def test_gmm_test():
# Tests a GMMMachine by computing scores against a model and compare to
# an old reference
ar = bob.io.base.load(datafile('dataforMAP.hdf5', __name__, path="../data/"))
# Initialize GMMMachine
n_gaussians = 5
n_inputs = 45
gmm = GMMMachine(n_gaussians, n_inputs)
gmm.means = bob.io.base.load(datafile('meansAfterML.hdf5', __name__, path="../data/"))
gmm.variances = bob.io.base.load(datafile('variancesAfterML.hdf5', __name__, path="../data/"))
gmm.weights = bob.io.base.load(datafile('weightsAfterML.hdf5', __name__, path="../data/"))
threshold = 0.001
gmm.set_variance_thresholds(threshold)
# Test against the model
score_mean_ref = -1.50379e+06
score = 0.
for v in ar: score += gmm(v)
score /= len(ar)
# Compare current results to torch3vision
assert abs(score-score_mean_ref)/score_mean_ref<1e-4
def test_gmm_MAP_2():
# Train a GMMMachine with MAP_GMMTrainer and compare with matlab reference
data = bob.io.base.load(datafile('data.hdf5', __name__, path="../data/"))
data = data.reshape((1, data.shape[0])) # make a 2D array out of it
means = bob.io.base.load(datafile('means.hdf5', __name__, path="../data/"))
variances = bob.io.base.load(datafile('variances.hdf5', __name__, path="../data/"))
weights = bob.io.base.load(datafile('weights.hdf5', __name__, path="../data/"))
gmm = GMMMachine(2,50)
gmm.means = means
gmm.variances = variances
gmm.weights = weights
map_adapt = MAP_GMMTrainer(update_means=True, update_variances=False, update_weights=False, mean_var_update_responsibilities_threshold=0.,prior_gmm=gmm, relevance_factor=4.)
gmm_adapted = GMMMachine(2,50)
gmm_adapted.means = means
gmm_adapted.variances = variances
gmm_adapted.weights = weights
#map_adapt.max_iterations = 1
#map_adapt.train(gmm_adapted, data)
bob.learn.em.train(map_adapt, gmm_adapted, data, max_iterations = 1)
new_means = bob.io.base.load(datafile('new_adapted_mean.hdf5', __name__, path="../data/"))
# print new_means[0,:]
# print gmm_adapted.means[:,0]
# Compare to matlab reference
assert equals(new_means[0,:], gmm_adapted.means[:,0], 1e-4)
assert equals(new_means[1,:], gmm_adapted.means[:,1], 1e-4)
def test_gmm_test():
# Tests a GMMMachine by computing scores against a model and compare to
# an old reference
ar = bob.io.base.load(
datafile('dataforMAP.hdf5', __name__, path="../data/"))
# Initialize GMMMachine
n_gaussians = 5
n_inputs = 45
gmm = GMMMachine(n_gaussians, n_inputs)
gmm.means = bob.io.base.load(
datafile('meansAfterML.hdf5', __name__, path="../data/"))
gmm.variances = bob.io.base.load(
datafile('variancesAfterML.hdf5', __name__, path="../data/"))
gmm.weights = bob.io.base.load(
datafile('weightsAfterML.hdf5', __name__, path="../data/"))
threshold = 0.001
gmm.set_variance_thresholds(threshold)
# Test against the model
score_mean_ref = -1.50379e+06
score = 0.
for v in ar:
score += gmm(v)
score /= len(ar)
# Compare current results to torch3vision
assert abs(score - score_mean_ref) / score_mean_ref < 1e-4
def test_netpbm():
transcode(test_utils.datafile('test.pbm', __name__)) # indexed, works fine
transcode(test_utils.datafile('test.pgm', __name__)) # indexed, works fine
transcode(test_utils.datafile('test.ppm', __name__)) # indexed, works fine
transcode(test_utils.datafile('test_2.pgm', __name__)) # indexed, works fine
transcode(test_utils.datafile('test_2.ppm', __name__)) # indexed, works fine
def test_matlab_baseline():
"""
Tests based on this matlab baseline
http://www-scf.usc.edu/~boqinggo/domainadaptation.html#intro
"""
import numpy
numpy.random.seed(10)
source_webcam = bob.io.matlab.read_matrix(datafile("webcam.mat", __name__))
webcam_labels = bob.io.matlab.read_matrix(
datafile("webcam_labels.mat", __name__))
target_dslr = bob.io.matlab.read_matrix(datafile("dslr.mat", __name__))
dslr_labels = bob.io.matlab.read_matrix(
datafile("dslr_labels.mat", __name__))
gfk_trainer = GFKTrainer(10,
subspace_dim_source=140,
subspace_dim_target=140)
gfk_machine = gfk_trainer.train(source_webcam, target_dslr)
accuracy = compute_accuracy(gfk_machine.G, source_webcam, webcam_labels,
target_dslr, dslr_labels) * 100
assert accuracy > 70
def test_cglogreg_norm_slow():
pos1 = bob.io.base.load(datafile('positives_isv.hdf5', __name__))
neg1 = bob.io.base.load(datafile('negatives_isv.hdf5', __name__))
pos2 = bob.io.base.load(datafile('positives_lda.hdf5', __name__))
neg2 = bob.io.base.load(datafile('negatives_lda.hdf5', __name__))
negatives = numpy.vstack((neg1, neg2)).T
positives = numpy.vstack((pos1, pos2)).T
T = CGLogRegTrainer(0.5, 1e-10, 10000, mean_std_norm=True)
# apply it to test data
test1 = [1., -50.]
test2 = [0.5, -86.]
res1 = machine(test1)
res2 = machine(test2)
# try the training without normalization
machine = T.train(negatives, positives)
# check that the results are at least approximately equal
# Note: lower values for epsilon and higher number of iterations improve the stability)
assert abs(machine(test1) - res1) < 1e-3
assert abs(machine(test2) - res2) < 1e-3
def test_uint8_histoPython():
# Compute the histogram of a uint8 image
input_image = bob.io.base.load(datafile('image.hdf5', "bob.ip.base"))
histo1 = bob.ip.base.histogram(input_image)
histo2 = bob.ip.base.histogram(input_image, 256)
histo3 = bob.ip.base.histogram(input_image, (0, 255), 256)
histo4 = numpy.ndarray((256,), numpy.uint64)
histo5 = numpy.ndarray((256,), numpy.uint64)
bob.ip.base.histogram(input_image, histo4)
bob.ip.base.histogram(input_image, (0, 255), histo5)
# Save the computed data
#bob.io.base.save(histo1, datafile('image_histo.hdf5', 'bob.ip.base', 'data/histo'))
histo_ref = bob.io.base.load(datafile('image_histo.hdf5', 'bob.ip.base', 'data/histo'))
assert input_image.size == histo1.sum()
assert input_image.size == histo2.sum()
assert input_image.size == histo3.sum()
assert input_image.size == histo4.sum()
assert input_image.size == histo5.sum()
assert (histo_ref == histo1).all()
assert (histo_ref == histo2).all()
assert (histo_ref == histo3).all()
assert (histo_ref == histo4).all()
assert (histo_ref == histo5).all()
def test_uint8_histoPython():
# Compute the histogram of a uint8 image
input_image = bob.io.base.load(datafile('image.hdf5', "bob.ip.base"))
histo1 = bob.ip.base.histogram(input_image)
histo2 = bob.ip.base.histogram(input_image, 256)
histo3 = bob.ip.base.histogram(input_image, (0, 255), 256)
histo4 = numpy.ndarray((256, ), numpy.uint64)
histo5 = numpy.ndarray((256, ), numpy.uint64)
bob.ip.base.histogram(input_image, histo4)
bob.ip.base.histogram(input_image, (0, 255), histo5)
# Save the computed data
#bob.io.base.save(histo1, datafile('image_histo.hdf5', 'bob.ip.base', 'data/histo'))
histo_ref = bob.io.base.load(
datafile('image_histo.hdf5', 'bob.ip.base', 'data/histo'))
assert input_image.size == histo1.sum()
assert input_image.size == histo2.sum()
assert input_image.size == histo3.sum()
assert input_image.size == histo4.sum()
assert input_image.size == histo5.sum()
assert (histo_ref == histo1).all()
assert (histo_ref == histo2).all()
assert (histo_ref == histo3).all()
assert (histo_ref == histo4).all()
assert (histo_ref == histo5).all()
def test_ztnorm_big():
my_A = bob.io.base.load(
datafile("ztnorm_eval_eval.hdf5", __name__, path="../data/"))
my_B = bob.io.base.load(
datafile("ztnorm_znorm_eval.hdf5", __name__, path="../data/"))
my_C = bob.io.base.load(
datafile("ztnorm_eval_tnorm.hdf5", __name__, path="../data/"))
my_D = bob.io.base.load(
datafile("ztnorm_znorm_tnorm.hdf5", __name__, path="../data/"))
# ZT-Norm
ref_scores = bob.io.base.load(
datafile("ztnorm_result.hdf5", __name__, path="../data/"))
scores = bob.learn.em.ztnorm(my_A, my_B, my_C, my_D)
assert (abs(scores - ref_scores) < 1e-7).all()
# T-Norm
scores = bob.learn.em.tnorm(my_A, my_C)
scores_py = tnorm(my_A, my_C)
assert (abs(scores - scores_py) < 1e-7).all()
# Z-Norm
scores = bob.learn.em.znorm(my_A, my_B)
scores_py = znorm(my_A, my_B)
assert (abs(scores - scores_py) < 1e-7).all()
def test_video_like_container():
path = datafile("testvideo.avi", "bob.bio.video.test")
video = bob.bio.video.VideoAsArray(path,
selection_style="spread",
max_number_of_frames=3)
container = bob.bio.video.VideoLikeContainer(video, video.indices)
