def extract_low_level_features(self):
""" docstring """
start = get_time()
np.random.seed(7)
sd = int(np.random.rand(1, 1)[0, 0] * 1e9)
self.n_jobs = N_JOBS
fnames_a = self.data_a.metainfo['all_fnames']
tasks = []
for fname in fnames_a:
for trial in self.llf_trials_a:
path = 'low_level_features/{0}/{1}/{2}'.format(
trial["NTV"], rois_available[trial["LGF"]],
measures_available[trial["M"]])
output_path = os.path.join(self.data_a.output_path, path)
tasks += [
LowLevelFeatures(self.data_a.dataset_path,
output_path,
fname,
trial["LGF"],
trial["M"],
n_cuboids=trial["NTV"],
seed=sd)
]
print "running %d tasks in parallel" % len(tasks)
sys.stdout.flush()
RunInParallel(tasks, self.n_jobs).run()
fnames_b = self.data_b.metainfo['all_fnames']
tasks = []
for fname in fnames_b:
for trial in self.llf_trials_b:
path = 'low_level_features/{0}/{1}/{2}'.format(
trial["NTV"], rois_available[trial["LGF"]],
measures_available[trial["M"]])
output_path = os.path.join(self.data_b.output_path, path)
tasks += [
LowLevelFeatures(self.data_b.dataset_path,
output_path,
fname,
trial["LGF"],
trial["M"],
n_cuboids=trial["NTV"],
seed=sd)
]
print "running %d tasks in parallel" % len(tasks)
RunInParallel(tasks, self.n_jobs).run()
total_time_elapsed(start, get_time())
def extract_low_level_features(self):
""" docstring """
start = get_time()
np.random.seed(7)
if self.data.__class__.__name__.lower() == 'maskattack':
fnames = self.data.metainfo['all_fnames']
else:
print 'This Protocol is not defined for this dataset!'
sys.exit(0)
tasks = []
sd = int(np.random.rand(1, 1)[0, 0] * 1e9)
for fname in fnames:
for trial in self.llf_trials:
path = 'low_level_features/{0}/{1}/{2}'.format(trial["NTV"],
rois_available[trial["LGF"]],
measures_available[trial["M"]])
output_path = os.path.join(self.data.output_path, path)
tasks += [LowLevelFeatures(self.data.dataset_path, output_path, fname,
trial["LGF"],
trial["M"],
n_cuboids=trial["NTV"],
frame_numbers=self.frame_numbers,
seed=sd)]
print "running %d tasks in parallel" % len(tasks)
RunInParallel(tasks, self.n_jobs).run()
total_time_elapsed(start, get_time())
def extract_low_level_features(self):
""" docstring """
start = get_time()
np.random.seed(7)
fnames = self.data.metainfo['all_fnames']
# if self.data.__class__.__name__.lower() == 'replayattack':
# # fnames = [fname for fname in self.data.metainfo['all_fnames']
# # if ('train/' in fname)or('devel/' in fname)or('test/' in fname)]
# elif self.data.__class__.__name__.lower() == 'casia':
# fnames = self.data.metainfo['all_fnames']
# elif self.data.__class__.__name__.lower() == 'maskattack':
# fnames = self.data.metainfo['all_fnames']
# elif self.data.__class__.__name__.lower() == 'uvad':
# fnames = self.data.metainfo['all_fnames']
# # # search substrings
# # fnames_camera = fnames[np.array(["attack/nikon" in s
# # for s in fnames.flat]).reshape(fnames.shape)]
# # fnames = [fname for fname in fnames_camera if ('monitor7' in fname)]
# else:
# pass
tasks = []
sd = int(np.random.rand(1, 1)[0, 0] * 1e9)
for fname in fnames:
for trial in self.llf_trials:
path = 'low_level_features/{0}/{1}/{2}'.format(trial["NTV"],
rois_available[trial["LGF"]],
measures_available[trial["M"]])
output_path = os.path.join(self.data.output_path, path)
