Skip to content
Snippets Groups Projects

Add pre-commit

Merged Add pre-commit
deprecation into master
Merged Amir MOHAMMADI requested to merge deprecation into master
Compare
and
161 files
+ 4558
− 2483
Compare changes
  • Side-by-side
  • Inline
Files
161
bob/bio/face/algorithm/Histogram.py
+ 211
− 190
@@ -6,210 +6,231 @@ import numpy
from bob.bio.base.algorithm import Algorithm
def chi_square(*args):
"""
Calculates the chi-square distance between two histograms.
@param hist1 The first histogram
@param hist2 The second histogram
@returns The chi-square distance between the two histograms
"""
if len(args) == 2:
h1, h2 = args
d = 0
for i in range(h1.shape[0]):
if h1[i] != h2[i]: d += int(((h1[i] - h2[i])**2) / (h1[i] + h2[i]))
return d
else:
# histogram intersection with sparse histograms
index_1, value_1, index_2, value_2 = args
raise NotImplementedError("chi_square distance with sparse histograms is not implemented yet")
def histogram_intersection(*args):
"""
Calculates the histogram intersection between two histograms.
@param hist1 The first histogram
@param hist2 The second histogram
@returns The histogram intersection between the two histograms
"""
if len(args) == 2:
hist1, hist2 = args
return numpy.sum(numpy.minimum(hist1, hist2))
else:
# histogram intersection with sparse histograms
index_1, value_1, index_2, value_2 = args
i1, i2, i1_end, i2_end = 0, 0, index_1.shape[0], index_2.shape[0]
p1, p2 = index_1[i1], index_2[i2]
sum = 0
while i1 < i1_end and i2 < i2_end:
p1 = index_1[i1]
p2 = index_2[i2]
if p1 == p2:
sum += numpy.minimum(value_1[i1], value_2[i2])
i1 += 1
i2 += 1
elif p1 < p2:
i1 += 1
else:
i2 += 1
return sum
class Histogram (Algorithm):
"""Computes the distance between histogram sequences.
Both sparse and non-sparse representations of histograms are supported.
For enrollment, to date only the averaging of histograms is implemented.
**Parameters:**
distance_function : function
The function to be used to compare two histograms.
This function should accept sparse histograms.
is_distance_function : bool
Is the given ``distance_function`` distance function (lower values are better) or a similarity function (higher values are better)?
multiple_probe_scoring : str or ``None``
The way, scores are fused when multiple probes are available.
See :py:func:`bob.bio.base.score_fusion_strategy` for possible values.
"""
def __init__(
self,
distance_function = 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 chi_square(*args):
"""
Calculates the chi-square distance between two histograms.
def _is_sparse(self, feature):
assert isinstance(feature, numpy.ndarray)
return feature.ndim == 2
@param hist1 The first histogram
@param hist2 The second histogram
def _check_feature(self, feature, sparse):
assert isinstance(feature, numpy.ndarray)
if sparse:
# check that we have a 2D array
assert feature.ndim == 2
assert feature.shape[0] == 2
@returns The chi-square distance between the two histograms
"""
if len(args) == 2:
h1, h2 = args
d = 0
for i in range(h1.shape[0]):
if h1[i] != h2[i]:
d += int(((h1[i] - h2[i]) ** 2) / (h1[i] + h2[i]))
return d
else:
assert feature.ndim == 1
# histogram intersection with sparse histograms
index_1, value_1, index_2, value_2 = args
raise NotImplementedError(
"chi_square distance with sparse histograms is not implemented yet"
)
def enroll(self, enroll_features):
"""enroll(enroll_features) -> model
Enrolls a model by taking the average of all histograms.
enroll_features : [1D or 2D :py:class:`numpy.ndarray`]
The histograms that should be averaged.
Histograms can be specified sparse (2D) or non-sparse (1D)
def histogram_intersection(*args):
"""
Calculates the histogram intersection between two histograms.