container_path = datafile("video_like.hdf5", "bob.bio.video.test")
if regenerate_refs:
container.save(container_path)
loaded_container = bob.bio.video.VideoLikeContainer.load(container_path)
if platform.machine() == "arm64" and platform.system() == "Darwin":
raise nose.SkipTest("Skipping test on arm64 macos")
np.testing.assert_allclose(loaded_container.indices, container.indices)
np.testing.assert_allclose(loaded_container.data, container.data)
assert container == loaded_container
# test saving and loading None arrays
with tempfile.NamedTemporaryFile(suffix=".pkl") as f:
data = [None] * 10 + [1]
indices = range(11)
frame_container = bob.bio.video.VideoLikeContainer(data, indices)
frame_container.save(f.name)
loaded = bob.bio.video.VideoLikeContainer.load(f.name)
np.testing.assert_equal(loaded.indices, frame_container.indices)
np.testing.assert_equal(loaded.data, frame_container.data)
assert loaded == frame_container
def test_cglogreg_norm_slow():
pos1 = bob.io.base.load(datafile('positives_isv.hdf5', __name__))
neg1 = bob.io.base.load(datafile('negatives_isv.hdf5', __name__))
pos2 = bob.io.base.load(datafile('positives_lda.hdf5', __name__))
neg2 = bob.io.base.load(datafile('negatives_lda.hdf5', __name__))
negatives = numpy.vstack((neg1, neg2)).T
positives = numpy.vstack((pos1, pos2)).T
T = CGLogRegTrainer(0.5, 1e-10, 10000, mean_std_norm=True)
# apply it to test data
test1 = [1., -50.]
test2 = [0.5, -86.]
res1 = machine(test1)
res2 = machine(test2)
# try the training without normalization
machine = T.train(negatives, positives)
# check that the results are at least approximately equal
# Note: lower values for epsilon and higher number of iterations improve the stability)
assert abs(machine(test1) - res1) < 1e-3
assert abs(machine(test2) - res2) < 1e-3
def test_netpbm():
transcode(test_utils.datafile('test.pbm', __name__)) # indexed, works fine
transcode(test_utils.datafile('test.pgm', __name__)) # indexed, works fine
transcode(test_utils.datafile('test.ppm', __name__)) # indexed, works fine
transcode(test_utils.datafile('test_2.pgm',
__name__)) # indexed, works fine
transcode(test_utils.datafile('test_2.ppm',
__name__)) # indexed, works fine
def loadGMM():
gmm = GMMMachine(2, 2)
gmm.weights = bob.io.base.load(datafile('gmm.init_weights.hdf5', __name__, path="../data/"))
gmm.means = bob.io.base.load(datafile('gmm.init_means.hdf5', __name__, path="../data/"))
gmm.variances = bob.io.base.load(datafile('gmm.init_variances.hdf5', __name__, path="../data/"))
#gmm.variance_thresholds = numpy.array([[0.001, 0.001],[0.001, 0.001]], 'float64')
return gmm
def loadGMM():
gmm = GMMMachine(2, 2)
gmm.weights = bob.io.base.load(
datafile('gmm.init_weights.hdf5', __name__, path="../data/"))
gmm.means = bob.io.base.load(
datafile('gmm.init_means.hdf5', __name__, path="../data/"))
gmm.variances = bob.io.base.load(
datafile('gmm.init_variances.hdf5', __name__, path="../data/"))
#gmm.variance_thresholds = numpy.array([[0.001, 0.001],[0.001, 0.001]], 'float64')
return gmm
def test_GMMMachine_3():
# Test a GMMMachine (log-likelihood computation)
data = bob.io.base.load(datafile('data.hdf5', __name__, path="../data/"))
gmm = GMMMachine(2, 50)
gmm.weights = bob.io.base.load(datafile('weights.hdf5', __name__, path="../data/"))
gmm.means = bob.io.base.load(datafile('means.hdf5', __name__, path="../data/"))
gmm.variances = bob.io.base.load(datafile('variances.hdf5', __name__, path="../data/"))
# Compare the log-likelihood with the one obtained using Chris Matlab
# implementation
matlab_ll_ref = -2.361583051672024e+02
assert abs(gmm(data) - matlab_ll_ref) < 1e-10
def test_kmeans_a():
# Trains a KMeansMachine
# This files contains draws from two 1D Gaussian distributions:
# * 100 samples from N(-10,1)
# * 100 samples from N(10,1)
data = bob.io.base.load(datafile("samplesFrom2G_f64.hdf5", __name__, path="../data/"))
machine = KMeansMachine(2, 1)
trainer = KMeansTrainer()
#trainer.train(machine, data)
bob.learn.em.train(trainer,machine,data)
[variances, weights] = machine.get_variances_and_weights_for_each_cluster(data)
variances_b = numpy.ndarray(shape=(2,1), dtype=numpy.float64)
weights_b = numpy.ndarray(shape=(2,), dtype=numpy.float64)
machine.__get_variances_and_weights_for_each_cluster_init__(variances_b, weights_b)
machine.__get_variances_and_weights_for_each_cluster_acc__(data, variances_b, weights_b)
machine.__get_variances_and_weights_for_each_cluster_fin__(variances_b, weights_b)
m1 = machine.get_mean(0)
m2 = machine.get_mean(1)
## Check means [-10,10] / variances [1,1] / weights [0.5,0.5]
if(m1<m2): means=numpy.array(([m1[0],m2[0]]), 'float64')
else: means=numpy.array(([m2[0],m1[0]]), 'float64')
assert equals(means, numpy.array([-10.,10.]), 2e-1)
assert equals(variances, numpy.array([1.,1.]), 2e-1)
assert equals(weights, numpy.array([0.5,0.5]), 1e-3)
assert equals(variances, variances_b, 1e-8)
assert equals(weights, weights_b, 1e-8)
def test_GMMMachine_3():
# Test a GMMMachine (log-likelihood computation)
data = bob.io.base.load(datafile('data.hdf5', __name__, path="../data/"))
gmm = GMMMachine(2, 50)
gmm.weights = bob.io.base.load(
datafile('weights.hdf5', __name__, path="../data/"))
gmm.means = bob.io.base.load(
datafile('means.hdf5', __name__, path="../data/"))
gmm.variances = bob.io.base.load(
datafile('variances.hdf5', __name__, path="../data/"))
# Compare the log-likelihood with the one obtained using Chris Matlab
# implementation
matlab_ll_ref = -2.361583051672024e+02
assert abs(gmm(data) - matlab_ll_ref) < 1e-10
def wrapper(*args, **kwargs):
dbfile = datafile("db.sql3", __name__, None)
if os.path.exists(dbfile):
return test(*args, **kwargs)
else:
raise SkipTest("The database file '%s' is not available; did you forget to run 'bob_dbmanage.py %s create' ?" % (
dbfile, 'replaymobile'))
def test_get_skin_pixels():
"""
Test the skin colored pixels detection
"""
# to run face detection
import bob.ip.facedetect
mod = sys.modules.get(__name__) or loader.load_module(__name__)
# load face image
face = load(datafile('001.jpg', 'bob.rppg.base'))
from bob.rppg.ssr.ssr_utils import get_skin_pixels
# zero threshold -> the number of skin pixels is the number of pixels in the cropped face
skin_pixels = get_skin_pixels(face, 0, True, 0.0)
bbox, quality = bob.ip.facedetect.detect_single_face(face)
assert skin_pixels.shape[1] == (bbox.size[0] - 1) * bbox.size[1] # -1 because of the cropping
# same as before, but with provided bbox
bounding_boxes = [bbox]
skin_pixels = get_skin_pixels(face, 0, True, 0, bounding_boxes)
assert skin_pixels.shape[1] == (bbox.size[0] - 1) * bbox.size[1] # -1 because of the cropping
# threshold of 1.0 -> zero skin pixels
skin_pixels = get_skin_pixels(face, 0, True, 1, bounding_boxes)
assert skin_pixels.shape[1] == 0
def test_kmeans_a():
# Trains a KMeansMachine
# This files contains draws from two 1D Gaussian distributions:
# * 100 samples from N(-10,1)
# * 100 samples from N(10,1)
data = bob.io.base.load(datafile("samplesFrom2G_f64.hdf5", __name__, path="../data/"))
machine = KMeansMachine(2, 1)
trainer = KMeansTrainer()
# trainer.train(machine, data)
bob.learn.em.train(trainer, machine, data)
[variances, weights] = machine.get_variances_and_weights_for_each_cluster(data)
variances_b = numpy.ndarray(shape=(2, 1), dtype=numpy.float64)
weights_b = numpy.ndarray(shape=(2,), dtype=numpy.float64)
machine.__get_variances_and_weights_for_each_cluster_init__(variances_b, weights_b)
machine.__get_variances_and_weights_for_each_cluster_acc__(data, variances_b, weights_b)
machine.__get_variances_and_weights_for_each_cluster_fin__(variances_b, weights_b)
m1 = machine.get_mean(0)
m2 = machine.get_mean(1)
## Check means [-10,10] / variances [1,1] / weights [0.5,0.5]
if (m1 < m2):
means = numpy.array(([m1[0], m2[0]]), 'float64')
else:
means = numpy.array(([m2[0], m1[0]]), 'float64')
assert equals(means, numpy.array([-10., 10.]), 2e-1)
assert equals(variances, numpy.array([1., 1.]), 2e-1)
assert equals(weights, numpy.array([0.5, 0.5]), 1e-3)
assert equals(variances, variances_b, 1e-8)
assert equals(weights, weights_b, 1e-8)
def test_cglogreg_norm_keyword():
