def test_larger_contour():
locus = pyefd.calculate_dc_coefficients(contour_1)
coeffs = pyefd.elliptic_fourier_descriptors(contour_1, order=50)
number_of_points = contour_1.shape[0]
reconstruction = pyefd.reconstruct_contour(coeffs, locus, number_of_points)
hausdorff_distance, _, _ = directed_hausdorff(contour_1, reconstruction)
assert hausdorff_distance < 0.4
def create_contour(mask):
"""Turn as mask into a contour polygon. The polygon is smoothed
via Fourier descriptors
Parameters
----------
mask : 2D numpy array
mask of a region
Returns
-------
recon : Nx2 numpy array
coordinates of contour
"""
# extract contour by recovering contour of filled mask
reg = skimage.measure.regionprops_table(
skimage.morphology.label(mask),
properties=(
"label",
"area",
"filled_area",
"eccentricity",
"extent",
"filled_image",
"coords",
"bbox",
),
)
regp = pd.DataFrame(reg).sort_values(by="filled_area", ascending=False)
new_mask2 = np.zeros(mask.shape)
new_mask2[regp.iloc[0]["bbox-0"]:regp.iloc[0]["bbox-2"],
regp.iloc[0]["bbox-1"]:regp.
iloc[0]["bbox-3"], ] = regp.iloc[0].filled_image
area = regp.iloc[0].filled_area
contour = skimage.measure.find_contours(new_mask2, 0.8)
sel_contour = contour[np.argmax([len(x) for x in contour])]
# calculate a smoothed verions using a reconstruction via Fourier descriptors
coeffs = elliptic_fourier_descriptors(sel_contour,
order=100,
normalize=False)
coef0 = calculate_dc_coefficients(sel_contour)
recon = reconstruct_contour(coeffs, locus=coef0, num_points=1500)
return recon, area
def test_fit_1():
n = 300
locus = pyefd.calculate_dc_coefficients(contour_1)
coeffs = pyefd.elliptic_fourier_descriptors(contour_1, order=20)
t = np.linspace(0, 1.0, n)
xt = np.ones((n, )) * locus[0]
yt = np.ones((n, )) * locus[1]
for n in pyefd._range(coeffs.shape[0]):
xt += (coeffs[n, 0] * np.cos(2 * (n + 1) * np.pi * t)) + \
(coeffs[n, 1] * np.sin(2 * (n + 1) * np.pi * t))
yt += (coeffs[n, 2] * np.cos(2 * (n + 1) * np.pi * t)) + \
(coeffs[n, 3] * np.sin(2 * (n + 1) * np.pi * t))
assert True
def test_reconstruct_simple_contour():
simple_polygon = np.array([[1., 1.], [0., 1.], [0., 0.], [1., 0.],
[1., 1.]])
number_of_points = simple_polygon.shape[0]
locus = pyefd.calculate_dc_coefficients(simple_polygon)
coeffs = pyefd.elliptic_fourier_descriptors(simple_polygon, order=30)
reconstruction = pyefd.reconstruct_contour(coeffs, locus, number_of_points)
# with only 2 decimal accuracy it is a bit of a course test, but it will do
# directly comparing the two polygons will only work here, because efd coefficients will be cycle-consistent
np.testing.assert_array_almost_equal(simple_polygon,
reconstruction,
decimal=2)
hausdorff_distance, _, _ = directed_hausdorff(reconstruction,
simple_polygon)
assert hausdorff_distance < 0.01
def test_fit_1():
n = 300
locus = pyefd.calculate_dc_coefficients(contour_1)
coeffs = pyefd.elliptic_fourier_descriptors(contour_1, order=20)
t = np.linspace(0, 1.0, n)
xt = np.ones((n,)) * locus[0]
yt = np.ones((n,)) * locus[1]
for n in pyefd._range(coeffs.shape[0]):
xt += (coeffs[n, 0] * np.cos(2 * (n + 1) * np.pi * t)) + \
(coeffs[n, 1] * np.sin(2 * (n + 1) * np.pi * t))
yt += (coeffs[n, 2] * np.cos(2 * (n + 1) * np.pi * t)) + \
(coeffs[n, 3] * np.sin(2 * (n + 1) * np.pi * t))
assert True
def test_centroid(self):
geom1 = 'POLYGON((0 0, 1 0, 1 1, 0 1, 0 0))'
geom1 = gv.vectorize_wkt(geom1)
with self.subTest('It does not accept a numpy ndarray'):
with self.assertRaises(AssertionError):
centroid(geom1)
with self.subTest('It rejects 3D geometries'):
with self.assertRaises(AssertionError):
centroid(torch.rand((10, 3)))
with self.subTest('Our stand-in centroid function does the same as pyefd'):
geom2 = 'POLYGON((1 1, 0 1, 0 0, 1 0, 1 1))'
geom2 = gv.vectorize_wkt(geom2)
coords_batch = geom2[:, :2]
coords_batch = coords_batch.reshape(1, geom2.shape[0], 2)
