def __init__(self, **kwargs):
"""Initializes the local UBM-GMM tool chain with the given file selector object"""
# logger.warn("This class must be checked. Please verify that I didn't do any mistake here. I had to rename 'train_projector' into a 'train_enroller'!")
# initialize the UBMGMM base class
GMM.__init__(self, **kwargs)
# register a different set of functions in the Tool base class
Algorithm.__init__(self, requires_enroller_training = True, performs_projection = False)
def __init__(self, **kwargs):
"""Generates a test value that is read and written"""
# call base class constructor registering that this tool performs everything.
Algorithm.__init__(self,
performs_projection=False,
requires_enroller_training=True)
def __init__(
self,
# JFA training
subspace_dimension_of_u, # U subspace dimension
subspace_dimension_of_v, # V subspace dimension
jfa_training_iterations=10, # Number of EM iterations for the JFA training
# JFA enrollment
jfa_enroll_iterations=1, # Number of iterations for the enrollment phase
# parameters of the GMM
**kwargs):
"""Initializes the local UBM-GMM tool with the given file selector object"""
# call base class constructor
GMM.__init__(self, **kwargs)
# call tool constructor to overwrite what was set before
Algorithm.__init__(self,
performs_projection=True,
use_projected_features_for_enrollment=True,
requires_enroller_training=True,
subspace_dimension_of_u=subspace_dimension_of_u,
subspace_dimension_of_v=subspace_dimension_of_v,
jfa_training_iterations=jfa_training_iterations,
jfa_enroll_iterations=jfa_enroll_iterations,
multiple_model_scoring=None,
multiple_probe_scoring=None,
**kwargs)
self.subspace_dimension_of_u = subspace_dimension_of_u
self.subspace_dimension_of_v = subspace_dimension_of_v
self.jfa_training_iterations = jfa_training_iterations
self.jfa_enroll_iterations = jfa_enroll_iterations
self.jfa_trainer = bob.learn.em.JFATrainer()
def __init__(
self,
# ISV training
subspace_dimension_of_u, # U subspace dimension
isv_training_iterations=10, # Number of EM iterations for the ISV training
# ISV enrollment
isv_enroll_iterations=1, # Number of iterations for the enrollment phase
multiple_probe_scoring=None, # scoring when multiple probe files are available
# parameters of the GMM
**kwargs):
"""Initializes the local UBM-GMM tool with the given file selector object"""
# call base class constructor with its set of parameters
GMM.__init__(self, **kwargs)
# call tool constructor to overwrite what was set before
Algorithm.__init__(
self,
performs_projection=True,
use_projected_features_for_enrollment=True,
requires_enroller_training=
False, # not needed anymore because it's done while training the projector
split_training_features_by_client=True,
subspace_dimension_of_u=subspace_dimension_of_u,
isv_training_iterations=isv_training_iterations,
isv_enroll_iterations=isv_enroll_iterations,
multiple_model_scoring=None,
multiple_probe_scoring=multiple_probe_scoring,
**kwargs)
self.subspace_dimension_of_u = subspace_dimension_of_u
self.isv_training_iterations = isv_training_iterations
self.isv_enroll_iterations = isv_enroll_iterations
self.isv_trainer = bob.learn.em.ISVTrainer(self.relevance_factor)
def __init__(self, **kwargs):
"""Initializes the local UBM-GMM tool chain with the given file selector object"""
# logger.warn("This class must be checked. Please verify that I didn't do any mistake here. I had to rename 'train_projector' into a 'train_enroller'!")
# initialize the UBMGMM base class
GMM.__init__(self, **kwargs)
# register a different set of functions in the Tool base class
Algorithm.__init__(self, requires_enroller_training = True, performs_projection = False)
def __init__(self, **kwargs):
"""Generates a test value that is read and written"""
