def test_data_loader_3D(self, batch_size=1):
for j in range(self.test_samples.size):
dwi = self.test_hdf5s[j]['dwi1'][:]
dwi = utils.min_max_normalize(dwi)
dwi = np.expand_dims(dwi.transpose(0, 3, 1, 2), axis=0)
t2ax = self.test_hdf5s[j]['t2ax_cropped'][:]
t2ax = utils.min_max_normalize(t2ax)
t2ax = t2ax.transpose(2, 0, 1)
t2ax = np.expand_dims(np.expand_dims(t2ax, axis=0), axis=0)
label = np.expand_dims(np.expand_dims(self.test_labels[j], axis=0),
axis=0)
yield dwi, t2ax, label
def data_loader_3D(self, batch_size=1):
for j in range(self.all_samples.size):
print(self.subject_ids[j])
try:
dwi = self.sample_hdf5s[j]['dwi1'][:]
except:
print('dwi1 for {} does not exist'.format(self.subject_ids[j]))
sys.exit(1)
dwi = utils.min_max_normalize(dwi)
dwi = np.expand_dims(dwi.transpose(0, 3, 1, 2), axis=0)
t2ax = self.sample_hdf5s[j]['t2ax_cropped'][:]
t2ax = utils.min_max_normalize(t2ax)
t2ax = t2ax.transpose(2, 0, 1)
t2ax = np.expand_dims(np.expand_dims(t2ax, axis=0), axis=0)
label = np.expand_dims(np.expand_dims(self.test_labels[j], axis=0),
axis=0)
yield dwi, t2ax, label
def generate_cam(activation, fc_weights, class_idx):
activation = reshape_for_removing_batch_size(activation)
class_idx = reshape_for_removing_batch_size(class_idx)
channel, height, width = activation.shape
cam = torch.zeros([height, width], dtype=torch.float)
fc_weight = fc_weights[class_idx]
fc_weight = fc_weight.view(fc_weight.shape[1])
for c in range(channel):
cam += fc_weight[c] * activation[c]
normalized_cam = min_max_normalize(cam)
scaled_cam = np.uint8((normalized_cam * 255).int().numpy())
return cv2.resize(scaled_cam, (IMG_SIZE, IMG_SIZE))
def main(args):
func = args.func
side = args.side
if args.file_in != None and args.file_out != None and arg.mask != None:
image_path = args.file_in
out_path = args.file_out
mask_path = args.mask
mask_file = utils.read_file(mask_path)
image_file = utils.read_file(image_path)
masked = utils.mask(image_file[0], mask_file[0], side)
if func == "mask":
utils.save_file(masked, image_file[1], image_file[2], out_path)
elif func == "normalize":
normalized = utils.min_max_normalize(masked)
utils.save_file(normalized, image_file[1], image_file[2], out_path)
else:
print('func error')
elif args.folder_in != None :
folder_in = args.folder_in
if args.folder_out != None:
folder_out = args.folder_out
else:
folder_out = folder_in
image_folder = os.path.join(folder_in, 'image')
mask_folder = os.path.join(folder_in, 'mask')
image_files = []
for file in os.listdir(image_folder):
if file.endswith((".mgz", ".nii", ".nii.gz")):
image_files.append(file)
for file in image_files:
image_path = os.path.join(image_folder, file)
mask_path = os.path.join(mask_folder, file)
if not os.path.isfile(mask_path):
break
image_file = utils.read_file(image_path)
mask_file = utils.read_file(mask_path)
masked = utils.mask(image_file[0], mask_file[0], side)
if func == "mask":
if file.endswith((".mgz", ".nii")):
out_file = file[:-4] + '_masked.nii.gz'
elif file.endswith([".nii.gz"]):
out_file = file[:-7] + '_masked.nii.gz'
out_path = os.path.join(folder_out, 'masked')
if not os.path.exists(out_path):
os.makedirs(out_path)
out_path = os.path.join(out_path, out_file)
utils.save_file(masked, image_file[1], image_file[2], out_path)
elif func == "normalize":
normalized = utils.min_max_normalize(masked)
if file.endswith((".mgz", ".nii")):
out_file = file[:-4] + '_normalized.nii.gz'
elif file.endswith([".nii.gz"]):
out_file = file[:-7] + '_normalized.nii.gz'
out_path = os.path.join(folder_out, 'normalized')
if not os.path.exists(out_path):
os.makedirs(out_path)
out_path = os.path.join(out_path, out_file)
utils.save_file(normalized, image_file[1], image_file[2], out_path)
else:
print('func error')
else:
print('parameter error')
def _normalize(image, output):
normalized = utils.min_max_normalize(image[0])
utils.save_file(normalized, image[1], image[2], output)
if timestamp not in turnstile:
turnstile[timestamp] = {}
if stamp['STATION'] not in turnstile[timestamp]:
turnstile[timestamp][stamp['STATION']] = 0
turnstile[timestamp][
stamp['STATION']] += stamp['ENTRIES'] - stamp['EXITS']
print('Normalizing...')
for time in list(turnstile.keys()):
values = []
for _, key in enumerate(list(turnstile[time].keys())):
values.append(turnstile[time][key])
values = min_max_normalize(values)
for index, key in enumerate(list(turnstile[time].keys())):
turnstile[time][key] = values[index]
print('Tensorizing...')
original_G = ox.graph_from_place('New York City, New York, USA',
network_type='drive')
for time in list(turnstile.keys()):
G = original_G
fig, ax = ox.plot_graph(G,
show=False,
close=False,
node_size=0,
edge_linewidth=0)