polygon2_tensor = torch.from_numpy(coords_batch)
pyefd_centroid = pyefd.calculate_dc_coefficients(coords_batch[0])
pytorch_centroid = centroid(polygon2_tensor)
np.testing.assert_array_almost_equal(pyefd_centroid, pytorch_centroid[0])
with self.subTest('It correctly calculates centroids for batches'):
geom2 = 'POLYGON((1 1, 0 1, 0 0, 1 0, 1 1))'
geom2 = gv.vectorize_wkt(geom2)
coords_batch = geom2[:, :2]
coords_batch = coords_batch.reshape(1, geom2.shape[0], 2)
polygon2_tensor = torch.from_numpy(coords_batch)
batch_size = 6
batch = polygon2_tensor.repeat(batch_size, 1, 1)
multiply_range = torch.range(1., 6., dtype=batch.dtype).reshape((batch_size, 1, 1))
batch = batch * multiply_range
reference_centroids = np.arange(1, batch_size + 1)
reference_centroids = reference_centroids.reshape(batch_size, 1) * 0.5
reference_centroids = reference_centroids.repeat(2, axis=1)
batch_centroids = centroid(batch)
np.testing.assert_array_almost_equal(reference_centroids, batch_centroids.numpy())
def test_locus():
locus = pyefd.calculate_dc_coefficients(contour_1)
np.testing.assert_array_almost_equal(locus,
np.mean(contour_1, axis=0),
decimal=0)
"""Elliptic Fourier Descriptors
parametrizing a closed countour (in red)"""
import vedo, pyefd
shapes = vedo.load(vedo.dataurl+'timecourse1d.npy')
sh = shapes[55]
sr = vedo.Line(sh).mirror('x').reverse()
sm = vedo.merge(sh, sr).c('red5').lw(3)
pts = sm.points()[:,(0,1)]
rlines = []
for order in range(5,30, 5):
coeffs = pyefd.elliptic_fourier_descriptors(pts, order=order, normalize=False)
a0, c0 = pyefd.calculate_dc_coefficients(pts)
rpts = pyefd.reconstruct_contour(coeffs, locus=(a0,c0), num_points=400)
color = vedo.colorMap(order, "Blues", 5,30)
rline = vedo.Line(rpts).lw(3).c(color)
rlines.append(rline)
sm.z(0.1) # move it on top so it's visible
vedo.show(sm, *rlines, __doc__, axes=1, bg='k', size=(1190, 630), zoom=1.8)
# create fourier descriptors and save statistics
normalized_coefficients = np.zeros((dataset.n_images, n_harmonics, 4))
errors = np.zeros(dataset.n_images)
for idx in range(dataset.n_images):
im = dataset.get_image(idx).squeeze().numpy()
contour = get_contour(im, threshold=threshold)
if contour is not None:
# elliptical Fourier descriptor's coefficients.
coeffs = pyefd.elliptic_fourier_descriptors(contour,
order=n_harmonics)
# normalize coeffs
normalized_coeffs, L = pyefd.normalize_efd(deepcopy(coeffs))
# recon_contour
if len(contour) > 0:
locus = pyefd.calculate_dc_coefficients(contour)
else:
locus = (0, 0)
recon_contour = pyefd.reconstruct_contour(coeffs,
locus=locus,
num_points=300)
# error measure
error = calc_contours_distance(contour,
recon_contour) / (2 * L) * 100
normalized_coefficients[idx] = normalized_coeffs
errors[idx] = error
print('{} harmonics: error {}(mean); {}(std); {} (min); {}(max); {}(90p)'.
format(n_harmonics, errors.mean(), errors.std(), errors.min(),
errors.max(), np.percentile(errors, 90)))
# diamond
# rCos = np.array([0.5, 0.427, 0.0, 0.0732])
# zSin = np.array([0.0, 0.427, 0.0, -0.0732])
# D
# rCos = np.array([3.0, 0.991, 0.136])
# zSin = np.array([0.0, 1.409, -0.118])
# belt
# rCos = np.array([3.0, 0.453, 0.0, 0.0 ])
# zSin = np.array([0.0, 0.6 , 0.0, 0.196])
# ellipse
# rCos = np.array([3.0, 1.0])
# zSin = np.array([0.0, 3.0])
# evaluate curve geometry given as Fourier coefficients above
n = 300
contour = np.zeros([n,2])
theta = np.linspace(0.0, 2.0*np.pi, n)
mMax = len(rCos)
for m in range(mMax):
contour[:,0] += rCos[m] * np.cos(m*theta)
contour[:,1] += zSin[m] * np.sin(m*theta)
# apply pyefd to get elliptic Fourier descriptors
coeffs = pyefd.elliptic_fourier_descriptors(contour)
a0, c0 = pyefd.calculate_dc_coefficients(contour)
pyefd.plot_efd(coeffs, locus=(a0,c0), contour=contour)
def test_locus():
locus = pyefd.calculate_dc_coefficients(contour_1)
np.testing.assert_array_almost_equal(locus, np.mean(contour_1, axis=0), decimal=0)