# call base class constructor registering that this tool performs everything.
Algorithm.__init__(
self,
performs_projection=False,
requires_enroller_training=True
)
def __init__(self,
distance_function=bob.math.chi_square,
is_distance_function=True,
multiple_probe_scoring='average'):
# call base class constructor
Algorithm.__init__(self,
distance_function=str(distance_function),
is_distance_function=is_distance_function,
multiple_model_scoring=None,
multiple_probe_scoring=multiple_probe_scoring)
# remember distance function
self.distance_function = distance_function
self.factor = -1. if is_distance_function else 1
def __init__(self,
machine_type='C_SVC',
kernel_type='LINEAR',
C=1,
**kwargs):
# initialize the UBMGMM base class
GMMRegular.__init__(self, **kwargs)
# register a different set of functions in the Tool base class
Algorithm.__init__(self,
requires_enroller_training=True,
performs_projection=False)
self.machine_type = machine_type
self.kernel_type = kernel_type
self.C = C
def __init__(
self,
distance_function = bob.math.chi_square,
is_distance_function = True,
multiple_probe_scoring = 'average'
):
# call base class constructor
Algorithm.__init__(
self,
distance_function = str(distance_function),
is_distance_function = is_distance_function,
multiple_model_scoring = None,
multiple_probe_scoring = multiple_probe_scoring
)
# remember distance function
self.distance_function = distance_function
self.factor = -1. if is_distance_function else 1
def __init__(
self,
# ISV training
subspace_dimension_of_u, # U subspace dimension
isv_training_iterations = 10, # Number of EM iterations for the ISV training
# ISV enrollment
isv_enroll_iterations = 1, # Number of iterations for the enrollment phase
multiple_probe_scoring = None, # scoring when multiple probe files are available
# parameters of the GMM
**kwargs
):
"""Initializes the local UBM-GMM tool with the given file selector object"""
# call base class constructor with its set of parameters
GMM.__init__(self, **kwargs)
# call tool constructor to overwrite what was set before
Algorithm.__init__(
self,
performs_projection = True,
use_projected_features_for_enrollment = True,
requires_enroller_training = False, # not needed anymore because it's done while training the projector
split_training_features_by_client = True,
subspace_dimension_of_u = subspace_dimension_of_u,
isv_training_iterations = isv_training_iterations,
isv_enroll_iterations = isv_enroll_iterations,
multiple_model_scoring = None,
multiple_probe_scoring = multiple_probe_scoring,
**kwargs
)
self.subspace_dimension_of_u = subspace_dimension_of_u
self.isv_training_iterations = isv_training_iterations
self.isv_enroll_iterations = isv_enroll_iterations
self.isv_trainer = bob.learn.em.ISVTrainer(self.relevance_factor)
def __init__(
self,
# JFA training
subspace_dimension_of_u, # U subspace dimension
subspace_dimension_of_v, # V subspace dimension
jfa_training_iterations = 10, # Number of EM iterations for the JFA training
# JFA enrollment
jfa_enroll_iterations = 1, # Number of iterations for the enrollment phase
# parameters of the GMM
**kwargs
):
"""Initializes the local UBM-GMM tool with the given file selector object"""
# call base class constructor
GMM.__init__(self, **kwargs)
# call tool constructor to overwrite what was set before
Algorithm.__init__(
self,
performs_projection = True,
use_projected_features_for_enrollment = True,
requires_enroller_training = True,
subspace_dimension_of_u = subspace_dimension_of_u,
subspace_dimension_of_v = subspace_dimension_of_v,
jfa_training_iterations = jfa_training_iterations,
jfa_enroll_iterations = jfa_enroll_iterations,
multiple_model_scoring = None,
multiple_probe_scoring = None,
**kwargs
)
self.subspace_dimension_of_u = subspace_dimension_of_u
self.subspace_dimension_of_v = subspace_dimension_of_v
self.jfa_training_iterations = jfa_training_iterations
self.jfa_enroll_iterations = jfa_enroll_iterations
self.jfa_trainer = bob.learn.em.JFATrainer()
def score_for_multiple_probes(self, model, probes):
"""This function computes the score between the given model and several given probe files."""
