def compute_gmm_given_dataset(src_cfg, nr_clusters, **kwargs):
""" Uses the conventions from the dataset for loading a subset of
descriptors, loading the PCA object and choosing an output file.
"""
# Set default parameters.
ip_type = kwargs.get('ip_type', 'dense5.track15mbh')
nr_iterations = kwargs.get('nr_iterations', 100)
nr_threads = kwargs.get('nr_threads', multiprocessing.cpu_count())
seed = kwargs.get('seed', 1)
nr_redos = kwargs.get('nr_redos', 4)
suffix = kwargs.get('suffix', '')
nr_pca_components = kwargs.get('nr_pca_components', 64)
dataset = Dataset(src_cfg, ip_type=ip_type, suffix=suffix)
filename_pca = os.path.join(dataset.FEAT_DIR, 'pca', 'pca_%d.pkl' % nr_pca_components)
data = load_subsample_descriptors(dataset)
pca = load_pca(filename_pca)
transformed_data = pca.transform(data)
# Do the computation.
gmm = compute_gmm(
transformed_data, nr_clusters, nr_iterations,
nr_threads, seed, nr_redos)
outfilename = os.path.join(
dataset.FEAT_DIR, 'gmm', 'gmm_%d' % nr_clusters)
save_gmm(gmm, outfilename)
def compute_gmm_given_dataset(src_cfg, nr_clusters, **kwargs):
""" Uses the conventions from the dataset for loading a subset of
descriptors, loading the PCA object and choosing an output file.
"""
# Set default parameters.
ip_type = kwargs.get('ip_type', 'dense5.track15mbh')
nr_iterations = kwargs.get('nr_iterations', 100)
nr_threads = kwargs.get('nr_threads', multiprocessing.cpu_count())
seed = kwargs.get('seed', 1)
nr_redos = kwargs.get('nr_redos', 4)
suffix = kwargs.get('suffix', '')
nr_pca_components = kwargs.get('nr_pca_components', 64)
dataset = Dataset(src_cfg, ip_type=ip_type, suffix=suffix)
filename_pca = os.path.join(dataset.FEAT_DIR, 'pca',
'pca_%d.pkl' % nr_pca_components)
data = load_subsample_descriptors(dataset)
pca = load_pca(filename_pca)
transformed_data = pca.transform(data)
# Do the computation.
gmm = compute_gmm(transformed_data, nr_clusters, nr_iterations, nr_threads,
seed, nr_redos)
outfilename = os.path.join(dataset.FEAT_DIR, 'gmm', 'gmm_%d' % nr_clusters)
save_gmm(gmm, outfilename)
def compute_pca_given_dataset(src_cfg, **kwargs):
""" Uses the conventions from the dataset for loading a subset of
descriptors and choosing an output file.
"""
# Set default parameters.
ip_type = kwargs.get('ip_type', 'dense5.track15mbh')
n_components = kwargs.get('n_components', 64)
suffix = kwargs.get('suffix', '')
dataset = Dataset(src_cfg, ip_type=ip_type, suffix=suffix)
data = load_subsample_descriptors(dataset)
# Do the computation.
pca = compute_pca(data, n_components)
outfilename = os.path.join(
dataset.FEAT_DIR, 'pca', 'pca_%d.pkl' % n_components)
save_pca(pca, outfilename)
