def resume_segmentation(iterations=10):
logger.debug('Resume segmentation')
corpus = Corpus(Q=opt.gmm, subaction=opt.subaction)
for iteration in range(iterations):
logger.debug('Iteration %d' % iteration)
corpus.iter = iteration
corpus.resume_segmentation()
corpus.accuracy_corpus()
corpus.accuracy_corpus()
def baseline(iterations=7):
"""Implementation of the paper"""
corpus = Corpus(Q=opt.gmm, subaction=opt.subaction)
for iteration in range(iterations):
logger.debug('Iteration %d' % iteration)
corpus.iter = iteration
corpus.accuracy_corpus()
if (opt.gt_training and iteration == 0) or not opt.gt_training:
corpus.embedding_training()
# one version of gaussian mixtures for the entire dataset
if opt.gmms == 'one':
corpus.one_gaussian_model()
# different gmm for different subsets of videos, i.e. leave one out for
# each video subset
elif opt.gmms == 'many':
corpus.many_gaussian_models()
# with multiprocessing package
# corpus.gaussians_mp(n_threads=3)
else:
raise RuntimeError('define number of gmms for the video collection')
if opt.viterbi:
# corpus.viterbi_decoding()
# corpus.accuracy_corpus(prefix='pure vit ')
# corpus.viterbi_ordering()
# take into account Mallow Model
corpus.ordering_sampler()
corpus.rho_sampling()
# corpus.accuracy_corpus(prefix='vit+ord ')
corpus.viterbi_decoding()
# corpus.viterbi_alex_decoding()
else:
corpus.subactivity_sampler()
# take into account Mallow Model
corpus.ordering_sampler()
corpus.rho_sampling()
logger.debug('Iteration %d' % iteration)
corpus.accuracy_corpus()
def temp_embed():
corpus = Corpus(subaction=opt.subaction)
logger.debug('Corpus with poses created')
if opt.model_name in ['mlp']:
corpus.regression_training()
if opt.model_name == 'nothing':
corpus.without_temp_emed()
corpus.clustering()
corpus.gaussian_model()
corpus.accuracy_corpus()
if opt.resume_segmentation:
corpus.resume_segmentation()
else:
corpus.viterbi_decoding()
corpus.accuracy_corpus('final')
return corpus.return_stat