assert isinstance(model, bob.learn.em.ISVMachine)
[self._check_projected(probe) for probe in probes]
if self.probe_fusion_function is not None:
# When a multiple probe fusion function is selected, use it
return Algorithm.score_for_multiple_probes(self, model, probes)
else:
# Otherwise: compute joint likelihood of all probe features
# create GMM statistics from first probe statistics
# import pdb; pdb.set_trace()
gmmstats_acc = bob.learn.em.GMMStats(probes[0][0])
# gmmstats_acc = probes[0][0]
# add all other probe statistics
for i in range(1,len(probes)):
gmmstats_acc += probes[i][0]
# compute ISV score with the accumulated statistics
projected_isv_acc = numpy.ndarray(shape=(self.ubm.shape[0]*self.ubm.shape[1],), dtype=numpy.float64)
model.estimate_ux(gmmstats_acc, projected_isv_acc)
return model.forward_ux(gmmstats_acc, projected_isv_acc)
def score_for_multiple_probes(self, model, probes):
"""This function computes the score between the given model and several given probe files."""
assert isinstance(model, bob.learn.em.ISVMachine)
[self._check_projected(probe) for probe in probes]
if self.probe_fusion_function is not None:
# When a multiple probe fusion function is selected, use it
return Algorithm.score_for_multiple_probes(self, model, probes)
else:
# Otherwise: compute joint likelihood of all probe features
# create GMM statistics from first probe statistics
# import pdb; pdb.set_trace()
gmmstats_acc = bob.learn.em.GMMStats(probes[0][0])
# gmmstats_acc = probes[0][0]
# add all other probe statistics
for i in range(1,len(probes)):
gmmstats_acc += probes[i][0]
# compute ISV score with the accumulated statistics
projected_isv_acc = numpy.ndarray(shape=(self.ubm.shape[0]*self.ubm.shape[1],), dtype=numpy.float64)
model.estimate_ux(gmmstats_acc, projected_isv_acc)
return model.forward_ux(gmmstats_acc, projected_isv_acc)
def __init__(
self,
# IVector training
subspace_dimension_of_t, # T subspace dimension
tv_training_iterations = 25, # Number of EM iterations for the JFA training
update_sigma = True,
use_whitening = True,
use_lda = False,
use_wccn = False,
use_plda = False,
lda_dim = 50,
plda_dim_F = 50,
plda_dim_G = 50,
plda_training_iterations = 50,
# parameters of the GMM
**kwargs
):
"""Initializes the local GMM tool with the given file selector object"""
# call base class constructor with its set of parameters
GMM.__init__(self, **kwargs)
# call tool constructor to overwrite what was set before
Algorithm.__init__(
self,
performs_projection = True,
use_projected_features_for_enrollment = True,
requires_enroller_training = False, # not needed anymore because it's done while training the projector
split_training_features_by_client = True,
subspace_dimension_of_t = subspace_dimension_of_t,
tv_training_iterations = tv_training_iterations,
update_sigma = update_sigma,
use_whitening = use_whitening,
use_lda = use_lda,
use_wccn = use_wccn,
use_plda = use_plda,
lda_dim = lda_dim,
plda_dim_F = plda_dim_F,
plda_dim_G = plda_dim_G,
plda_training_iterations = plda_training_iterations,
multiple_model_scoring = None,
multiple_probe_scoring = None,
**kwargs
)
self.update_sigma = update_sigma
self.use_whitening = use_whitening
self.use_lda = use_lda
self.use_wccn = use_wccn
self.use_plda = use_plda
self.subspace_dimension_of_t = subspace_dimension_of_t
self.tv_training_iterations = tv_training_iterations
self.ivector_trainer = bob.learn.em.IVectorTrainer(update_sigma=update_sigma)
self.whitening_trainer = bob.learn.linear.WhiteningTrainer()
self.lda_dim = lda_dim
self.lda_trainer = bob.learn.linear.FisherLDATrainer(strip_to_rank=False)
self.wccn_trainer = bob.learn.linear.WCCNTrainer()
self.plda_trainer = bob.learn.em.PLDATrainer()
self.plda_dim_F = plda_dim_F
self.plda_dim_G = plda_dim_G
self.plda_training_iterations = plda_training_iterations
def __init__(
self,
# IVector training
subspace_dimension_of_t, # T subspace dimension
tv_training_iterations = 25, # Number of EM iterations for the JFA training
update_sigma = True,