**Returns:**
@param hist1 The first histogram
@param hist2 The second histogram
model : 1D or 2D :py:class:`numpy.ndarray`
The averaged histogram, sparse (2D) or non-sparse (1D).
@returns The histogram intersection between the two histograms
"""
assert len(enroll_features)
sparse = self._is_sparse(enroll_features[0])
[self._check_feature(feature, sparse) for feature in enroll_features]
if sparse:
# get all indices for the sparse model
values = {}
# iterate through all sparse features
for feature in enroll_features:
# collect the values by index
for j in range(feature.shape[1]):
index = int(feature[0,j])
value = feature[1,j] / float(len(enroll_features))
# add up values
if index in values:
values[index] += value
else:
values[index] = value
# create model containing all the used indices
model = numpy.ndarray((2, len(values)), dtype = numpy.float64)
for i, index in enumerate(sorted(values.keys())):
model[0,i] = index
model[1,i] = values[index]
if len(args) == 2:
hist1, hist2 = args
return numpy.sum(numpy.minimum(hist1, hist2))
else:
model = numpy.zeros(enroll_features[0].shape, dtype = numpy.float64)
# add up models
for feature in enroll_features:
model += feature
# normalize by number of models
model /= float(len(enroll_features))
# return averaged model
return model
def score(self, model, probe):
"""score(model, probe) -> score
Computes the score of the probe and the model using the desired histogram distance function.
The resulting score is the negative distance, if ``is_distance_function = True``.
Both sparse and non-sparse models and probes are accepted, but their sparseness must agree.
# histogram intersection with sparse histograms
index_1, value_1, index_2, value_2 = args
i1, i2, i1_end, i2_end = 0, 0, index_1.shape[0], index_2.shape[0]
p1, p2 = index_1[i1], index_2[i2]
sum = 0
while i1 < i1_end and i2 < i2_end:
p1 = index_1[i1]
p2 = index_2[i2]
if p1 == p2:
sum += numpy.minimum(value_1[i1], value_2[i2])
i1 += 1
i2 += 1
elif p1 < p2:
i1 += 1
else:
i2 += 1
return sum
class Histogram(Algorithm):
"""Computes the distance between histogram sequences.
Both sparse and non-sparse representations of histograms are supported.
For enrollment, to date only the averaging of histograms is implemented.
**Parameters:**
model : 1D or 2D :py:class:`numpy.ndarray`
The model enrolled by the :py:meth:`enroll` function.
probe : 1D or 2D :py:class:`numpy.ndarray`
The probe histograms, which can be specified sparse (2D) or non-sparse (1D)
distance_function : function
The function to be used to compare two histograms.
This function should accept sparse histograms.
**Returns:**
is_distance_function : bool
Is the given ``distance_function`` distance function (lower values are better) or a similarity function (higher values are better)?
score : float
The resulting similarity score.
multiple_probe_scoring : str or ``None``
The way, scores are fused when multiple probes are available.
See :py:func:`bob.bio.base.score_fusion_strategy` for possible values.
"""
sparse = self._is_sparse(probe)
self._check_feature(model, sparse)
self._check_feature(probe, sparse)
if sparse:
# assure that the probe is sparse as well
return self.factor * self.distance_function(model[0,:], model[1,:], probe[0,:], probe[1,:])
else:
return self.factor * self.distance_function(model, probe)
# overwrite functions to avoid them being documented.
def train_projector(*args, **kwargs) : raise NotImplementedError("This function is not implemented and should not be called.")
def load_projector(*args, **kwargs) : pass
def project(*args, **kwargs) : raise NotImplementedError("This function is not implemented and should not be called.")
def write_feature(*args, **kwargs) : raise NotImplementedError("This function is not implemented and should not be called.")
def read_feature(*args, **kwargs) : raise NotImplementedError("This function is not implemented and should not be called.")
