diff --git a/bob/bio/gmm/algorithm/GMM.py b/bob/bio/gmm/algorithm/GMM.py
index bc6c1434ba411571de4d95f643ba944dc635d189..a122da9bf79cde683ddcf6819afbe9cf1037818e 100644
--- a/bob/bio/gmm/algorithm/GMM.py
+++ b/bob/bio/gmm/algorithm/GMM.py
@@ -14,6 +14,7 @@ import copy
import logging
from typing import Callable
+from typing import Union
import dask
import dask.array as da
@@ -23,10 +24,10 @@ from h5py import File as HDF5File
from sklearn.base import BaseEstimator
from bob.bio.base.pipelines.vanilla_biometrics.abstract_classes import BioAlgorithm
-from bob.learn.em.cluster import KMeansMachine
-from bob.learn.em.mixture import GMMMachine
-from bob.learn.em.mixture import GMMStats
-from bob.learn.em.mixture import linear_scoring
+from bob.learn.em import GMMMachine
+from bob.learn.em import GMMStats
+from bob.learn.em import KMeansMachine
+from bob.learn.em import linear_scoring
logger = logging.getLogger(__name__)
@@ -49,19 +50,23 @@ class GMM(BioAlgorithm, BaseEstimator):
number_of_gaussians: int,
# parameters of UBM training
kmeans_training_iterations: int = 25, # Maximum number of iterations for K-Means
+ kmeans_init_iterations: Union[
+ int, None
+ ] = None, # Maximum number of iterations for K-Means init
+ kmeans_oversampling_factor: int = 64,
ubm_training_iterations: int = 25, # Maximum number of iterations for GMM Training
training_threshold: float = 5e-4, # Threshold to end the ML training
variance_threshold: float = 5e-4, # Minimum value that a variance can reach
update_means: bool = True,
update_variances: bool = True,
update_weights: bool = True,
- # parameters of the GMM enrollment
- gmm_enroll_iterations: int = 1, # Number of iterations for the enrollment phase
+ # parameters of the GMM enrollment (MAP)
+ gmm_enroll_iterations: int = 1,
enroll_update_means: bool = True,
enroll_update_variances: bool = False,
enroll_update_weights: bool = False,
- relevance_factor: float = 4, # Relevance factor as described in Reynolds paper
- responsibility_threshold: float = 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.
+ enroll_relevance_factor: Union[float, None] = 4,
+ enroll_alpha: float = 0.5,
# scoring
scoring_function: Callable = linear_scoring,
# RNG
@@ -75,6 +80,11 @@ class GMM(BioAlgorithm, BaseEstimator):
The number of Gaussians used in the UBM and the models.
kmeans_training_iterations
Number of e-m iterations to train k-means initializing the UBM.
+ kmeans_init_iterations
+ Number of iterations used for setting the k-means initial centroids.
+ if None, will use the same as kmeans_training_iterations.
+ kmeans_oversampling_factor
+ Oversampling factor used by k-means initializer.
ubm_training_iterations
Number of e-m iterations for training the UBM.
training_threshold
@@ -95,12 +105,11 @@ class GMM(BioAlgorithm, BaseEstimator):
Decides wether the means of the Gaussians are updated while enrolling.
enroll_update_variances
Decides wether the variancess of the Gaussians are updated while enrolling.
- relevance_factor
- Relevance factor as described in Reynolds paper.
- responsibility_threshold
- If set, the weight of a particular Gaussian will at least be greater than
- this threshold. In the case where the real weight is lower, the prior mean
- value will be used to estimate the current mean and variance.
+ enroll_relevance_factor
+ For enrollment: MAP relevance factor as described in Reynolds paper.
+ If None, will not apply Reynolds adaptation.
+ enroll_alpha
+ For enrollment: MAP adaptation coefficient.
init_seed
Seed for the random number generation.
scoring_function
@@ -110,20 +119,26 @@ class GMM(BioAlgorithm, BaseEstimator):
# Copy parameters
self.number_of_gaussians = number_of_gaussians
self.kmeans_training_iterations = kmeans_training_iterations
+ self.kmeans_init_iterations = (
+ kmeans_training_iterations
+ if kmeans_init_iterations is None
+ else kmeans_init_iterations
+ )
+ self.kmeans_oversampling_factor = kmeans_oversampling_factor
self.ubm_training_iterations = ubm_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.enroll_relevance_factor = enroll_relevance_factor
+ self.enroll_alpha = enroll_alpha
self.gmm_enroll_iterations = gmm_enroll_iterations
self.enroll_update_means = enroll_update_means
self.enroll_update_weights = enroll_update_weights
self.enroll_update_variances = enroll_update_variances
self.init_seed = init_seed
self.rng = self.init_seed
- self.responsibility_threshold = responsibility_threshold
self.scoring_function = scoring_function
@@ -186,7 +201,6 @@ class GMM(BioAlgorithm, BaseEstimator):
# Use the array to train a GMM and return it
logger.info("Enrolling with %d feature vectors", array.shape[0])
- # TODO accept responsibility_threshold in bob.learn.em
with dask.config.set(scheduler="threads"):
gmm = GMMMachine(
n_gaussians=self.number_of_gaussians,
@@ -199,6 +213,8 @@ class GMM(BioAlgorithm, BaseEstimator):
update_variances=self.enroll_update_variances,
update_weights=self.enroll_update_weights,
mean_var_update_threshold=self.variance_threshold,
+ relevance_factor=self.enroll_relevance_factor,
+ alpha=self.enroll_alpha,
)
gmm.fit(array)
return gmm
@@ -304,7 +320,7 @@ class GMM(BioAlgorithm, BaseEstimator):
convergence_threshold=self.training_threshold,
max_iter=self.kmeans_training_iterations,
init_method="k-means||",
- init_max_iter=5,
+ init_max_iter=self.kmeans_init_iterations,
random_state=self.init_seed,
),
)
@@ -319,10 +335,21 @@ class GMM(BioAlgorithm, BaseEstimator):
def transform(self, X, **kwargs):
"""Passthrough. Enroll applies a different transform as score."""