use_whitening = True,
use_lda = False,
use_wccn = False,
use_plda = False,
lda_dim = 50,
plda_dim_F = 50,
plda_dim_G = 50,
plda_training_iterations = 50,
# parameters of the GMM
**kwargs
):
"""Initializes the local GMM tool with the given file selector object"""
# call base class constructor with its set of parameters
GMM.__init__(self, **kwargs)
# call tool constructor to overwrite what was set before
Algorithm.__init__(
self,
performs_projection = True,
use_projected_features_for_enrollment = True,
requires_enroller_training = False, # not needed anymore because it's done while training the projector
split_training_features_by_client = True,
subspace_dimension_of_t = subspace_dimension_of_t,
tv_training_iterations = tv_training_iterations,
update_sigma = update_sigma,
use_whitening = use_whitening,
use_lda = use_lda,
use_wccn = use_wccn,
use_plda = use_plda,
lda_dim = lda_dim,
plda_dim_F = plda_dim_F,
plda_dim_G = plda_dim_G,
plda_training_iterations = plda_training_iterations,
multiple_model_scoring = None,
multiple_probe_scoring = None,
**kwargs
)
self.update_sigma = update_sigma
self.use_whitening = use_whitening
self.use_lda = use_lda
self.use_wccn = use_wccn
self.use_plda = use_plda
self.subspace_dimension_of_t = subspace_dimension_of_t
self.tv_training_iterations = tv_training_iterations
self.ivector_trainer = bob.learn.em.IVectorTrainer(update_sigma=update_sigma)
self.whitening_trainer = bob.learn.linear.WhiteningTrainer()
self.lda_dim = lda_dim
self.lda_trainer = bob.learn.linear.FisherLDATrainer(strip_to_rank=False)
self.wccn_trainer = bob.learn.linear.WCCNTrainer()
self.plda_trainer = bob.learn.em.PLDATrainer()
self.plda_dim_F = plda_dim_F
self.plda_dim_G = plda_dim_G
self.plda_training_iterations = plda_training_iterations
def __init__(
self,
# parameters for the tool
gabor_jet_similarity_type,
multiple_feature_scoring='max_jet',
# some similarity functions might need a GaborWaveletTransform class, so we have to provide the parameters here as well...
gabor_directions=8,
gabor_scales=5,
gabor_sigma=2. * math.pi,
gabor_maximum_frequency=math.pi / 2.,
gabor_frequency_step=math.sqrt(.5),
gabor_power_of_k=0,
gabor_dc_free=True):
# call base class constructor
Algorithm.__init__(self,
gabor_jet_similarity_type=gabor_jet_similarity_type,
multiple_feature_scoring=multiple_feature_scoring,
gabor_directions=gabor_directions,
gabor_scales=gabor_scales,
gabor_sigma=gabor_sigma,
gabor_maximum_frequency=gabor_maximum_frequency,
gabor_frequency_step=gabor_frequency_step,
gabor_power_of_k=gabor_power_of_k,
gabor_dc_free=gabor_dc_free,
multiple_model_scoring=None,
multiple_probe_scoring=None)
# the Gabor wavelet transform; used by (some of) the Gabor jet similarities
gwt = bob.ip.gabor.Transform(number_of_scales=gabor_scales,
number_of_directions=gabor_directions,
sigma=gabor_sigma,
k_max=gabor_maximum_frequency,
k_fac=gabor_frequency_step,
power_of_k=gabor_power_of_k,
dc_free=gabor_dc_free)
# jet comparison function
self.similarity_function = bob.ip.gabor.Similarity(
gabor_jet_similarity_type, gwt)
# how to proceed with multiple features per model
self.jet_scoring = {
'average_model': None, # compute an average model
'average': numpy.average, # compute the average similarity
'min_jet':
min, # for each jet location, compute the minimum similarity
'max_jet':
max, # for each jet location, compute the maximum similarity
'med_jet': numpy.
median, # for each jet location, compute the median similarity
'min_graph': numpy.
average, # for each model graph, compute the minimum average similarity
'max_graph': numpy.
average, # for each model graph, compute the maximum average similarity
'med_graph': numpy.
average, # for each model graph, compute the median average similarity
}[multiple_feature_scoring]
self.graph_scoring = {
'average_model': None, # compute an average model
'average': numpy.average, # compute the average similarity
'min_jet': numpy.
average, # for each jet location, compute the minimum similarity
'max_jet': numpy.