tasks += [LowLevelFeatures(self.data.dataset_path, output_path, fname,
trial["LGF"],
trial["M"],
n_cuboids=trial["NTV"],
only_face=self.only_face,
frame_numbers=self.frame_numbers,
seed=sd,
get_faceloc=self.data.get_faceloc)]
if self.n_jobs > 1:
print "running %d tasks in parallel" % len(tasks)
RunInParallel(tasks, self.n_jobs).run()
else:
print "running %d tasks in sequence" % len(tasks)
for idx in range(len(fnames)):
tasks[idx].run()
progressbar('-- RunInSequence', idx, len(fnames))
elapsed = total_time_elapsed(start, get_time())
print 'spent time: {0}!'.format(elapsed)
sys.stdout.flush()
def extract_low_level_features(self):
""" docstring """
start = get_time()
np.random.seed(7)
fnames = []
if self.data.__class__.__name__.lower() == 'replayattack':
fnames = self.data.metainfo['all_fnames']
fnames = [
fname for fname in fnames if (self.sample in fname) and (
('hand/' in fname) or ('real' in fname))
]
elif self.data.__class__.__name__.lower() == 'casia':
fnames = self.data.metainfo['all_fnames']
elif self.data.__class__.__name__.lower() == 'maskattack':
fnames = self.data.metainfo['all_fnames']
elif self.data.__class__.__name__.lower() == 'uvad':
fnames = self.data.metainfo['all_fnames']
else:
pass
for realization in [1]:
sd = int(np.random.rand(1, 1)[0, 0] * 1e9)
for fname in fnames:
for trial in self.llf_trials:
path = 'realization_{0}/low_level_features/{1}/{2}/{3}'.format(
realization, trial["NTV"],
rois_available[trial["LGF"]],
measures_available[trial["M"]])
output_path = os.path.join(self.data.output_path, path)
LowLevelFeatures(self.data.dataset_path,
output_path,
fname,
trial["LGF"],
trial["M"],
n_cuboids=trial["NTV"],
analize=True,
frame_numbers=10,
seed=sd).analize_one_sample()
total_time_elapsed(start, get_time())
def extract_low_level_features(self):
""" docstring """
start = get_time()
np.random.seed(7)
fnames = self.data.metainfo['all_fnames']
tasks = []
sd = int(np.random.rand(1, 1)[0, 0] * 1e9)
for fname in fnames:
for trial in self.llf_trials:
path = 'low_level_features/{0}/{1}/{2}'.format(
trial["NTV"], rois_available[trial["LGF"]],
measures_available[trial["M"]])
output_path = os.path.join(self.data.output_path, path)
tasks += [
LowLevelFeatures(self.data.dataset_path,
output_path,
fname,
trial["LGF"],
trial["M"],
n_cuboids=trial["NTV"],
only_face=self.only_face,
frame_numbers=self.frame_numbers,
seed=sd,
get_faceloc=self.data.get_faceloc)
]
if self.n_jobs > 1:
print "running %d tasks in parallel" % len(tasks)
RunInParallel(tasks, self.n_jobs).run()
else:
print "running %d tasks in sequence" % len(tasks)
for idx in range(len(fnames)):
tasks[idx].run()
progressbar('-- RunInSequence', idx, len(fnames))
elapsed = total_time_elapsed(start, get_time())
print 'spent time: {0}!'.format(elapsed)
sys.stdout.flush()
def extract_low_level_features(self):
""" docstring """
start = get_time()
np.random.seed(7)
seeds = []
for _ in self.realizations:
seeds += [int(np.random.rand(1, 1)[0, 0] * 1e9)]
fnames = self.data.metainfo['all_fnames']
tasks = []
for realization, sd in zip(self.realizations, seeds):
print "realization, sd", realization, sd
for fname in fnames:
for trial in self.llf_trials:
path = 'realization_{0}/low_level_features/{1}/{2}/{3}'.format(