# read some real test data;
# for toy examples the results are quite different...
pos1 = bob.io.base.load(datafile('positives_isv.hdf5', __name__))
neg1 = bob.io.base.load(datafile('negatives_isv.hdf5', __name__))
pos2 = bob.io.base.load(datafile('positives_lda.hdf5', __name__))
neg2 = bob.io.base.load(datafile('negatives_lda.hdf5', __name__))
negatives = numpy.vstack((neg1, neg2)).T
positives = numpy.vstack((pos1, pos2)).T
# Train the machine after mean-std norm
T = CGLogRegTrainer(0.5, 1e-10, 10000, reg=0.0001, mean_std_norm=True)
machine = T.train(negatives, positives)
# assert that mean and variance are correct
mean = numpy.mean(numpy.vstack((positives, negatives)), 0)
std = numpy.std(numpy.vstack((positives, negatives)), 0)
assert (abs(machine.input_subtract - mean) < 1e-10).all()
assert (abs(machine.input_divide - std) < 1e-10).all()
# apply it to test data
test1 = [1., -50.]
test2 = [0.5, -86.]
res1 = machine(test1)
res2 = machine(test2)
# normalize training data
pos = numpy.vstack([(positives[i] - mean) / std
for i in range(len(positives))])
neg = numpy.vstack([(negatives[i] - mean) / std
for i in range(len(negatives))])
# re-train the machine; should give identical results
T.mean_std_norm = False
machine = T.train(neg, pos)
machine.input_subtract = mean
machine.input_divide = std
# assert that the result is the same
assert abs(machine(test1) - res1) < 1e-10
assert abs(machine(test2) - res2) < 1e-10
def test_gmm_MAP_1():
# Train a GMMMachine with MAP_GMMTrainer
ar = bob.io.base.load(datafile('faithful.torch3_f64.hdf5', __name__, path="../data/"))
gmm = GMMMachine(bob.io.base.HDF5File(datafile("gmm_ML.hdf5", __name__, path="../data/")))
gmmprior = GMMMachine(bob.io.base.HDF5File(datafile("gmm_ML.hdf5", __name__, path="../data/")))
map_gmmtrainer = MAP_GMMTrainer(update_means=True, update_variances=False, update_weights=False, prior_gmm=gmmprior, relevance_factor=4.)
#map_gmmtrainer.train(gmm, ar)
bob.learn.em.train(map_gmmtrainer, gmm, ar)
gmm_ref = GMMMachine(bob.io.base.HDF5File(datafile('gmm_MAP.hdf5', __name__, path="../data/")))
assert (equals(gmm.means,gmm_ref.means,1e-3) and equals(gmm.variances,gmm_ref.variances,1e-3) and equals(gmm.weights,gmm_ref.weights,1e-3))
def wrapper(*args, **kwargs):
dbfile = datafile("db.sql3", __name__, None)
if os.path.exists(dbfile):
return test(*args, **kwargs)
else:
raise SkipTest(
"The database file '%s' is not available; did you forget to run 'bob_dbmanage.py %s create' ?" % (
dbfile, 'avspoof'))
def wrapper(*args, **kwargs):
dbfile = datafile("db.sql3", __name__, None)
if os.path.exists(dbfile):
return test(*args, **kwargs)
else:
raise SkipTest(
"The interface SQL file (%s) is not available; did you forget to run 'bob_dbmanage.py %s create' ?"
% (dbfile, 'vera'))
def test_cglogreg_norm_keyword():
# read some real test data;
# for toy examples the results are quite different...
pos1 = bob.io.base.load(datafile('positives_isv.hdf5', __name__))
neg1 = bob.io.base.load(datafile('negatives_isv.hdf5', __name__))
pos2 = bob.io.base.load(datafile('positives_lda.hdf5', __name__))
neg2 = bob.io.base.load(datafile('negatives_lda.hdf5', __name__))
negatives = numpy.vstack((neg1, neg2)).T
positives = numpy.vstack((pos1, pos2)).T
# Train the machine after mean-std norm
T = CGLogRegTrainer(0.5, 1e-10, 10000, reg=0.0001, mean_std_norm=True)
machine = T.train(negatives,positives)
# assert that mean and variance are correct
mean = numpy.mean(numpy.vstack((positives, negatives)), 0)
std = numpy.std(numpy.vstack((positives, negatives)), 0)
assert (abs(machine.input_subtract - mean) < 1e-10).all()
assert (abs(machine.input_divide - std) < 1e-10).all()
# apply it to test data
test1 = [1., -50.]
test2 = [0.5, -86.]
