def degrade(reference, rotation, translation, scale, drop, duplications, noise):
from numpy import delete
from coherent_point_drift.geometry import rotationMatrix, rigidXform
from itertools import chain, repeat
points = delete(reference, drop, axis=0)
rotation_matrix = rotationMatrix(*rotation)
indices = chain.from_iterable(repeat(i, n) for i, n in enumerate(duplications))
return rigidXform(points, rotation_matrix, translation, scale)[list(indices)] + noise
def test_cpd_prior():
from coherent_point_drift.align import driftRigid
rng = np.random.RandomState(4)
ndim = 3
R = rotationMatrix(*next(randomRotations(ndim, rng)))
t = rng.normal(size=ndim)
s = rng.normal(size=1)[0]
X = rng.normal(size=(10, ndim))
Y = RigidXform(R, t, s) @ X
_, cpd = last(islice(driftRigid(X, Y, w=np.eye(len(X))), 200))
ls = align(X, Y)
np.testing.assert_almost_equal(cpd.R, ls.R)
np.testing.assert_almost_equal(cpd.t, ls.t)
np.testing.assert_almost_equal(cpd.s, ls.s)
def test_cpd_prior():
from coherent_point_drift.align import driftRigid
rng = np.random.RandomState(4)
ndim = 3
R = rotationMatrix(*next(randomRotations(ndim, rng)))
t = rng.normal(size=ndim)
s = rng.normal(size=1)[0]
X = rng.normal(size=(10, ndim))
Y = RigidXform(R, t, s) @ X
_, cpd = last(islice(driftRigid(X, Y, w=np.eye(len(X))), 200))
ls = align(X, Y)
np.testing.assert_almost_equal(cpd.R, ls.R)
np.testing.assert_almost_equal(cpd.t, ls.t)
np.testing.assert_almost_equal(cpd.s, ls.s)
def test_least_squares():
rng = np.random.RandomState(4)
ndim = 3
R = rotationMatrix(*next(randomRotations(ndim, rng)))
t = rng.normal(size=ndim)
s = rng.lognormal(size=1)[0]
xform = RigidXform(R, t, s)
X = rng.normal(size=(10, ndim))
Y = xform @ X
alignment = align(X, Y)
expected = xform.inverse
np.testing.assert_almost_equal(alignment.R, expected.R)
np.testing.assert_almost_equal(alignment.t, expected.t)
np.testing.assert_almost_equal(alignment.s, expected.s)
np.testing.assert_almost_equal(alignment @ Y, X)
def test_least_squares():
rng = np.random.RandomState(4)
ndim = 3
R = rotationMatrix(*next(randomRotations(ndim, rng)))
t = rng.normal(size=ndim)
s = rng.lognormal(size=1)[0]
xform = RigidXform(R, t, s)
X = rng.normal(size=(10, ndim))
Y = xform @ X
alignment = align(X, Y)
expected = xform.inverse
np.testing.assert_almost_equal(alignment.R, expected.R)
np.testing.assert_almost_equal(alignment.t, expected.t)
np.testing.assert_almost_equal(alignment.s, expected.s)
np.testing.assert_almost_equal(alignment @ Y, X)