def test_check_drop_3d():
image = np.zeros((10, 10, 10, 2))
image[5:, :, :, 0] = 1
image[5:, :5, :, 1] = 1
images = np.array([image for _ in range(2)])
indice = get_patch_indice((10, 10, 10), (5, 5, 5), 0)
indice = list(product([0, 1], indice))
res = check_drop(images, indice, (5, 5, 5), 0.4, None)[0]
expected = [
False, False, False, False, True, True, True, True
] * 2
assert np.all(res == expected)
res = check_drop(images, indice, (5, 5, 5), 0.9, None)[0]
expected = [
False, False, False, False, True, True, False, False
] * 2
assert np.all(res == expected)
res = check_drop(images, indice, (5, 5, 5), 0.9, 0)[0]
expected = [
False, False, False, False, True, True, True, True
] * 2
assert np.all(res == expected)
def test_get_patch_indice():
indice = get_patch_indice((20, 20, 20), (10, 10, 10), 0)
expected = [(0, 0, 0), (0, 0, 10),
(0, 10, 0), (0, 10, 10),
(10, 0, 0), (10, 0, 10),
(10, 10, 0), (10, 10, 10)]
assert np.all(indice == expected)
indice = get_patch_indice((20, 20, 20), (10, 10, 10), 0.5)
expected = [(0, 0, 0), (0, 0, 5), (0, 0, 10),
(0, 5, 0), (0, 5, 5), (0, 5, 10),
(0, 10, 0), (0, 10, 5), (0, 10, 10),
(5, 0, 0), (5, 0, 5), (5, 0, 10),
(5, 5, 0), (5, 5, 5), (5, 5, 10),
(5, 10, 0), (5, 10, 5), (5, 10, 10),
(10, 0, 0), (10, 0, 5), (10, 0, 10),
(10, 5, 0), (10, 5, 5), (10, 5, 10),
(10, 10, 0), (10, 10, 5), (10, 10, 10)]
assert np.all(indice == expected)
indice = get_patch_indice((20, 20, 19), (10, 10, 10), 0.5)
expected = [(0, 0, 0), (0, 0, 5), (0, 0, 9),
(0, 5, 0), (0, 5, 5), (0, 5, 9),
(0, 10, 0), (0, 10, 5), (0, 10, 9),
(5, 0, 0), (5, 0, 5), (5, 0, 9),
(5, 5, 0), (5, 5, 5), (5, 5, 9),
(5, 10, 0), (5, 10, 5), (5, 10, 9),
(10, 0, 0), (10, 0, 5), (10, 0, 9),
(10, 5, 0), (10, 5, 5), (10, 5, 9),
(10, 10, 0), (10, 10, 5), (10, 10, 9)]
assert np.all(indice == expected)
def test_get_patches_2d():
image = np.zeros((10, 10, 2))
image[5:, :, 0] = 1
image[5:, :5, 1] = 1
images = np.array([image for _ in range(2)])
target = np.zeros((10, 10, 1))
target[5:, :5, 0] = 1
targets = np.array([target for _ in range(2)])
indice = get_patch_indice((10, 10), (5, 5), 0)
patches, labels = get_patches(images, target=targets,
patch_indice=indice, patch_size=(5, 5),
stratified=True, drop_fraction=0.4,
check_drop_channel=0, batch_size=2)
half_patches = np.zeros((5, 5, 2))
half_patches[..., 0] = 1
full_patches = np.ones((5, 5, 2))
assert len(patches) == 4
full_patches_num = 0
half_patches_num = 0
for patch, label in zip(patches[:2], labels[:2]):
if np.all(full_patches == patch):
assert np.all(label == np.ones((5, 5, 1)))
full_patches_num += 1
elif np.all(half_patches == patch):
assert np.all(label == np.zeros((5, 5, 1)))
half_patches_num += 1
assert full_patches_num == 1 and half_patches_num == 1
full_patches_num = 0
half_patches_num = 0
for patch, label in zip(patches[2:], labels[2:]):
if np.all(full_patches == patch):
assert np.all(label == np.ones((5, 5, 1)))
full_patches_num += 1
elif np.all(half_patches == patch):
assert np.all(label == np.zeros((5, 5, 1)))
half_patches_num += 1
assert full_patches_num == 1 and half_patches_num == 1
def test_get_patch_indice_error():
with pytest.raises(ValueError):
get_patch_indice((2, 3, 4), (13, 13), 0)
def test_get_patches_3d():
image = np.zeros((10, 10, 10, 2))
image[5:, :, :, 0] = 1
image[5:, :5, :, 1] = 1
images = np.array([image for _ in range(2)])
indice = get_patch_indice((10, 10, 10), (5, 5, 5), 0)
# Only patches, in order
patches = get_patches(images,
patch_indice=indice, patch_size=(5, 5, 5),
stratified=False, drop_fraction=0,
check_drop_channel=None)