res1 = machine(test1)
res2 = machine(test2)
# normalize training data
pos = numpy.vstack([(positives[i] - mean) / std for i in range(len(positives))])
neg = numpy.vstack([(negatives[i] - mean) / std for i in range(len(negatives))])
# re-train the machine; should give identical results
T.mean_std_norm = False
machine = T.train(neg, pos)
machine.input_subtract = mean
machine.input_divide = std
# assert that the result is the same
assert abs(machine(test1) - res1) < 1e-10
assert abs(machine(test2) - res2) < 1e-10
def test_GMMMachine_4():
import numpy
numpy.random.seed(3) # FIXING A SEED
data = numpy.random.rand(100,50) #Doesn't matter if it is ramdom. The average of 1D array (in python) MUST output the same result for the 2D array (in C++)
gmm = GMMMachine(2, 50)
gmm.weights = bob.io.base.load(datafile('weights.hdf5', __name__, path="../data/"))
gmm.means = bob.io.base.load(datafile('means.hdf5', __name__, path="../data/"))
gmm.variances = bob.io.base.load(datafile('variances.hdf5', __name__, path="../data/"))
ll = 0
for i in range(data.shape[0]):
ll += gmm(data[i,:])
ll /= data.shape[0]
assert ll==gmm(data)
def test_gmm_ML_1():
# Trains a GMMMachine with ML_GMMTrainer
ar = bob.io.base.load(datafile("faithful.torch3_f64.hdf5", __name__, path="../data/"))
gmm = loadGMM()
ml_gmmtrainer = ML_GMMTrainer(True, True, True)
#ml_gmmtrainer.train(gmm, ar)
bob.learn.em.train(ml_gmmtrainer, gmm, ar, convergence_threshold=0.001)
#config = bob.io.base.HDF5File(datafile('gmm_ML.hdf5", __name__), 'w')
#gmm.save(config)
gmm_ref = GMMMachine(bob.io.base.HDF5File(datafile('gmm_ML.hdf5', __name__, path="../data/")))
gmm_ref_32bit_debug = GMMMachine(bob.io.base.HDF5File(datafile('gmm_ML_32bit_debug.hdf5', __name__, path="../data/")))
gmm_ref_32bit_release = GMMMachine(bob.io.base.HDF5File(datafile('gmm_ML_32bit_release.hdf5', __name__, path="../data/")))
assert (gmm == gmm_ref) or (gmm == gmm_ref_32bit_release) or (gmm == gmm_ref_32bit_release)
def test_io():
raise SkipTest("TODO: Not fully implemented yet")
# Checks that the IO functionality of LBP works
test_file = datafile("LBP.hdf5", __name__)
temp_file = temporary_filename()
# create file
lbp1 = bob.ip.base.LBP(8, (2, 3),
elbp_type="transitional",
to_average=True,
add_average_bit=True)
lbp2 = bob.ip.base.LBP(16,
4.,
2.,
uniform=True,
rotation_invariant=True,
circular=True)
# re-generate the reference file, if wanted
f = bob.io.base.HDF5File(temp_file, 'w')
f.create_group("LBP1")
f.create_group("LBP2")
f.cd("/LBP1")
lbp1.save(f)
f.cd("/LBP2")
lbp2.save(f)
del f
# load the file again
f = bob.io.base.HDF5File(temp_file)
f.cd("/LBP1")
read1 = bob.ip.base.LBP(f)
f.cd("/LBP2")
read2 = bob.ip.base.LBP(f)
del f
# assert that the created and the read object are identical
assert lbp1 == read1
assert lbp2 == read2
# load the reference file
f = bob.io.base.HDF5File(test_file)
f.cd("/LBP1")
ref1 = bob.ip.base.LBP(f)
f.cd("/LBP2")
ref2 = bob.ip.base.LBP(f)
del f
# assert that the lbp objects and the reference ones are identical
assert lbp1 == ref1
assert lbp2 == ref2
assert read1 == ref1
assert read2 == ref2
def test_gmm_MAP_2():
# Train a GMMMachine with MAP_GMMTrainer and compare with matlab reference
data = bob.io.base.load(datafile('data.hdf5', __name__, path="../data/"))
data = data.reshape((1, data.shape[0])) # make a 2D array out of it
means = bob.io.base.load(datafile('means.hdf5', __name__, path="../data/"))
variances = bob.io.base.load(
datafile('variances.hdf5', __name__, path="../data/"))
weights = bob.io.base.load(
datafile('weights.hdf5', __name__, path="../data/"))
gmm = GMMMachine(2, 50)
gmm.means = means
gmm.variances = variances
gmm.weights = weights
map_adapt = MAP_GMMTrainer(update_means=True,
update_variances=False,
update_weights=False,
mean_var_update_responsibilities_threshold=0.,
prior_gmm=gmm,
relevance_factor=4.)
gmm_adapted = GMMMachine(2, 50)
gmm_adapted.means = means
gmm_adapted.variances = variances
gmm_adapted.weights = weights
#map_adapt.max_iterations = 1
#map_adapt.train(gmm_adapted, data)
bob.learn.em.train(map_adapt, gmm_adapted, data, max_iterations=1)
new_means = bob.io.base.load(
datafile('new_adapted_mean.hdf5', __name__, path="../data/"))
# print new_means[0,:]
# print gmm_adapted.means[:,0]
# Compare to matlab reference
assert equals(new_means[0, :], gmm_adapted.means[:, 0], 1e-4)
assert equals(new_means[1, :], gmm_adapted.means[:, 1], 1e-4)
def test_processing():
# Processing tests
A = bob.io.base.load(datafile("vlimg_ref.hdf5", 'bob.ip.base', 'data/sift'))
No = 3
Ns = 3
sigma0 = 1.6
sigma_n = 0.5
cont_t = 0.03
edge_t = 10.
norm_t = 0.2
f=4.
op = bob.ip.base.SIFT(A.shape,Ns,No,0,sigma_n,sigma0,cont_t,edge_t,norm_t,f,bob.sp.BorderType.NearestNeighbour)
kp=[bob.ip.base.GSSKeypoint(1.6,(326,270))]
B = numpy.ndarray(op.output_shape(1), numpy.float64)
op.compute_descriptor(A,kp,B)
C=B[0]
#bob.io.base.save(C, datafile(os.path.join("sift","vlimg_ref_cmp.hdf5"), __name__)) # Generated using initial bob version
C_ref = bob.io.base.load(datafile("vlimg_ref_cmp.hdf5", 'bob.ip.base', 'data/sift'))
assert numpy.allclose(C, C_ref, 1e-5, 1e-5)
"""
def test_ztnorm_big():
my_A = bob.io.base.load(datafile("ztnorm_eval_eval.hdf5", __name__, path="../data/"))
my_B = bob.io.base.load(datafile("ztnorm_znorm_eval.hdf5", __name__, path="../data/"))
my_C = bob.io.base.load(datafile("ztnorm_eval_tnorm.hdf5", __name__, path="../data/"))
my_D = bob.io.base.load(datafile("ztnorm_znorm_tnorm.hdf5", __name__, path="../data/"))
# ZT-Norm
ref_scores = bob.io.base.load(datafile("ztnorm_result.hdf5", __name__, path="../data/"))
scores = bob.learn.em.ztnorm(my_A, my_B, my_C, my_D)
assert (abs(scores - ref_scores) < 1e-7).all()
# T-Norm
scores = bob.learn.em.tnorm(my_A, my_C)
scores_py = tnorm(my_A, my_C)
assert (abs(scores - scores_py) < 1e-7).all()
# Z-Norm
scores = bob.learn.em.znorm(my_A, my_B)
scores_py = znorm(my_A, my_B)
assert (abs(scores - scores_py) < 1e-7).all()
def test_GMMMachine_2():
# Test a GMMMachine (statistics)
arrayset = bob.io.base.load(datafile("faithful.torch3_f64.hdf5", __name__, path="../data/"))
gmm = GMMMachine(2, 2)
gmm.weights = numpy.array([0.5, 0.5], 'float64')
gmm.means = numpy.array([[3, 70], [4, 72]], 'float64')
gmm.variances = numpy.array([[1, 10], [2, 5]], 'float64')
gmm.variance_thresholds = numpy.array([[0, 0], [0, 0]], 'float64')
stats = GMMStats(2, 2)
gmm.acc_statistics(arrayset, stats)
stats_ref = GMMStats(bob.io.base.HDF5File(datafile("stats.hdf5",__name__, path="../data/")))
assert stats.t == stats_ref.t
assert numpy.allclose(stats.n, stats_ref.n, atol=1e-10)
#assert numpy.array_equal(stats.sumPx, stats_ref.sumPx)
#Note AA: precision error above
assert numpy.allclose(stats.sum_px, stats_ref.sum_px, atol=1e-10)
assert numpy.allclose(stats.sum_pxx, stats_ref.sum_pxx, atol=1e-10)
def test_int32_histoPython():
# Compute the histogram of a int32 random array
# Generate random int32 array
#input_array = numpy.ndarray((50, 70), 'int32')
#random_int(input_array, -20,20)
#bob.io.base.save(input_array,os.path.join('histo','input_int32.hdf5'))
input_array = bob.io.base.load(datafile('input_int32.hdf5', 'bob.ip.base', 'data/histo'))
histo2 = numpy.ndarray((41,), numpy.uint64)
histo1 = bob.ip.base.histogram(input_array, (-20, 20), 41)
bob.ip.base.histogram(input_array, (-20, 20), histo2)
# Save computed data
#bob.io.base.save(histo, os.path.join('histo','input_int32.histo.hdf5'))
histo_ref = bob.io.base.load(datafile('input_int32.histo.hdf5', 'bob.ip.base', 'data/histo'))
assert input_array.size == histo1.sum()
assert input_array.size == histo2.sum()
assert (histo_ref == histo1).all()
assert (histo_ref == histo2).all()
def test_processing():
# Processing tests
A = bob.io.base.load(datafile("vlimg_ref.hdf5", "bob.ip.base",
"data/sift"))
No = 3
Ns = 3
sigma0 = 1.6
sigma_n = 0.5
f = 4.
