def test_outliers_line_msac():
a, b, x, samples = line_data()
sigma = 2.0
y = a * x + b + np.random.normal(scale=sigma, size=x.shape[0])
multiplier = 1.96
ratio_outliers = 0.4
outliers_max = 5.0
add_outliers(ratio_outliers, x, multiplier * sigma,
multiplier * outliers_max)
data = np.array([x, y]).transpose()
params = pyrobust.RobustEstimatorParams()
result = pyrobust.ransac_line(data, multiplier * sigma, params,
pyrobust.RansacType.MSAC)
inliers_count = (1 - ratio_outliers) * samples
confidence = 0.95 # 1.96*MAD -> 95% rejecting inliers
assert np.isclose(len(result.inliers_indices),
inliers_count,
rtol=(1 - confidence),
atol=5)
def test_outliers_line_LMedS():
a, b, x, samples = line_data()
sigma = 2.0
y = a * x + b + np.random.normal(scale=sigma, size=x.shape[0])
multiplier = 1.96
ratio_outliers = 0.4
outliers_max = 5.0
add_outliers(ratio_outliers, x, multiplier * sigma,
multiplier * outliers_max)
data = np.array([x, y]).transpose()
params = pyrobust.RobustEstimatorParams()
# can't be used with LMedS as an over-estimated sigma will make it stop early
params.use_iteration_reduction = False
result = pyrobust.ransac_line(data, multiplier, params,
pyrobust.RansacType.LMedS)
inliers_count = (1 - ratio_outliers) * samples
confidence = 0.95 # 1.96*MAD -> 95% rejecting inliers
assert np.isclose(len(result.inliers_indices),
inliers_count,
rtol=(1 - confidence),
atol=8)
def test_uniform_line_ransac():
a, b, x, samples = line_data()
scale = 2.0
y = a * x + b + np.random.rand(x.shape[0]) * scale
data = np.array([x, y]).transpose()
params = pyrobust.RobustEstimatorParams()
result = pyrobust.ransac_line(data, scale, params,
pyrobust.RansacType.RANSAC)
assert result.score == samples
assert len(result.inliers_indices) == samples
def test_normal_line_msac():
a, b, x, samples = line_data()
sigma = 2.0
y = a * x + b + np.random.normal(scale=sigma, size=x.shape[0])
multiplier = 1.96
data = np.array([x, y]).transpose()
params = pyrobust.RobustEstimatorParams()
result = pyrobust.ransac_line(data, multiplier * sigma, params,
pyrobust.RansacType.MSAC)
confidence = 0.95 # 1.96*MAD -> 95% rejecting inliers
assert np.isclose(len(result.inliers_indices),
samples,
rtol=(1 - confidence),
atol=7)
def test_outliers_line_ransac():
a, b, x, samples = line_data()
scale = 2.0
y = a * x + b + np.random.rand(x.shape[0]) * scale
ratio_outliers = 0.4
outliers_max = 5.0
add_outliers(ratio_outliers, x, scale, outliers_max)
data = np.array([x, y]).transpose()
params = pyrobust.RobustEstimatorParams()
result = pyrobust.ransac_line(data, scale, params,
pyrobust.RansacType.RANSAC)
inliers_count = (1 - ratio_outliers) * samples
assert result.score == inliers_count
assert len(result.inliers_indices) == inliers_count