def example_checkerboard():
# generate a checkerboard
x = np.ones((10, 12))
x[::2, ::2] = -1
x[1::2, 1::2] = -1
x_noisy = x + np.random.normal(0, 1.5, size=x.shape)
x_thresh = x_noisy > .0
# create unaries
unaries = x_noisy
# as we convert to int, we need to multipy to get sensible values
unaries = (10 * np.dstack([unaries, -unaries]).copy("C")).astype(np.int32)
# create potts pairwise
pairwise = 100 * np.eye(2, dtype=np.int32)
# do simple cut
result_qpbo = binary_grid(unaries, pairwise)
# make non-labeled zero (instead of negative)
result_qpbo_vis = result_qpbo.copy()
result_qpbo_vis[result_qpbo < 0] = 0
result_qpbo_vis[result_qpbo == 0] = -1
# plot results
plt.subplot(231, title="original")
plt.imshow(x, interpolation='nearest')
plt.subplot(232, title="noisy version")
plt.imshow(x_noisy, interpolation='nearest')
plt.subplot(233, title="rounded to integers")
plt.imshow(unaries[:, :, 0], interpolation='nearest')
plt.subplot(234, title="thresholding result")
plt.imshow(x_thresh, interpolation='nearest')
plt.subplot(235, title="qpbo")
plt.imshow(result_qpbo_vis, interpolation='nearest')
plt.show()
def test_binary_grid_unaries():
# test handling on unaries for binary grid CRFs
for ds in toy.binary:
X, Y = ds(n_samples=1)
x, y = X[0], Y[0]
crf = GridCRF(inference_method="lp")
w_unaries_only = np.zeros(5)
w_unaries_only[:2] = 1.0
# test that inference with unaries only is the
# same as argmax
inf_unaries = crf.inference(x, w_unaries_only)
pw_z = np.zeros((2, 2), dtype=np.int32)
un = np.ascontiguousarray(-1000 * x).astype(np.int32)
unaries = binary_grid(un, pw_z)
assert_array_equal(inf_unaries, unaries)
assert_array_equal(inf_unaries, np.argmax(x, axis=2))
try:
assert np.mean(inf_unaries == y) > 0.5
except:
print(ds)
# check that the right thing happens on noise-free data
X, Y = ds(n_samples=1, noise=0)
inf_unaries = crf.inference(X[0], w_unaries_only)
assert_array_equal(inf_unaries, Y[0])
def example_checkerboard():
# generate a checkerboard
x = np.ones((10, 12))
x[::2, ::2] = -1
x[1::2, 1::2] = -1
x_noisy = x + np.random.normal(0, 1.5, size=x.shape)
x_thresh = x_noisy > .0
# create unaries
unaries = x_noisy
# as we convert to int, we need to multipy to get sensible values
unaries = (10 * np.dstack([unaries, -unaries]).copy("C")).astype(np.int32)
# create potts pairwise
pairwise = 100 * np.eye(2, dtype=np.int32)
# do simple cut
result_qpbo = binary_grid(unaries, pairwise)
# make non-labeled zero (instead of negative)
result_qpbo_vis = result_qpbo.copy()
result_qpbo_vis[result_qpbo < 0] = 0
result_qpbo_vis[result_qpbo == 0] = -1
# plot results
plt.subplot(231, title="original")
plt.imshow(x, interpolation='nearest')
plt.subplot(232, title="noisy version")
plt.imshow(x_noisy, interpolation='nearest')
plt.subplot(233, title="rounded to integers")
plt.imshow(unaries[:, :, 0], interpolation='nearest')
plt.subplot(234, title="thresholding result")
plt.imshow(x_thresh, interpolation='nearest')
plt.subplot(235, title="qpbo")
plt.imshow(result_qpbo_vis, interpolation='nearest')
plt.show()
def inference(self, x, w):
if w.shape != (self.size_psi,):
raise ValueError("Got w of wrong shape. Expected %s, got %s" %
(self.size_psi, w.shape))
self.inference_calls += 1
unary_param = w[0]
pairwise_params = np.array([[0, w[1]], [w[1], 0]])
if (x[:, :, 1] != 0).any():
raise ValueError("For simplicty, in binary CRFS,"
"all entries in the second feature should be 0.")
unaries = - 1000 * unary_param * x.copy()
pairwise = -1000 * pairwise_params
y = binary_grid(unaries.astype(np.int32), pairwise.astype(np.int32))
return y
def test_multinomial_grid_binary():
# test handling on unaries for multinomial grid CRFs
# on binary datasets
for ds in toy_datasets.binary:
X, Y = ds(n_samples=1)
x, y = X[0], Y[0]
crf = MultinomialGridCRF()
w_unaries_only = np.zeros(5)
w_unaries_only[:2] = 1.
# test that inference with unaries only is the
# same as argmax
inf_unaries = crf.inference(x, w_unaries_only)
pw_z = np.zeros((2, 2), dtype=np.int32)
un = np.ascontiguousarray(-1000 * x).astype(np.int32)
unaries = binary_grid(un, pw_z)
assert_array_equal(inf_unaries, unaries)
assert_array_equal(inf_unaries, np.argmax(x, axis=2))
def test_multinomial_grid_binary():
# test handling on unaries for multinomial grid CRFs
# on binary datasets
for ds in toy_datasets.binary:
X, Y = ds(n_samples=1)
x, y = X[0], Y[0]
crf = MultinomialGridCRF()
w_unaries_only = np.zeros(5)
w_unaries_only[:2] = 1.