op = bob.ip.base.GaussianScaleSpace(A.shape, Ns, No, 0, sigma_n, sigma0, f)
pyr = op(A)
import math
# Assumes that octave_min = 0
dsigma0 = sigma0 * math.sqrt(1. - math.pow(2, -2. / float(Ns)))
Aa = A
for o in range(No):
for s in range(-1, Ns + 2):
# Get Gaussian for this scale
g_pyr = op.get_gaussian(s + 1)
# Filtering step
if s != -1 or (o == 0 and s == -1):
# Compute scale and radius
if (o == 0 and s == -1):
sa = sigma0 #* math.pow(2.,s/float(Ns))
sb = sigma_n
sigma = math.sqrt(sa * sa - sb * sb)
else:
sigma = dsigma0 * math.pow(2, s / float(Ns))
radius = int(math.ceil(f * sigma))
# Check values
assert abs(sigma - g_pyr.sigma[0]) < eps
assert abs(sigma - g_pyr.sigma[1]) < eps
assert abs(radius - g_pyr.radius[0]) < eps
assert abs(radius - g_pyr.radius[1]) < eps
g = bob.ip.base.Gaussian((sigma, sigma), (radius, radius))
B = g(Aa)
# Downsampling step
else:
# Select image as by VLfeat (seems wrong to me)
Aa = pyr[o - 1][Ns, :, :]
# Downsample using a trick to make sure that if the length is l=2p+1,
# the new one is p and not p+1.
B = Aa[:2 * (int(Aa.shape[0] / 2)):2, :2 *
(int(Aa.shape[1] / 2)):2]
# Compare image of the pyramids (Python implementation vs. C++)
Bpyr = pyr[o][s + 1, :, :]
assert numpy.allclose(B, Bpyr, eps)
Aa = B
def test_matlab_baseline():
"""
Tests based on this matlab baseline
http://www-scf.usc.edu/~boqinggo/domainadaptation.html#intro
"""
import numpy
numpy.random.seed(10)
source_webcam = bob.io.matlab.read_matrix(datafile("webcam.mat", __name__))
webcam_labels = bob.io.matlab.read_matrix(datafile("webcam_labels.mat", __name__))
target_dslr = bob.io.matlab.read_matrix(datafile("dslr.mat", __name__))
dslr_labels = bob.io.matlab.read_matrix(datafile("dslr_labels.mat", __name__))
gfk_trainer = GFKTrainer(10, subspace_dim_source=140, subspace_dim_target=140)
gfk_machine = gfk_trainer.train(source_webcam, target_dslr)
accuracy = compute_accuracy(gfk_machine.G, source_webcam, webcam_labels, target_dslr, dslr_labels) * 100
assert accuracy > 70
def test_custom_trainer():
# Custom python trainer
ar = bob.io.base.load(datafile("faithful.torch3_f64.hdf5", __name__, path="../data/"))
mytrainer = MyTrainer1()
machine = KMeansMachine(2, 2)
mytrainer.train(machine, ar)
for i in range(0, 2):
assert (ar[i+1] == machine.means[i, :]).all()
def test_float_histoPython():
# Compute the histogram of a float random array
# Generate random float32 array
#input_array = numpy.ndarray((50, 70), 'float32')
#random_float(input_array, 0, 1)
#bob.io.base.save(input_array, os.path.join('histo','input_float.hdf5'))
input_array = bob.io.base.load(datafile('input_float.hdf5', 'bob.ip.base', 'data/histo'))
histo2 = numpy.ndarray((10,), numpy.uint64)
histo1 = bob.ip.base.histogram(input_array, (0, 1), 10)
bob.ip.base.histogram(input_array, (0, 1), histo2)
# Save computed data
#bob.io.base.save(histo1,os.path.join('histo','input_float.histo.hdf5'))
histo_ref = bob.io.base.load(datafile('input_float.histo.hdf5', 'bob.ip.base', 'data/histo'))
assert input_array.size == histo1.sum()
assert input_array.size == histo2.sum()
assert (histo_ref == histo1).all()
assert (histo_ref == histo2).all()
def test_kmeans_b():
# Trains a KMeansMachine
(arStd,std) = NormalizeStdArray(datafile("faithful.torch3.hdf5", __name__, path="../data/"))
machine = KMeansMachine(2, 2)
trainer = KMeansTrainer()
#trainer.seed = 1337
bob.learn.em.train(trainer,machine, arStd, convergence_threshold=0.001)
[variances, weights] = machine.get_variances_and_weights_for_each_cluster(arStd)
means = numpy.array(machine.means)
variances = numpy.array(variances)
multiplyVectorsByFactors(means, std)
multiplyVectorsByFactors(variances, std ** 2)
gmmWeights = bob.io.base.load(datafile('gmm.init_weights.hdf5', __name__, path="../data/"))
gmmMeans = bob.io.base.load(datafile('gmm.init_means.hdf5', __name__, path="../data/"))
gmmVariances = bob.io.base.load(datafile('gmm.init_variances.hdf5', __name__, path="../data/"))
if (means[0, 0] < means[1, 0]):
means = flipRows(means)
variances = flipRows(variances)
weights = flipRows(weights)
assert equals(means, gmmMeans, 1e-3)
assert equals(weights, gmmWeights, 1e-3)
assert equals(variances, gmmVariances, 1e-3)
# Check that there is no duplicate means during initialization
machine = KMeansMachine(2, 1)
trainer = KMeansTrainer()
trainer.initialization_method = 'RANDOM_NO_DUPLICATE'
data = numpy.array([[1.], [1.], [1.], [1.], [1.], [1.], [2.], [3.]])
bob.learn.em.train(trainer, machine, data)
assert (numpy.isnan(machine.means).any()) == False
def test_kmeans_b():
# Trains a KMeansMachine
(arStd, std) = NormalizeStdArray(datafile("faithful.torch3.hdf5", __name__, path="../data/"))
machine = KMeansMachine(2, 2)
trainer = KMeansTrainer()
# trainer.seed = 1337
bob.learn.em.train(trainer, machine, arStd, convergence_threshold=0.001)
[variances, weights] = machine.get_variances_and_weights_for_each_cluster(arStd)
means = numpy.array(machine.means)
variances = numpy.array(variances)
multiplyVectorsByFactors(means, std)
multiplyVectorsByFactors(variances, std ** 2)
gmmWeights = bob.io.base.load(datafile('gmm.init_weights.hdf5', __name__, path="../data/"))
gmmMeans = bob.io.base.load(datafile('gmm.init_means.hdf5', __name__, path="../data/"))
gmmVariances = bob.io.base.load(datafile('gmm.init_variances.hdf5', __name__, path="../data/"))
if (means[0, 0] < means[1, 0]):
means = flipRows(means)
variances = flipRows(variances)
weights = flipRows(weights)
assert equals(means, gmmMeans, 1e-3)
assert equals(weights, gmmWeights, 1e-3)
assert equals(variances, gmmVariances, 1e-3)
# Check that there is no duplicate means during initialization
machine = KMeansMachine(2, 1)
trainer = KMeansTrainer()
trainer.initialization_method = 'RANDOM_NO_DUPLICATE'
data = numpy.array([[1.], [1.], [1.], [1.], [1.], [1.], [2.], [3.]])
bob.learn.em.train(trainer, machine, data)
assert (numpy.isnan(machine.means).any()) == False
def test_uint16_histoPython():
# Compute the histogram of a uint16 random array
# Generate random uint16 array
#input_array = numpy.ndarray((50, 70), 'uint16')
#random_int(input_array, 0, 65535)
#bob.io.base.save(input_array, os.path.join('histo','input_uint16.hdf5'))
input_array = bob.io.base.load(
datafile('input_uint16.hdf5', 'bob.ip.base', 'data/histo'))
histo1 = bob.ip.base.histogram(input_array)
histo2 = bob.ip.base.histogram(input_array, 65536)
histo3 = bob.ip.base.histogram(input_array, (0, 65535), 65536)
histo4 = numpy.ndarray((65536, ), numpy.uint64)
histo5 = numpy.ndarray((65536, ), numpy.uint64)
bob.ip.base.histogram(input_array, histo4)
bob.ip.base.histogram(input_array, (0, 65535), histo5)
# Save computed data
#bob.io.base.save(histo1, os.path.join('histo','input_uint16.histo.hdf5'))
histo_ref = bob.io.base.load(
datafile('input_uint16.histo.hdf5', 'bob.ip.base', 'data/histo'))
assert input_array.size == histo1.sum()
assert input_array.size == histo2.sum()
assert input_array.size == histo3.sum()
assert input_array.size == histo4.sum()
assert input_array.size == histo5.sum()
assert (histo_ref == histo1).all()
assert (histo_ref == histo2).all()
assert (histo_ref == histo3).all()
assert (histo_ref == histo4).all()
assert (histo_ref == histo5).all()
def test_processing():
# Processing tests
A = bob.io.base.load(datafile("vlimg_ref.hdf5", "bob.ip.base", "data/sift"))
No = 3
Ns = 3
sigma0 = 1.6
sigma_n = 0.5
f=4.