realization, trial["NTV"],
rois_available[trial["LGF"]],
measures_available[trial["M"]])
output_path = os.path.join(self.data.output_path, path)
tasks += [
LowLevelFeatures(self.data.dataset_path,
output_path,
fname,
trial["LGF"],
trial["M"],
n_cuboids=trial["NTV"],
seed=sd)
]
print "running %d tasks in parallel" % len(tasks)
RunInParallel(tasks, self.n_jobs).run()
total_time_elapsed(start, get_time())
def extract_low_level_features(self):
start = get_time()
np.random.seed(7)
sd = int(np.random.rand(1, 1)[0, 0] * 1e9)
self.n_jobs = N_JOBS
# if self.data_a.__class__.__name__.lower() == 'replayattack':
# # fnames = [fname for fname in self.data_a.metainfo['all_fnames']
# # if ('train/' in fname)or('devel/' in fname)or('test/' in fname)]
# fnames_a = self.data_a.metainfo['all_fnames']
# elif self.data_a.__class__.__name__.lower() == 'casia':
# fnames = self.data_a.metainfo['all_fnames']
# elif self.data_a.__class__.__name__.lower() == 'maskattack':
# fnames = self.data_a.metainfo['all_fnames']
# elif self.data_a.__class__.__name__.lower() == 'uvad':
# fnames = self.data_a.metainfo['all_fnames']
# # # search substrings
# # fnames_camera = fnames[np.array(["attack/nikon" in s
# # for s in fnames.flat]).reshape(fnames.shape)]
# # fnames = [fname for fname in fnames_camera if ('monitor7' in fname)]
# else:
# pass
# if self.data_b.__class__.__name__.lower() == 'replayattack':
# # fnames = [fname for fname in self.data_b.metainfo['all_fnames']
# # if ('train/' in fname)or('devel/' in fname)or('test/' in fname)]
# fnames_a = self.data_b.metainfo['all_fnames']
# elif self.data_b.__class__.__name__.lower() == 'casia':
# fnames = self.data_b.metainfo['all_fnames']
# elif self.data_b.__class__.__name__.lower() == 'maskattack':
# fnames = self.data_b.metainfo['all_fnames']
# elif self.data_b.__class__.__name__.lower() == 'uvad':
# fnames = self.data_b.metainfo['all_fnames']
# # # search substrings
# # fnames_camera = fnames[np.array(["attack/nikon" in s
# # for s in fnames.flat]).reshape(fnames.shape)]
# # fnames = [fname for fname in fnames_camera if ('monitor7' in fname)]
# else:
# pass
fnames_a = self.data_a.metainfo['all_fnames']
tasks = []
for fname in fnames_a:
for trial in self.llf_trials_a:
path = 'low_level_features/{0}/{1}/{2}'.format(
trial["NTV"], rois_available[trial["LGF"]],
measures_available[trial["M"]])
output_path = os.path.join(self.data_a.output_path, path)
tasks += [
LowLevelFeatures(
self.data_a.dataset_path,
output_path,
fname,
trial["LGF"],
trial["M"],
n_cuboids=trial["NTV"],
seed=sd,
)
]
print "running %d tasks in parallel" % len(tasks)
sys.stdout.flush()
RunInParallel(tasks, self.n_jobs).run()
fnames_b = self.data_b.metainfo['all_fnames']
tasks = []
for fname in fnames_b:
for trial in self.llf_trials_b:
path = 'low_level_features/{0}/{1}/{2}'.format(
trial["NTV"], rois_available[trial["LGF"]],
measures_available[trial["M"]])
output_path = os.path.join(self.data_b.output_path, path)
tasks += [
LowLevelFeatures(self.data_b.dataset_path,
output_path,
fname,
trial["LGF"],
trial["M"],
n_cuboids=trial["NTV"],
seed=sd)
]
print "running %d tasks in parallel" % len(tasks)
RunInParallel(tasks, self.n_jobs).run()
total_time_elapsed(start, get_time())