average, # for each jet location, compute the maximum similarity
'med_jet': numpy.
average, # for each jet location, compute the median similarity
'min_graph':
min, # for each model graph, compute the minimum average similarity
'max_graph':
max, # for each model graph, compute the maximum average similarity
'med_graph': numpy.
median, # for each model graph, compute the median average similarity
}[multiple_feature_scoring]
def __init__(
self,
# parameters for the GMM
number_of_gaussians,
# parameters of UBM training
kmeans_training_iterations = 25, # Maximum number of iterations for K-Means
gmm_training_iterations = 25, # Maximum number of iterations for ML GMM Training
training_threshold = 5e-4, # Threshold to end the ML training
variance_threshold = 5e-4, # Minimum value that a variance can reach
update_weights = True,
update_means = True,
update_variances = True,
# parameters of the GMM enrollment
relevance_factor = 4, # Relevance factor as described in Reynolds paper
gmm_enroll_iterations = 1, # Number of iterations for the enrollment phase
responsibility_threshold = 0, # If set, the weight of a particular Gaussian will at least be greater than this threshold. In the case the real weight is lower, the prior mean value will be used to estimate the current mean and variance.
INIT_SEED = 5489,
# scoring
scoring_function = bob.learn.em.linear_scoring
):
"""Initializes the local UBM-GMM tool chain with the given file selector object"""
# call base class constructor and register that this tool performs projection
Algorithm.__init__(
self,
performs_projection = True,
use_projected_features_for_enrollment = False,
number_of_gaussians = number_of_gaussians,
kmeans_training_iterations = kmeans_training_iterations,
gmm_training_iterations = gmm_training_iterations,
training_threshold = training_threshold,
variance_threshold = variance_threshold,
update_weights = update_weights,
update_means = update_means,
update_variances = update_variances,
relevance_factor = relevance_factor,
gmm_enroll_iterations = gmm_enroll_iterations,
responsibility_threshold = responsibility_threshold,
INIT_SEED = INIT_SEED,
scoring_function = str(scoring_function),
multiple_model_scoring = None,
multiple_probe_scoring = 'average'
)
# copy parameters
self.gaussians = number_of_gaussians
self.kmeans_training_iterations = kmeans_training_iterations
self.gmm_training_iterations = gmm_training_iterations
self.training_threshold = training_threshold
self.variance_threshold = variance_threshold
self.update_weights = update_weights
self.update_means = update_means
self.update_variances = update_variances
self.relevance_factor = relevance_factor
self.gmm_enroll_iterations = gmm_enroll_iterations
self.init_seed = INIT_SEED
self.rng = bob.core.random.mt19937(self.init_seed)
self.responsibility_threshold = responsibility_threshold
self.scoring_function = scoring_function
self.ubm = None
self.kmeans_trainer = bob.learn.em.KMeansTrainer()
self.ubm_trainer = bob.learn.em.ML_GMMTrainer(self.update_means, self.update_variances, self.update_weights, self.responsibility_threshold)
def __init__(
self,
# parameters for the tool
gabor_jet_similarity_type,
multiple_feature_scoring = 'max_jet',
# some similarity functions might need a GaborWaveletTransform class, so we have to provide the parameters here as well...