op = bob.ip.base.GaussianScaleSpace(A.shape,Ns,No,0,sigma_n,sigma0,f)
pyr = op(A)
import math
# Assumes that octave_min = 0
dsigma0 = sigma0 * math.sqrt(1.-math.pow(2,-2./float(Ns)))
Aa = A
for o in range(No):
for s in range(-1,Ns+2):
# Get Gaussian for this scale
g_pyr = op.get_gaussian(s+1)
# Filtering step
if s!=-1 or (o==0 and s==-1):
# Compute scale and radius
if(o==0 and s==-1):
sa = sigma0 #* math.pow(2.,s/float(Ns))
sb = sigma_n
sigma = math.sqrt(sa*sa - sb*sb)
else:
sigma = dsigma0 * math.pow(2,s/float(Ns))
radius = int(math.ceil(f*sigma))
# Check values
assert abs(sigma - g_pyr.sigma[0]) < eps
assert abs(sigma - g_pyr.sigma[1]) < eps
assert abs(radius - g_pyr.radius[0]) < eps
assert abs(radius - g_pyr.radius[1]) < eps
g = bob.ip.base.Gaussian((sigma, sigma), (radius, radius))
B = g(Aa)
# Downsampling step
else:
# Select image as by VLfeat (seems wrong to me)
Aa = pyr[o-1][Ns,:,:]
# Downsample using a trick to make sure that if the length is l=2p+1,
# the new one is p and not p+1.
B = Aa[:2*(int(Aa.shape[0]/2)):2,:2*(int(Aa.shape[1]/2)):2]
# Compare image of the pyramids (Python implementation vs. C++)
Bpyr = pyr[o][s+1,:,:]
assert numpy.allclose(B, Bpyr, eps)
Aa = B
def test_int32_histoPython():
# Compute the histogram of a int32 random array
# Generate random int32 array
#input_array = numpy.ndarray((50, 70), 'int32')
#random_int(input_array, -20,20)
#bob.io.base.save(input_array,os.path.join('histo','input_int32.hdf5'))
input_array = bob.io.base.load(
datafile('input_int32.hdf5', 'bob.ip.base', 'data/histo'))
histo2 = numpy.ndarray((41, ), numpy.uint64)
histo1 = bob.ip.base.histogram(input_array, (-20, 20), 41)
bob.ip.base.histogram(input_array, (-20, 20), histo2)
# Save computed data
#bob.io.base.save(histo, os.path.join('histo','input_int32.histo.hdf5'))
histo_ref = bob.io.base.load(
datafile('input_int32.histo.hdf5', 'bob.ip.base', 'data/histo'))
assert input_array.size == histo1.sum()
assert input_array.size == histo2.sum()
assert (histo_ref == histo1).all()
assert (histo_ref == histo2).all()
def test_custom_trainer():
# Custom python trainer
ar = bob.io.base.load(
datafile("faithful.torch3_f64.hdf5", __name__, path="../data/"))
mytrainer = MyTrainer1()
machine = KMeansMachine(2, 2)
mytrainer.train(machine, ar)
for i in range(0, 2):
assert (ar[i + 1] == machine.means[i, :]).all()
def test_GMMMachine_4():
import numpy
numpy.random.seed(3) # FIXING A SEED
data = numpy.random.rand(
100, 50
) #Doesn't matter if it is ramdom. The average of 1D array (in python) MUST output the same result for the 2D array (in C++)
gmm = GMMMachine(2, 50)
gmm.weights = bob.io.base.load(
datafile('weights.hdf5', __name__, path="../data/"))
gmm.means = bob.io.base.load(
datafile('means.hdf5', __name__, path="../data/"))
gmm.variances = bob.io.base.load(
datafile('variances.hdf5', __name__, path="../data/"))
ll = 0
for i in range(data.shape[0]):
ll += gmm(data[i, :])
ll /= data.shape[0]
assert ll == gmm(data)
def test_processing():
# Processing tests
A = bob.io.base.load(datafile("vlimg_ref.hdf5", 'bob.ip.base',
'data/sift'))
No = 3
Ns = 3
sigma0 = 1.6
sigma_n = 0.5
cont_t = 0.03
edge_t = 10.
norm_t = 0.2
f = 4.
op = bob.ip.base.SIFT(A.shape, Ns, No, 0, sigma_n, sigma0, cont_t, edge_t,
norm_t, f, bob.sp.BorderType.NearestNeighbour)
kp = [bob.ip.base.GSSKeypoint(1.6, (326, 270))]
B = numpy.ndarray(op.output_shape(1), numpy.float64)
op.compute_descriptor(A, kp, B)
C = B[0]
#bob.io.base.save(C, datafile(os.path.join("sift","vlimg_ref_cmp.hdf5"), __name__)) # Generated using initial bob version
C_ref = bob.io.base.load(
datafile("vlimg_ref_cmp.hdf5", 'bob.ip.base', 'data/sift'))
assert numpy.allclose(C, C_ref, 1e-5, 1e-5)
"""
def test_gmm_ML_1():
# Trains a GMMMachine with ML_GMMTrainer
ar = bob.io.base.load(
datafile("faithful.torch3_f64.hdf5", __name__, path="../data/"))
gmm = loadGMM()
# test rng handling
ml_gmmtrainer = ML_GMMTrainer(True, True, True)
rng = bob.core.random.mt19937(12345)
bob.learn.em.train(ml_gmmtrainer,
gmm,
ar,
convergence_threshold=0.001,
rng=rng)
gmm = loadGMM()
ml_gmmtrainer = ML_GMMTrainer(True, True, True)
#ml_gmmtrainer.train(gmm, ar)
bob.learn.em.train(ml_gmmtrainer, gmm, ar, convergence_threshold=0.001)
#config = bob.io.base.HDF5File(datafile('gmm_ML.hdf5", __name__), 'w')
#gmm.save(config)
gmm_ref = GMMMachine(
bob.io.base.HDF5File(datafile('gmm_ML.hdf5', __name__,
path="../data/")))
gmm_ref_32bit_debug = GMMMachine(
bob.io.base.HDF5File(
datafile('gmm_ML_32bit_debug.hdf5', __name__, path="../data/")))
gmm_ref_32bit_release = GMMMachine(
bob.io.base.HDF5File(
datafile('gmm_ML_32bit_release.hdf5', __name__, path="../data/")))
assert (gmm == gmm_ref) or (gmm == gmm_ref_32bit_release) or (
gmm == gmm_ref_32bit_release)
def test_uint16_histoPython():
# Compute the histogram of a uint16 random array
# Generate random uint16 array
#input_array = numpy.ndarray((50, 70), 'uint16')
#random_int(input_array, 0, 65535)
#bob.io.base.save(input_array, os.path.join('histo','input_uint16.hdf5'))
input_array = bob.io.base.load(datafile('input_uint16.hdf5', 'bob.ip.base', 'data/histo'))
histo1 = bob.ip.base.histogram(input_array)
histo2 = bob.ip.base.histogram(input_array, 65536)
histo3 = bob.ip.base.histogram(input_array, (0, 65535), 65536)
histo4 = numpy.ndarray((65536,), numpy.uint64)
histo5 = numpy.ndarray((65536,), numpy.uint64)
bob.ip.base.histogram(input_array, histo4)
bob.ip.base.histogram(input_array, (0, 65535), histo5)
# Save computed data
#bob.io.base.save(histo1, os.path.join('histo','input_uint16.histo.hdf5'))
histo_ref = bob.io.base.load(datafile('input_uint16.histo.hdf5', 'bob.ip.base', 'data/histo'))
assert input_array.size == histo1.sum()
assert input_array.size == histo2.sum()
assert input_array.size == histo3.sum()
assert input_array.size == histo4.sum()
assert input_array.size == histo5.sum()
assert (histo_ref == histo1).all()
assert (histo_ref == histo2).all()
assert (histo_ref == histo3).all()
assert (histo_ref == histo4).all()
assert (histo_ref == histo5).all()
def test_GMMMachine_2():
# Test a GMMMachine (statistics)
arrayset = bob.io.base.load(
datafile("faithful.torch3_f64.hdf5", __name__, path="../data/"))
gmm = GMMMachine(2, 2)
gmm.weights = numpy.array([0.5, 0.5], 'float64')
gmm.means = numpy.array([[3, 70], [4, 72]], 'float64')
gmm.variances = numpy.array([[1, 10], [2, 5]], 'float64')
gmm.variance_thresholds = numpy.array([[0, 0], [0, 0]], 'float64')
stats = GMMStats(2, 2)
gmm.acc_statistics(arrayset, stats)
stats_ref = GMMStats(
bob.io.base.HDF5File(datafile("stats.hdf5", __name__,
path="../data/")))
assert stats.t == stats_ref.t
assert numpy.allclose(stats.n, stats_ref.n, atol=1e-10)