# test that inference with unaries only is the
# same as argmax
inf_unaries = crf.inference(x, w_unaries_only)
pw_z = np.zeros((2, 2), dtype=np.int32)
un = np.ascontiguousarray(
-1000 * x).astype(np.int32)
unaries = binary_grid(un, pw_z)
assert_array_equal(inf_unaries, unaries)
assert_array_equal(inf_unaries, np.argmax(x, axis=2))
def example_binary():
# generate trivial data
x = np.ones((10, 12))
x[:, 6:] = -1
x_noisy = x + np.random.normal(0, 0.8, size=x.shape)
x_thresh = x_noisy > .0
# create unaries
unaries = x_noisy
# as we convert to int, we need to multipy to get sensible values
unaries = (10 * np.dstack([unaries, -unaries]).copy("C")).astype(np.int32)
# create potts pairwise
pairwise = -10 * np.eye(2, dtype=np.int32)
# do simple cut
result_qpbo = binary_grid(unaries, pairwise)
#result_qpbo = alpha_expansion_grid(unaries, pairwise)
#result_gc = cut_simple(unaries, pairwise)
# use the gerneral graph algorithm
# first, we construct the grid graph
inds = np.arange(x.size).reshape(x.shape)
horz = np.c_[inds[:, :-1].ravel(), inds[:, 1:].ravel()]
vert = np.c_[inds[:-1, :].ravel(), inds[1:, :].ravel()]
edges = np.vstack([horz, vert]).astype(np.int32)
# we flatten the unaries
result_graph = binary_graph(edges, unaries.reshape(-1, 2), pairwise)
# plot results
plt.subplot(231, title="original")
plt.imshow(x, interpolation='nearest')
plt.subplot(232, title="noisy version")
plt.imshow(x_noisy, interpolation='nearest')
plt.subplot(233, title="rounded to integers")
plt.imshow(unaries[:, :, 0], interpolation='nearest')
plt.subplot(234, title="thresholding result")
plt.imshow(x_thresh, interpolation='nearest')
plt.subplot(235, title="qpbo")
plt.imshow(result_qpbo, interpolation='nearest')
plt.subplot(236, title="qpbo graph")
plt.imshow(result_graph.reshape(x.shape), interpolation='nearest')
plt.show()
def example_binary():
# generate trivial data
x = np.ones((10, 12))
x[:, 6:] = -1
x_noisy = x + np.random.normal(0, 0.8, size=x.shape)
x_thresh = x_noisy > .0
# create unaries
unaries = x_noisy
# as we convert to int, we need to multipy to get sensible values
unaries = (10 * np.dstack([unaries, -unaries]).copy("C")).astype(np.int32)
# create potts pairwise
pairwise = -10 * np.eye(2, dtype=np.int32)
# do simple cut
result_qpbo = binary_grid(unaries, pairwise)
#result_qpbo = alpha_expansion_grid(unaries, pairwise)
#result_gc = cut_simple(unaries, pairwise)
# use the gerneral graph algorithm
# first, we construct the grid graph
inds = np.arange(x.size).reshape(x.shape)
horz = np.c_[inds[:, :-1].ravel(), inds[:, 1:].ravel()]
vert = np.c_[inds[:-1, :].ravel(), inds[1:, :].ravel()]
edges = np.vstack([horz, vert]).astype(np.int32)
# we flatten the unaries
result_graph = binary_graph(edges, unaries.reshape(-1, 2), pairwise)
# plot results
plt.subplot(231, title="original")
plt.imshow(x, interpolation='nearest')
plt.subplot(232, title="noisy version")
plt.imshow(x_noisy, interpolation='nearest')
plt.subplot(233, title="rounded to integers")
plt.imshow(unaries[:, :, 0], interpolation='nearest')
plt.subplot(234, title="thresholding result")
plt.imshow(x_thresh, interpolation='nearest')
plt.subplot(235, title="qpbo")
plt.imshow(result_qpbo, interpolation='nearest')
plt.subplot(236, title="qpbo graph")
plt.imshow(result_graph.reshape(x.shape), interpolation='nearest')
plt.show()
def test_binary_grid_unaries():
# test handling on unaries for binary grid CRFs
for ds in toy_datasets.binary:
X, Y = ds(n_samples=1)
x, y = X[0], Y[0]
crf = BinaryGridCRF()
w_unaries_only = np.zeros(2)
w_unaries_only[0] = 1.
# test that inference with unaries only is the
# same as argmax
inf_unaries = crf.inference(x, w_unaries_only)
pw_z = np.zeros((2, 2), dtype=np.int32)
un = np.ascontiguousarray(-1000 * x).astype(np.int32)
unaries = binary_grid(un, pw_z)
assert_array_equal(inf_unaries, unaries)
assert_array_equal(inf_unaries, np.argmax(x, axis=2))
try:
assert (np.mean(inf_unaries == y) > 0.5)
except:
print(ds)
def test_binary_grid_unaries():
# test handling on unaries for binary grid CRFs
for ds in toy_datasets.binary:
X, Y = ds(n_samples=1)
x, y = X[0], Y[0]
crf = BinaryGridCRF()
w_unaries_only = np.zeros(2)
w_unaries_only[0] = 1.
# test that inference with unaries only is the
# same as argmax
inf_unaries = crf.inference(x, w_unaries_only)
pw_z = np.zeros((2, 2), dtype=np.int32)
un = np.ascontiguousarray(
-1000 * x).astype(np.int32)
unaries = binary_grid(un, pw_z)
assert_array_equal(inf_unaries, unaries)
assert_array_equal(inf_unaries, np.argmax(x, axis=2))
try:
assert(np.mean(inf_unaries == y) > 0.5)
except:
print(ds)
