def test_stress_test_slic():
lengths = range(100, 200, 3)
for i in range(len(lengths)):
n_segments = lengths[i] ** 3 / 10 ** 3
blob = data.binary_blobs(length=lengths[i], n_dim=3, seed=2)
blob = np.float32(blob)
slic(blob, n_segments=n_segments, compactness=2)
def test_cherry_picked_results_without_enforce_connectivity():
# fmt: off
plane = [[1,1,0,0],
[1,1,0,0],
[1,1,1,1],
[1,1,1,1]]
vol = np.asarray([plane, plane, plane])
expected = np.array([[[3, 3, 2, 2],
[3, 3, 2, 2],
[3, 3, 4, 4],
[3, 3, 4, 4]],
[[3, 3, 2, 2],
[3, 3, 2, 2],
[3, 3, 4, 4],
[3, 3, 4, 4]],
[[3, 3, 2, 2],
[3, 3, 2, 2],
[3, 3, 4, 4],
[3, 3, 4, 4]]])
# fmt: on
labels = slic(
vol,
n_segments=2,
compactness=0.1,
multichannel=False,
enforce_connectivity=False,
)
assert (labels == expected).all()
def test_slic_raises_value_error_when_input_dimention_more_than_4():
blob = data.binary_blobs(length=33, n_dim=5, seed=2)
blob = np.float32(blob)
with pytest.raises(ValueError):
labels = slic(blob, n_segments=100, compactness=3)
def test_slic_3channels_give_good_number_of_superpixels():
blob = data.binary_blobs(length=33, n_dim=4, seed=2)
blob = blob[:3]
blob = filters.gaussian(blob)
blob = img_as_float32(blob)
labels = slic(blob, n_segments=100, compactness=3)
assert len(np.unique(labels)) < 500
def test_slic_3channels_runs():
blob = data.binary_blobs(length=50, n_dim=4, seed=2)
blob = blob[..., :3]
blob = filters.gaussian(blob)
blob = img_as_float32(blob)
labels = slic(blob, n_segments=100, multichannel=False, compactness=3)
assert isinstance(labels, np.ndarray)
assert labels.ndim == 3
def test_slic_with_saved_results():
n = 200
blob = data.binary_blobs(length=n, n_dim=3, seed=2)
blob = np.float32(blob)
n_segments = n**3 // 5**3 # 5x5x5 initial segment size
labels = slic(
blob,
n_segments=n_segments,
compactness=0.01,
multichannel=False,
enforce_connectivity=True,
)
expected = np.load("tests/test_big_data.npy")
assert (labels == expected).all()
def __init__(self, img, ROI, region_size=50, compactness=5):
h, w = img.shape[0:2]
img = cv2.cvtColor(img, cv2.COLOR_BGR2LAB)
img = cv2.GaussianBlur(img, (3, 3), 0)
n_segment = int((h * w) / (region_size**2))
labels = slic(img, n_segment, compactness=compactness)
contour_mask = find_boundaries(labels)
contour_mask[ROI == 0] = 0
labels[ROI == 0] = -1
self.labels_position = self.__get_labels_position(labels)
self.__remap_labels(labels)
self.labels = labels
self.numOfPixel = np.max(labels) + 1
self.contour_mask = contour_mask
def __init__(self, img, ROI, region_size=50):
h, w = img.shape[0:2]
img = cv2.cvtColor(img, cv2.COLOR_BGR2LAB)
img = cv2.GaussianBlur(img, (3, 3), 0)
n_segment = int((h * w) / (region_size**2))
labels = slic(img, n_segment, compactness=10)
contour_mask = find_boundaries(labels)
contour_mask[ROI == 0] = 0
labels[ROI == 0] = -1
print(f'finished slic')
self.labels_position = self.__get_labels_position(labels)
print(f'finished get position')
self.__remap_labels(labels)
print(f'finished get position')
self.labels = labels
self.numOfPixel = np.max(labels) + 1
self.contour_mask = contour_mask
self.adjacent_pairs = self.__construct_adjacency(labels)
def test_slic_grayscale_runs():
blob = data.binary_blobs(length=50, n_dim=3, seed=2)
blob = img_as_float32(blob)
labels = slic(blob, n_segments=100, multichannel=False, compactness=3)
assert isinstance(labels, np.ndarray)
assert labels.ndim == 3
def test_slic_grayscale_give_good_number_of_superpixels():
blob = data.binary_blobs(length=50, n_dim=3, seed=2)
blob = img_as_float32(blob)
labels = slic(blob, n_segments=100, multichannel=False, compactness=3)
assert len(np.unique(labels)) < 150