#assert numpy.array_equal(stats.sumPx, stats_ref.sumPx)
#Note AA: precision error above
assert numpy.allclose(stats.sum_px, stats_ref.sum_px, atol=1e-10)
assert numpy.allclose(stats.sum_pxx, stats_ref.sum_pxx, atol=1e-10)
def test_float_histoPython():
# Compute the histogram of a float random array
# Generate random float32 array
#input_array = numpy.ndarray((50, 70), 'float32')
#random_float(input_array, 0, 1)
#bob.io.base.save(input_array, os.path.join('histo','input_float.hdf5'))
input_array = bob.io.base.load(
datafile('input_float.hdf5', 'bob.ip.base', 'data/histo'))
histo2 = numpy.ndarray((10, ), numpy.uint64)
histo1 = bob.ip.base.histogram(input_array, (0, 1), 10)
bob.ip.base.histogram(input_array, (0, 1), histo2)
# Save computed data
#bob.io.base.save(histo1,os.path.join('histo','input_float.histo.hdf5'))
histo_ref = bob.io.base.load(
datafile('input_float.histo.hdf5', 'bob.ip.base', 'data/histo'))
assert input_array.size == histo1.sum()
assert input_array.size == histo2.sum()
assert (histo_ref == histo1).all()
assert (histo_ref == histo2).all()
def test_all():
# array writing tests
a1 = numpy.random.normal(size=(2, 3)).astype('float32')
a2 = numpy.random.normal(size=(2, 3, 4)).astype('float64')
a3 = numpy.random.normal(size=(2, 3, 4, 5)).astype('complex128')
a4 = (10 * numpy.random.normal(size=(3, 3))).astype('uint64')
array_readwrite('.mat', a1)
array_readwrite(".mat", a2)
array_readwrite('.mat', a3)
array_readwrite(".mat", a4)
# arrayset writing tests
a1 = []
a2 = []
a3 = []
a4 = []
for k in range(10):
a1.append(numpy.random.normal(size=(2, 3)).astype('float32'))
a2.append(numpy.random.normal(size=(2, 3, 4)).astype('float64'))
a3.append(numpy.random.normal(size=(2, 3, 4, 5)).astype('complex128'))
a4.append((10 * numpy.random.normal(size=(3, 3))).astype('uint64'))
arrayset_readwrite('.mat', a1)
arrayset_readwrite(".mat", a2)
arrayset_readwrite('.mat', a3)
arrayset_readwrite(".mat", a4)
# complete transcoding tests
transcode(test_utils.datafile(
'test_1d.mat', __name__)) #pseudo 1D - matlab does not support true 1D
transcode(test_utils.datafile('test_2d.mat', __name__))
transcode(test_utils.datafile('test_3d.mat', __name__))
transcode(test_utils.datafile('test_4d.mat', __name__))
transcode(test_utils.datafile(
'test_1d_cplx.mat', __name__)) #pseudo 1D - matlab does not support 1D
transcode(test_utils.datafile('test_2d_cplx.mat', __name__))
transcode(test_utils.datafile('test_3d_cplx.mat', __name__))
transcode(test_utils.datafile('test_4d_cplx.mat', __name__))
transcode(test_utils.datafile('test.mat', __name__)) #3D complex, large
def test_interface():
# test that we can read the 'x' variable in the test file
cell_file = test_utils.datafile('test_2d.mat', __name__)
sorted(['x', 'y']) == sorted(read_varnames(cell_file))
# read x matrix
x = read_matrix(cell_file, 'x')
assert x.shape == (2,3)
y = read_matrix(cell_file, 'y')
assert y.shape == (3,2)
for i in range(2):
for j in range(3):
assert x[i,j] == float(j*2+i+1)
assert y[j,i] == float(j*2+i+1)
def test_gmm_ML_2():
# Trains a GMMMachine with ML_GMMTrainer; compares to an old reference
ar = bob.io.base.load(
datafile('dataNormalized.hdf5', __name__, path="../data/"))
# Initialize GMMMachine
gmm = GMMMachine(5, 45)
gmm.means = bob.io.base.load(
datafile('meansAfterKMeans.hdf5', __name__,
path="../data/")).astype('float64')
gmm.variances = bob.io.base.load(
datafile('variancesAfterKMeans.hdf5', __name__,
path="../data/")).astype('float64')
gmm.weights = numpy.exp(
bob.io.base.load(
datafile('weightsAfterKMeans.hdf5', __name__,
path="../data/")).astype('float64'))
threshold = 0.001
gmm.set_variance_thresholds(threshold)
# Initialize ML Trainer
prior = 0.001
max_iter_gmm = 25
accuracy = 0.00001
ml_gmmtrainer = ML_GMMTrainer(True, True, True, prior)
# Run ML
#ml_gmmtrainer.train(gmm, ar)
bob.learn.em.train(ml_gmmtrainer,
gmm,
ar,
max_iterations=max_iter_gmm,
convergence_threshold=accuracy)
# Test results
# Load torch3vision reference
meansML_ref = bob.io.base.load(
datafile('meansAfterML.hdf5', __name__, path="../data/"))
variancesML_ref = bob.io.base.load(
datafile('variancesAfterML.hdf5', __name__, path="../data/"))
weightsML_ref = bob.io.base.load(
datafile('weightsAfterML.hdf5', __name__, path="../data/"))
# Compare to current results
assert equals(gmm.means, meansML_ref, 3e-3)
assert equals(gmm.variances, variancesML_ref, 3e-3)
assert equals(gmm.weights, weightsML_ref, 1e-4)
def test_interface():
# test that we can read the 'x' variable in the test file
cell_file = test_utils.datafile('test_2d.mat', __name__)
sorted(['x', 'y']) == sorted(read_varnames(cell_file))
# read x matrix
x = read_matrix(cell_file, 'x')
assert x.shape == (2, 3)
y = read_matrix(cell_file, 'y')
assert y.shape == (3, 2)
for i in range(2):
for j in range(3):
assert x[i, j] == float(j * 2 + i + 1)
assert y[j, i] == float(j * 2 + i + 1)
def test_io():
raise SkipTest("TODO: Not fully implemented yet")
# Checks that the IO functionality of LBP works
test_file = datafile("LBP.hdf5", __name__)
temp_file = temporary_filename()
# create file
lbp1 = bob.ip.base.LBP(8, (2,3), elbp_type="transitional", to_average=True, add_average_bit=True)
lbp2 = bob.ip.base.LBP(16, 4., 2., uniform=True, rotation_invariant=True, circular=True)
# re-generate the reference file, if wanted
f = bob.io.base.HDF5File(temp_file, 'w')
f.create_group("LBP1")
f.create_group("LBP2")
f.cd("/LBP1")
lbp1.save(f)
f.cd("/LBP2")
lbp2.save(f)
del f
# load the file again
f = bob.io.base.HDF5File(temp_file)
f.cd("/LBP1")
read1 = bob.ip.base.LBP(f)
f.cd("/LBP2")
read2 = bob.ip.base.LBP(f)
del f
# assert that the created and the read object are identical
assert lbp1 == read1
assert lbp2 == read2
# load the reference file
f = bob.io.base.HDF5File(test_file)
f.cd("/LBP1")
ref1 = bob.ip.base.LBP(f)
f.cd("/LBP2")
ref2 = bob.ip.base.LBP(f)
del f
# assert that the lbp objects and the reference ones are identical
assert lbp1 == ref1
assert lbp2 == ref2
assert read1 == ref1
assert read2 == ref2
def test_all():
# array writing tests
a1 = numpy.random.normal(size=(2,3)).astype('float32')
a2 = numpy.random.normal(size=(2,3,4)).astype('float64')
a3 = numpy.random.normal(size=(2,3,4,5)).astype('complex128')
a4 = (10 * numpy.random.normal(size=(3,3))).astype('uint64')
array_readwrite('.mat', a1)
array_readwrite(".mat", a2)
array_readwrite('.mat', a3)
array_readwrite(".mat", a4)
# arrayset writing tests
a1 = []
a2 = []
a3 = []
a4 = []
for k in range(10):
a1.append(numpy.random.normal(size=(2,3)).astype('float32'))
a2.append(numpy.random.normal(size=(2,3,4)).astype('float64'))
a3.append(numpy.random.normal(size=(2,3,4,5)).astype('complex128'))
a4.append((10*numpy.random.normal(size=(3,3))).astype('uint64'))
