atep_bovw = kernels.compute_ATEP_kernels(feats_path + '/bovwtree/', videonames, traintest_parts, xml_config['features_list'], \
kernels_path + '/atep-bovw/', kernel_type='intersection', norm='l1', power_norm=False, \
use_disk=False, nt=args.nt, verbose=args.verbose)
params = [[[1]], [1], np.linspace(0,1,21), desc_weights_gbl]
results = classification.classify(atep_bovw, \
class_labels, traintest_parts, params, \
xml_config['features_list'], \
C=C_gbl,
strategy=strategy_gbl,
opt_criterion=opt_criterion,
verbose=args.verbose)
classification.print_results(results)
if 'atep-fv' in args.methods:
tracklet_representation.train_fv_gmms(tracklets_path, videonames, traintest_parts, xml_config['features_list'], intermediates_path, pca_reduction=True, nt=args.nt, verbose=args.verbose)
tracklet_representation.compute_fv_descriptors_multithread(tracklets_path, intermediates_path, videonames, traintest_parts, xml_config['features_list'], \
feats_path + '/fvtree/', \
treelike=True, pca_reduction=True, clusters_path=clusters_path, nt=args.nt, verbose=args.verbose)
atep_fv = kernels.compute_ATEP_kernels(feats_path + '/fvtree/', videonames, traintest_parts, xml_config['features_list'], \
kernels_path + '/atep-fv/', use_disk=False, nt=args.nt, verbose=args.verbose)
params = [[[1]], [1], np.linspace(0,1,21), desc_weights_gbl]
results = classification.classify(atep_fv, \
class_labels, traintest_parts, params, \
xml_config['features_list'], \
C=C_gbl,
strategy=strategy_gbl,
opt_criterion=opt_criterion,
verbose=args.verbose)
params = [[[1]], [1], np.linspace(0, 1, 21), desc_weights_gbl]
results = classification.classify(atep_bovw, \
class_labels, traintest_parts, params, \
xml_config['features_list'], \
C=C_gbl,
strategy=strategy_gbl,
opt_criterion=opt_criterion,
verbose=args.verbose)
classification.print_results(results)
if 'atep-fv' in args.methods:
tracklet_representation.train_fv_gmms(tracklets_path,
videonames,
traintest_parts,
xml_config['features_list'],
intermediates_path,
pca_reduction=True,
nt=args.nt,
verbose=args.verbose)
tracklet_representation.compute_fv_descriptors_multithread(tracklets_path, intermediates_path, videonames, traintest_parts, xml_config['features_list'], \
feats_path + '/fvtree/', \
treelike=True, pca_reduction=True, clusters_path=clusters_path, nt=args.nt, verbose=args.verbose)
atep_fv = kernels.compute_ATEP_kernels(feats_path + '/fvtree/', videonames, traintest_parts, xml_config['features_list'], \
kernels_path + '/atep-fv/', use_disk=False, nt=args.nt, verbose=args.verbose)
params = [[[1]], [1], np.linspace(0, 1, 21), desc_weights_gbl]
results = classification.classify(atep_fv, \
class_labels, traintest_parts, params, \
xml_config['features_list'], \
C=C_gbl,
tracklet_clustering.cluster(tracklets_path,
videonames,
INSTANCE_ST,
INSTANCE_TOTAL,
clusters_path,
visualize=False)
tracklet_representation.train_bovw_codebooks(
tracklets_path,
videonames,
traintest_parts,
INTERNAL_PARAMETERS['feature_types'],
intermediates_path,
pca_reduction=False)
tracklet_representation.train_fv_gmms(tracklets_path, videonames,
traintest_parts,
INTERNAL_PARAMETERS['feature_types'],
intermediates_path)
tracklet_representation.compute_bovw_descriptors(tracklets_path, intermediates_path, videonames, traintest_parts, \
INSTANCE_ST, INSTANCE_TOTAL, \
INTERNAL_PARAMETERS['feature_types'], feats_path + 'bovwtree/', \
pca_reduction=False, treelike=True, global_repr=True, clusters_path=clusters_path)
tracklet_representation.compute_fv_descriptors(tracklets_path, intermediates_path, videonames, traintest_parts, \
INSTANCE_ST, INSTANCE_TOTAL, \
INTERNAL_PARAMETERS['feature_types'], feats_path + 'fvtree/', \
treelike=True, global_repr=True, clusters_path=clusters_path)
c = [
0.001, 0.01, 0.05, 0.1, 0.5, 1, 5, 10, 50, 100, 500, 1000, 5000, 10000
]
elif len(sys.argv) < 4:
INSTANCE_ST = int(sys.argv[1])
if INSTANCE_ST > len(videonames) - 1:
INSTANCE_ST = len(videonames) - 1
INSTANCE_TOTAL = int(sys.argv[2])
if INSTANCE_ST + int(sys.argv[2]) > len(videonames):
INSTANCE_TOTAL = len(videonames) - INSTANCE_ST
print('INSTANCE_ST: %d, INSTANCE_TOTAL: %d' % (INSTANCE_ST, INSTANCE_TOTAL))
tracklet_extraction.extract(fullvideonames, videonames, INSTANCE_ST, INSTANCE_TOTAL, INTERNAL_PARAMETERS['feature_types'], tracklets_path)
tracklet_clustering.cluster(tracklets_path, videonames, INSTANCE_ST, INSTANCE_TOTAL, clusters_path, visualize=False)
tracklet_representation.train_bovw_codebooks(tracklets_path, videonames, traintest_parts, INTERNAL_PARAMETERS['feature_types'], intermediates_path, pca_reduction=False)
tracklet_representation.train_fv_gmms(tracklets_path, videonames, traintest_parts, INTERNAL_PARAMETERS['feature_types'], intermediates_path)
tracklet_representation.compute_bovw_descriptors(tracklets_path, intermediates_path, videonames, traintest_parts, \
INSTANCE_ST, INSTANCE_TOTAL, \
INTERNAL_PARAMETERS['feature_types'], feats_path + 'bovwtree/', \
pca_reduction=False, treelike=True, global_repr=True, clusters_path=clusters_path)
tracklet_representation.compute_fv_descriptors(tracklets_path, intermediates_path, videonames, traintest_parts, \
INSTANCE_ST, INSTANCE_TOTAL, \
INTERNAL_PARAMETERS['feature_types'], feats_path + 'fvtree/', \
treelike=True, global_repr=True, clusters_path=clusters_path)
c = [0.001, 0.01, 0.05, 0.1, 0.5, 1, 5, 10, 50, 100, 500, 1000, 5000, 10000]
st_time = time.time()
results = classification.classify(feats_path + 'bovwtree/', videonames, class_labels, traintest_parts, \
np.linspace(0, 1, 11), INTERNAL_PARAMETERS['feature_types'], \