empty_patches = np.zeros((5, 5, 5, 2))
half_patches = np.zeros((5, 5, 5, 2))
half_patches[..., 0] = 1
full_patches = np.ones((5, 5, 5, 2))
expected = np.array([
empty_patches, empty_patches, empty_patches, empty_patches,
full_patches, full_patches, half_patches, half_patches
] * 2)
assert len(patches) == 16
assert np.all(patches == expected)
# Only patches, in order, drop half
patches = get_patches(images,
patch_indice=indice, patch_size=(5, 5, 5),
stratified=False, drop_fraction=0.4,
check_drop_channel=None)
expected = np.array([
full_patches, full_patches, half_patches, half_patches
] * 2)
assert len(patches) == 8
assert np.all(patches == expected)
# Only patches, in order, only keep full
patches = get_patches(images,
patch_indice=indice, patch_size=(5, 5, 5),
stratified=False, drop_fraction=0.9,
check_drop_channel=None)
expected = np.array([
full_patches, full_patches
] * 2)
assert len(patches) == 4
assert np.all(patches == expected)
# Only patches, in order, drop half based on one channel
patches = get_patches(images,
patch_indice=indice, patch_size=(5, 5, 5),
stratified=False, drop_fraction=0.4,
check_drop_channel=0)
expected = np.array([
full_patches, full_patches, half_patches, half_patches
] * 2)
assert len(patches) == 8
assert np.all(patches == expected)
# patches, targets, stratified, without batch_size
target = np.zeros((10, 10, 10, 1))
target[5:, :5, :, 0] = 1
targets = np.array([target for _ in range(2)])
patches, labels = get_patches(images, target=targets,
patch_indice=indice, patch_size=(5, 5, 5),
stratified=True, drop_fraction=0.4,
check_drop_channel=0)
assert len(patches) == 8
full_patches_num = 0
half_patches_num = 0
for patch, label in zip(patches, labels):
if np.all(full_patches == patch):
assert np.all(label == np.ones((5, 5, 5, 1)))
full_patches_num += 1
elif np.all(half_patches == patch):
assert np.all(label == np.zeros((5, 5, 5, 1)))
half_patches_num += 1
assert full_patches_num == 4 and half_patches_num == 4
# patches, targets, stratified based on target
patches, labels = get_patches(images, target=targets,
patch_indice=indice, patch_size=(5, 5, 5),
stratified=True, drop_fraction=0.4,
check_drop_channel=0, batch_size=4)
assert len(patches) == 8
full_patches_num = 0
half_patches_num = 0
for patch, label in zip(patches[:4], labels[:4]):
if np.all(full_patches == patch):
assert np.all(label == np.ones((5, 5, 5, 1)))
full_patches_num += 1
elif np.all(half_patches == patch):
assert np.all(label == np.zeros((5, 5, 5, 1)))
half_patches_num += 1
assert full_patches_num == 2 and half_patches_num == 2
full_patches_num = 0
half_patches_num = 0
for patch, label in zip(patches[4:], labels[4:]):
if np.all(full_patches == patch):
assert np.all(label == np.ones((5, 5, 5, 1)))
full_patches_num += 1
elif np.all(half_patches == patch):
assert np.all(label == np.zeros((5, 5, 5, 1)))
half_patches_num += 1
assert full_patches_num == 2 and half_patches_num == 2
targets = f['y'][:index]
return normalize(images), targets
if __name__ == "__main__":
images, targets = load_images()
patch_size = (128, 128, 128)
overlap = 0.5
stratified = True
batch_size = 8
drop_fraction = 0.1
check_drop_channel = 0
indice = get_patch_indice(images.shape[1:-1], patch_size, overlap)
# print(indice, len(indice))
patches, labels = get_patches(images, targets, indice, patch_size,
stratified=stratified,
batch_size=batch_size,
drop_fraction=drop_fraction,
check_drop_channel=check_drop_channel)
print(patches.shape)
fig, ax = plt.subplots(1, 1)
ax.axis('off')
initialize = False
mask_levels = [0.5]