def train_enroller(*args, **kwargs) : raise NotImplementedError("This function is not implemented and should not be called.")
def load_enroller(*args, **kwargs) : pass
def score_for_multiple_models(self, models, probe):
self._check_feature(probe, self._is_sparse(probe))
scores = [ self.score(m, probe) for m in models ]
return scores
def __init__(
self,
distance_function=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 = -1.0 if is_distance_function else 1
def _is_sparse(self, feature):
assert isinstance(feature, numpy.ndarray)
return feature.ndim == 2
def _check_feature(self, feature, sparse):
assert isinstance(feature, numpy.ndarray)
if sparse:
# check that we have a 2D array
assert feature.ndim == 2
assert feature.shape[0] == 2
else:
assert feature.ndim == 1
def enroll(self, enroll_features):
"""enroll(enroll_features) -> model
Enrolls a model by taking the average of all histograms.
enroll_features : [1D or 2D :py:class:`numpy.ndarray`]
The histograms that should be averaged.
Histograms can be specified sparse (2D) or non-sparse (1D)
**Returns:**
model : 1D or 2D :py:class:`numpy.ndarray`
The averaged histogram, sparse (2D) or non-sparse (1D).
"""
assert len(enroll_features)
sparse = self._is_sparse(enroll_features[0])
[self._check_feature(feature, sparse) for feature in enroll_features]
if sparse:
# get all indices for the sparse model
values = {}
# iterate through all sparse features
for feature in enroll_features:
# collect the values by index
for j in range(feature.shape[1]):
index = int(feature[0, j])
value = feature[1, j] / float(len(enroll_features))
# add up values
if index in values:
values[index] += value
else:
values[index] = value
# create model containing all the used indices
model = numpy.ndarray((2, len(values)), dtype=numpy.float64)
for i, index in enumerate(sorted(values.keys())):
model[0, i] = index
model[1, i] = values[index]
else:
model = numpy.zeros(enroll_features[0].shape, dtype=numpy.float64)
# add up models
for feature in enroll_features:
model += feature
# normalize by number of models
model /= float(len(enroll_features))
# return averaged model
return model
def score(self, model, probe):
"""score(model, probe) -> score
Computes the score of the probe and the model using the desired histogram distance function.
The resulting score is the negative distance, if ``is_distance_function = True``.
Both sparse and non-sparse models and probes are accepted, but their sparseness must agree.
**Parameters:**
model : 1D or 2D :py:class:`numpy.ndarray`
The model enrolled by the :py:meth:`enroll` function.
probe : 1D or 2D :py:class:`numpy.ndarray`
The probe histograms, which can be specified sparse (2D) or non-sparse (1D)
**Returns:**
score : float
The resulting similarity score.
"""
sparse = self._is_sparse(probe)
self._check_feature(model, sparse)
self._check_feature(probe, sparse)
if sparse:
# assure that the probe is sparse as well
return self.factor * self.distance_function(
model[0, :], model[1, :], probe[0, :], probe[1, :]
)
else:
return self.factor * self.distance_function(model, probe)
# overwrite functions to avoid them being documented.
def train_projector(*args, **kwargs):
raise NotImplementedError(
"This function is not implemented and should not be called."
)
def load_projector(*args, **kwargs):
pass
def project(*args, **kwargs):
raise NotImplementedError(
"This function is not implemented and should not be called."
)
def write_feature(*args, **kwargs):
raise NotImplementedError(
"This function is not implemented and should not be called."
)
def read_feature(*args, **kwargs):
raise NotImplementedError(
"This function is not implemented and should not be called."
)
def train_enroller(*args, **kwargs):
raise NotImplementedError(
"This function is not implemented and should not be called."
)
def load_enroller(*args, **kwargs):
pass
def score_for_multiple_models(self, models, probe):
self._check_feature(probe, self._is_sparse(probe))
scores = [self.score(m, probe) for m in models]
return scores