arrayset_readwrite('.mat', a1)
arrayset_readwrite(".mat", a2)
arrayset_readwrite('.mat', a3)
arrayset_readwrite(".mat", a4)
# complete transcoding tests
transcode(test_utils.datafile('test_1d.mat', __name__)) #pseudo 1D - matlab does not support true 1D
transcode(test_utils.datafile('test_2d.mat', __name__))
transcode(test_utils.datafile('test_3d.mat', __name__))
transcode(test_utils.datafile('test_4d.mat', __name__))
transcode(test_utils.datafile('test_1d_cplx.mat', __name__)) #pseudo 1D - matlab does not support 1D
transcode(test_utils.datafile('test_2d_cplx.mat', __name__))
transcode(test_utils.datafile('test_3d_cplx.mat', __name__))
transcode(test_utils.datafile('test_4d_cplx.mat', __name__))
transcode(test_utils.datafile('test.mat', __name__)) #3D complex, large
def test_image_load():
# test that the generic bob.io.image.load function works as expected
for filename in ('test.jpg', 'cmyk.jpg', 'test.pbm', 'test_corrupted.pbm',
'test.pgm', 'test_corrupted.pgm', 'test_spaces.pgm', 'test.ppm',
'test_corrupted.ppm', 'img_rgba_color.png', 'test.gif'):
full_file = test_utils.datafile(filename, __name__)
# load with just image name
i1 = bob.io.image.load(full_file)
# load with image name and extension
i2 = bob.io.image.load(full_file, os.path.splitext(full_file)[1])
assert numpy.array_equal(i1,i2)
# load with image name and automatically estimated extension
i3 = bob.io.image.load(full_file, 'auto')
assert numpy.array_equal(i1,i3)
# assert that unknown extensions raise exceptions
nose.tools.assert_raises(RuntimeError, lambda x: bob.io.image.load(x, ".unknown"), full_file)
def test_gmm_MAP_3():
# Train a GMMMachine with MAP_GMMTrainer; compares to old reference
ar = bob.io.base.load(datafile('dataforMAP.hdf5', __name__, path="../data/"))
# Initialize GMMMachine
n_gaussians = 5
n_inputs = 45
prior_gmm = GMMMachine(n_gaussians, n_inputs)
prior_gmm.means = bob.io.base.load(datafile('meansAfterML.hdf5', __name__, path="../data/"))
prior_gmm.variances = bob.io.base.load(datafile('variancesAfterML.hdf5', __name__, path="../data/"))
prior_gmm.weights = bob.io.base.load(datafile('weightsAfterML.hdf5', __name__, path="../data/"))
threshold = 0.001
prior_gmm.set_variance_thresholds(threshold)
# Initialize MAP Trainer
relevance_factor = 0.1
prior = 0.001
max_iter_gmm = 1
accuracy = 0.00001
map_factor = 0.5
map_gmmtrainer = MAP_GMMTrainer(prior_gmm, alpha=map_factor, update_means=True, update_variances=False, update_weights=False, mean_var_update_responsibilities_threshold=accuracy)
#map_gmmtrainer.max_iterations = max_iter_gmm
#map_gmmtrainer.convergence_threshold = accuracy
gmm = GMMMachine(n_gaussians, n_inputs)
gmm.set_variance_thresholds(threshold)
# Train
#map_gmmtrainer.train(gmm, ar)
bob.learn.em.train(map_gmmtrainer, gmm, ar, max_iterations = max_iter_gmm, convergence_threshold=prior)
# Test results
# Load torch3vision reference
meansMAP_ref = bob.io.base.load(datafile('meansAfterMAP.hdf5', __name__, path="../data/"))
variancesMAP_ref = bob.io.base.load(datafile('variancesAfterMAP.hdf5', __name__, path="../data/"))
weightsMAP_ref = bob.io.base.load(datafile('weightsAfterMAP.hdf5', __name__, path="../data/"))
# Compare to current results
# Gaps are quite large. This might be explained by the fact that there is no
# adaptation of a given Gaussian in torch3 when the corresponding responsibilities
# are below the responsibilities threshold
assert equals(gmm.means, meansMAP_ref, 2e-1)
assert equals(gmm.variances, variancesMAP_ref, 1e-4)
assert equals(gmm.weights, weightsMAP_ref, 1e-4)
def test_video_as_array_vs_dask():
import dask
path = datafile("testvideo.avi", "bob.bio.video.test")
start = time.time()
video = bob.bio.video.VideoAsArray(path, selection_style="all")
video = dask.array.from_array(video, (20, 1, 480, 640))
video = video.compute(scheduler="single-threaded")
load_time = time.time() - start
start = time.time()
reference = to_bob(np.array(list((imageio.get_reader(path).iter_data()))))
load_time2 = time.time() - start
# Here, we're also chunking each frame, but normally we would only chunk the first axis.
print(
f"FYI: It took {load_time:.2f} s to load the video with dask and {load_time2:.2f} s "
"to load directly. The slower loading with dask is expected.")
np.testing.assert_allclose(reference, video)
def test_gmm_ML_2():
# Trains a GMMMachine with ML_GMMTrainer; compares to an old reference
ar = bob.io.base.load(datafile('dataNormalized.hdf5', __name__, path="../data/"))
# Initialize GMMMachine
gmm = GMMMachine(5, 45)
gmm.means = bob.io.base.load(datafile('meansAfterKMeans.hdf5', __name__, path="../data/")).astype('float64')
gmm.variances = bob.io.base.load(datafile('variancesAfterKMeans.hdf5', __name__, path="../data/")).astype('float64')
gmm.weights = numpy.exp(bob.io.base.load(datafile('weightsAfterKMeans.hdf5', __name__, path="../data/")).astype('float64'))
threshold = 0.001
gmm.set_variance_thresholds(threshold)
# Initialize ML Trainer
prior = 0.001
max_iter_gmm = 25
accuracy = 0.00001
ml_gmmtrainer = ML_GMMTrainer(True, True, True, prior)
# Run ML
#ml_gmmtrainer.train(gmm, ar)
bob.learn.em.train(ml_gmmtrainer, gmm, ar, max_iterations = max_iter_gmm, convergence_threshold=accuracy)
# Test results
# Load torch3vision reference
meansML_ref = bob.io.base.load(datafile('meansAfterML.hdf5', __name__, path="../data/"))
variancesML_ref = bob.io.base.load(datafile('variancesAfterML.hdf5', __name__, path="../data/"))
weightsML_ref = bob.io.base.load(datafile('weightsAfterML.hdf5', __name__, path="../data/"))
# Compare to current results
assert equals(gmm.means, meansML_ref, 3e-3)
assert equals(gmm.variances, variancesML_ref, 3e-3)
assert equals(gmm.weights, weightsML_ref, 1e-4)
import numpy
import math
import bob.io.base
import bob.ip.base
from bob.io.base.test_utils import datafile
# load a test image
face_image = bob.io.base.load(datafile('image_r10.hdf5', 'bob.ip.base', 'data/affine'))
# create FaceEyesNorm class
face_eyes_norm = bob.ip.base.FaceEyesNorm(eyes_distance = 65, crop_size = (128, 128), eyes_center = (32, 63.5))
# normalize image
normalized_image = face_eyes_norm( face_image, right_eye = (66, 47), left_eye = (62, 70) )
# plot results, including eye locations in original and normalized image
from matplotlib import pyplot
pyplot.figure(figsize=(8,4))
pyplot.subplot(121) ; pyplot.imshow(face_image, cmap='gray') ; pyplot.plot([47, 70], [66, 62], 'rx', ms=10, mew=2); pyplot.axis('tight'); pyplot.title('Original Image')
pyplot.subplot(122) ; pyplot.imshow(normalized_image, cmap='gray') ; pyplot.plot([31, 96], [32, 32], 'rx', ms=10, mew=2); pyplot.axis('tight'); pyplot.title('Cropped Image')
pyplot.show()
