def _augment(img, lbl):
"""An image augmentation function."""
img = add_gaussian_offset(img, sigma=1.0)
for a in range(3):
[img, lbl] = flip([img, lbl], axis=a)
return img, lbl
def __getitem__(self, index):
num_file = index // 70
num_index = index % 70
batch_file = self.datafiles[index // 70]
t1 = sitk.GetArrayFromImage(sitk.ReadImage(batch_file))
t1 = t1[30:-30, ::-1, :]
t1 = t1[num_index]
t1 = whitening(t1)
t1 = normalise_one_one(t1)
t1 = flip(t1)
im = np.expand_dims(t1, axis=-1).astype(np.float32)
im = resize(im, [128, 128], mode='constant', anti_aliasing=True)
im = torch.tensor(im, dtype=torch.float).permute(2, 0, 1)
return im
def _augment(img, lbl):
"""An image augmentation function"""
img = add_gaussian_noise(img, sigma=0.1)
[img, lbl] = flip([img, lbl], axis=1)
return img, lbl
def _augment(img):
"""An image augmentation function"""
return flip(img, axis=2)
def _augment(img, lbl):
"""An image augmentation function"""
img = add_gaussian_noise(img, sigma=0.1)
[img, lbl] = flip([img, lbl], axis=1)
return img, lbl
def _augment(img):
"""An image augmentation function"""
return flip(img, axis=2)
def _augment(img):
return flip(img, axis=2)
def _augment(img, lbl):
if (np.random.randint(0, 10) < 3):
img = scipy.ndimage.gaussian_filter(img, sigma=0.25)
[img, lbl] = flip([img, lbl], axis=0)
return img, lbl
