
def execute(self):
predicted_array_list = self.makeFeatureMap()
label = sitk.Image(self.image.GetSize(), sitk.sitkUInt8)
for ch in range(self.num_channel):
etr = Extractor(
image = predicted_array_list[ch],
label = label,
mask = self.mask,
image_patch_size = self.image_patch_size,
label_patch_size = self.label_patch_size
)
etr.execute()
predicted_array_list[ch], _ = etr.output(kind="Array")
predicted_array_list = np.array(predicted_array_list)
num_len = predicted_array_list.shape[1]
self.feature_map_list= []
for l in range(num_len):
self.feature_map_list.append(predicted_array_list[:, l, ...])
def execute(self, base_path, window, window_overlap, subjects):
self.window_overlap = window_overlap
self.window = window
self.calc_registers()
for i in subjects:
subject = 'S' + str(i)
print('iniciando sujeito = ', subject)
self.path = base_path + subject + '/data/'
os.chdir(self.path)
labels700 = np.loadtxt('chest_labels_filtered.txt')
self.process(labels700, 'chest', 'ecg', self.registers700)
self.process(labels700, 'chest', 'emg', self.registers700)
self.process(labels700, 'chest', 'eda', self.registers700)
self.process(labels700, 'chest', 'resp', self.registers700)
from extractor import Extractor
if __name__ == '__main__':
window = 30
window_overlap = True
path = '/Volumes/My Passport/TCC/WESAD/'
subjects = [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 14, 15, 16, 17]
extract = Extractor()
extract.execute(path, window, window_overlap, subjects)
#%%