def test_weighted_sampler_negative_weights(self):
"""Test for exception with negative weights."""
vertices, faces = mesh_test_utils.create_square_triangle_mesh()
face_wts = np.array([-0.3, 0.1, 0.5, 0.6], dtype=np.float32)
num_samples = 10
error_msg = "Condition x >= y did not hold."
with self.assertRaisesRegexp(tf.errors.InvalidArgumentError,
error_msg):
sampler.weighted_random_sample_triangle_mesh(vertices,
faces,
num_samples,
face_weights=face_wts)
def sampler_fn(vertices):
sample_pts, _ = sampler.weighted_random_sample_triangle_mesh(
vertices,
index_tensor,
num_samples,
weights_tensor,
seed=[0, 1],
stateless=True)
return sample_pts
def test_weighted_random_sample(self):
"""Test for provided face weights."""
faces = np.array([[0, 1, 2], [2, 1, 3]], dtype=np.int32)
vertex_attributes = np.array([[0.], [0.], [1.], [1.]],
dtype=np.float32)
# Equal face weights, mean of sampled attributes = 0.5.
expected_mean = np.array([0.5], dtype=np.float32)
sample_pts, _ = sampler.weighted_random_sample_triangle_mesh(
vertex_attributes,
faces,
num_samples=1000000,
face_weights=(0.5, 0.5))
self.assertAllClose(expected_mean,
tf.reduce_mean(input_tensor=sample_pts, axis=-2),
atol=1e-3)
# Face weights biased towards second face, mean > 0.5
sample_pts, _ = sampler.weighted_random_sample_triangle_mesh(
vertex_attributes,
faces,
num_samples=1000000,
face_weights=(0.2, 0.8))
self.assertGreater(tf.reduce_mean(input_tensor=sample_pts, axis=-2),
expected_mean)