gabor_directions = 8,
gabor_scales = 5,
gabor_sigma = 2. * math.pi,
gabor_maximum_frequency = math.pi / 2.,
gabor_frequency_step = math.sqrt(.5),
gabor_power_of_k = 0,
gabor_dc_free = True
):
# call base class constructor
Algorithm.__init__(
self,
gabor_jet_similarity_type = gabor_jet_similarity_type,
multiple_feature_scoring = multiple_feature_scoring,
gabor_directions = gabor_directions,
gabor_scales = gabor_scales,
gabor_sigma = gabor_sigma,
gabor_maximum_frequency = gabor_maximum_frequency,
gabor_frequency_step = gabor_frequency_step,
gabor_power_of_k = gabor_power_of_k,
gabor_dc_free = gabor_dc_free,
multiple_model_scoring = None,
multiple_probe_scoring = None
)
# the Gabor wavelet transform; used by (some of) the Gabor jet similarities
gwt = bob.ip.gabor.Transform(
number_of_scales = gabor_scales,
number_of_directions = gabor_directions,
sigma = gabor_sigma,
k_max = gabor_maximum_frequency,
k_fac = gabor_frequency_step,
power_of_k = gabor_power_of_k,
dc_free = gabor_dc_free
)
# jet comparison function
self.similarity_function = bob.ip.gabor.Similarity(gabor_jet_similarity_type, gwt)
# how to proceed with multiple features per model
self.jet_scoring = {
'average_model' : None, # compute an average model
'average' : numpy.average, # compute the average similarity
'min_jet' : min, # for each jet location, compute the minimum similarity
'max_jet' : max, # for each jet location, compute the maximum similarity
'med_jet' : numpy.median, # for each jet location, compute the median similarity
'min_graph' : numpy.average, # for each model graph, compute the minimum average similarity
'max_graph' : numpy.average, # for each model graph, compute the maximum average similarity
'med_graph' : numpy.average, # for each model graph, compute the median average similarity
}[multiple_feature_scoring]
self.graph_scoring = {
'average_model' : None, # compute an average model
'average' : numpy.average, # compute the average similarity
'min_jet' : numpy.average, # for each jet location, compute the minimum similarity
'max_jet' : numpy.average, # for each jet location, compute the maximum similarity
'med_jet' : numpy.average, # for each jet location, compute the median similarity
'min_graph' : min, # for each model graph, compute the minimum average similarity
'max_graph' : max, # for each model graph, compute the maximum average similarity
'med_graph' : numpy.median, # for each model graph, compute the median average similarity
}[multiple_feature_scoring]
def __init__(
self,
# parameters for the GMM
number_of_gaussians,
# parameters of UBM training
kmeans_training_iterations = 25, # Maximum number of iterations for K-Means
gmm_training_iterations = 25, # Maximum number of iterations for ML GMM Training
training_threshold = 5e-4, # Threshold to end the ML training
variance_threshold = 5e-4, # Minimum value that a variance can reach
update_weights = True,
update_means = True,
update_variances = True,
# parameters of the GMM enrollment
relevance_factor = 4, # Relevance factor as described in Reynolds paper
gmm_enroll_iterations = 1, # Number of iterations for the enrollment phase
responsibility_threshold = 0, # If set, the weight of a particular Gaussian will at least be greater than this threshold. In the case the real weight is lower, the prior mean value will be used to estimate the current mean and variance.
INIT_SEED = 5489,
# scoring
scoring_function = bob.learn.em.linear_scoring
):
"""Initializes the local UBM-GMM tool chain with the given file selector object"""
# call base class constructor and register that this tool performs projection
Algorithm.__init__(
self,
performs_projection = True,
use_projected_features_for_enrollment = False,
number_of_gaussians = number_of_gaussians,
kmeans_training_iterations = kmeans_training_iterations,
gmm_training_iterations = gmm_training_iterations,
training_threshold = training_threshold,
variance_threshold = variance_threshold,
update_weights = update_weights,
update_means = update_means,
update_variances = update_variances,
relevance_factor = relevance_factor,
gmm_enroll_iterations = gmm_enroll_iterations,
responsibility_threshold = responsibility_threshold,
INIT_SEED = INIT_SEED,
scoring_function = str(scoring_function),
multiple_model_scoring = None,
multiple_probe_scoring = 'average'
)
# copy parameters
self.gaussians = number_of_gaussians
self.kmeans_training_iterations = kmeans_training_iterations
self.gmm_training_iterations = gmm_training_iterations
self.training_threshold = training_threshold
self.variance_threshold = variance_threshold
self.update_weights = update_weights
self.update_means = update_means
self.update_variances = update_variances
self.relevance_factor = relevance_factor
self.gmm_enroll_iterations = gmm_enroll_iterations
self.init_seed = INIT_SEED
self.rng = bob.core.random.mt19937(self.init_seed)
self.responsibility_threshold = responsibility_threshold
self.scoring_function = scoring_function
self.ubm = None
self.kmeans_trainer = bob.learn.em.KMeansTrainer()
self.ubm_trainer = bob.learn.em.ML_GMMTrainer(self.update_means, self.update_variances, self.update_weights, self.responsibility_threshold)