# The idea would be to apply the projection in Transform (going from extracted
- # to GMMStats), but we must not apply this during the training (fit requires
- # extracted data directly).
+ # to GMMStats), but we must not apply this during the training or enrollment
+ # (those require extracted data directly, not projected).
# `project` is applied in the score function directly.
return X
+ @classmethod
+ def custom_enrolled_save_fn(cls, data, path):
+ data.save(path)
+
+ def custom_enrolled_load_fn(self, path):
+ return GMMMachine.from_hdf5(path, ubm=self.ubm)
+
def _more_tags(self):
- return {"bob_fit_supports_dask_array": True}
+ return {
+ "bob_fit_supports_dask_array": True,
+ "bob_enrolled_save_fn": self.custom_enrolled_save_fn,
+ "bob_enrolled_load_fn": self.custom_enrolled_load_fn,
+ }
diff --git a/bob/bio/gmm/test/data/gmm_enrolled.hdf5 b/bob/bio/gmm/test/data/gmm_enrolled.hdf5
index 4b48c0fa3c6376c2e5c41aa164dac481c847e380..94190e2b4c1cfb7bdf36627b5fd770fb9079ded0 100644
Binary files a/bob/bio/gmm/test/data/gmm_enrolled.hdf5 and b/bob/bio/gmm/test/data/gmm_enrolled.hdf5 differ
diff --git a/bob/bio/gmm/test/data/gmm_projected.hdf5 b/bob/bio/gmm/test/data/gmm_projected.hdf5
index 1bdbba5769c5a7195efe81d6b0b14685ef980097..894d3f7e6594fab0c9720dd71b7f5082f338ad4e 100644
Binary files a/bob/bio/gmm/test/data/gmm_projected.hdf5 and b/bob/bio/gmm/test/data/gmm_projected.hdf5 differ
diff --git a/bob/bio/gmm/test/data/gmm_ubm.hdf5 b/bob/bio/gmm/test/data/gmm_ubm.hdf5
index 780baa0277bd2d7cf68ab048c7df69900fa6b875..6f40b3b024d24a0349f122864a1b3c254b81955f 100644
Binary files a/bob/bio/gmm/test/data/gmm_ubm.hdf5 and b/bob/bio/gmm/test/data/gmm_ubm.hdf5 differ
diff --git a/bob/bio/gmm/test/test_gmm.py b/bob/bio/gmm/test/test_gmm.py
index ea7b6a4c9cc21853924d5eb8fbabb6d8f70d3da9..0c9f126a3c9162e9cc32afb7f198f526860ab51f 100644
--- a/bob/bio/gmm/test/test_gmm.py
+++ b/bob/bio/gmm/test/test_gmm.py
@@ -18,7 +18,6 @@
# along with this program. If not, see <http://www.gnu.org/licenses/>.
import logging
-import os
import tempfile
import numpy
@@ -28,8 +27,8 @@ import bob.bio.gmm
from bob.bio.base.test import utils
from bob.bio.gmm.algorithm import GMM
-from bob.learn.em.mixture.gmm import GMMMachine
-from bob.learn.em.mixture.gmm import GMMStats
+from bob.learn.em import GMMMachine
+from bob.learn.em import GMMStats
logger = logging.getLogger(__name__)
@@ -97,7 +96,7 @@ def test_projector():
# Generate and project random feature
feature = utils.random_array((20, 45), -5.0, 5.0, seed=seed_value)
projected = gmm1.project(feature)
- assert isinstance(projected, bob.learn.em.mixture.GMMStats)
+ assert isinstance(projected, GMMStats)
reference_file = pkg_resources.resource_filename(
"bob.bio.gmm.test", "data/gmm_projected.hdf5"
@@ -137,8 +136,8 @@ def test_enroll():
with tempfile.NamedTemporaryFile(prefix="bob_", suffix="_bioref.hdf5") as fd:
temp_file = fd.name
- gmm1.write_biometric_reference(biometric_reference, reference_file)
- assert os.path.exists(temp_file)
+ gmm1.write_biometric_reference(biometric_reference, temp_file)
+ assert GMMMachine.from_hdf5(temp_file, ubm).is_similar_to(gmm2)
def test_score():