def _setup_data(self, opt):
# Load map from image ID to gender
data = self._load_gender_data(opt)
extra_data = []
if self.add_unknown_classes:
for ex in data:
self_ex = deepcopy(ex)
self_ex['label'] = f'SELF:{gend_utils.UNKNOWN}'
self_ex['class_type'] = 'self'
extra_data.append(self_ex)
partner_ex = deepcopy(ex)
partner_ex['label'] = f'PARTNER:{gend_utils.UNKNOWN}'
partner_ex['class_type'] = 'partner'
extra_data.append(partner_ex)
if len(extra_data) > 0:
# possibly sample unknown classes
sample_rate = self.opt['unknown_temp']
if sample_rate < 1.0:
to_samp = int(sample_rate * len(extra_data))
sampled = random.sample(extra_data, to_samp)
data += sampled
else:
data += extra_data
if self.is_train:
random.shuffle(data)
gend_utils.get_data_stats(data, key='label', lst=False)
return data
def _setup_data(self, opt):
# Load map from image ID to gender
data = self._load_gender_data(opt)
# Possibly add extra examples
extra_data = []
if self.add_unknown_classes:
for ex in data:
# add self examples
self_ex = deepcopy(ex)
self_ex['class_type'] = 'self'
# not True neutral, so we flip between
self_ex['labels'] = gend_utils.UNKNOWN_LABELS['self']
extra_data.append(self_ex)
# add partner examples
partner_ex = deepcopy(ex)
partner_ex['labels'] = gend_utils.UNKNOWN_LABELS['partner']
partner_ex['class_type'] = 'partner'
extra_data.append(partner_ex)
# now sample the data
sample_rate = self.opt['unknown_temp']
if sample_rate < 1.0:
to_samp = int(sample_rate * len(extra_data))
sampled = random.sample(extra_data, to_samp)
data += sampled
else:
data += extra_data
if self.is_train:
random.shuffle(data)
gend_utils.get_data_stats(data, key='labels')
return data
def _setup_data(self, opt):
# Load map from image ID to gender
# TODO: proper train/test/valid splits
data = []
convos = self._get_convos(opt)
for convo in convos:
class_type = convo['class_type']
if self.labels_to_use in ['all', class_type]:
data.append(convo)
if self.is_train:
random.shuffle(data)
gend_utils.get_data_stats(data, key='labels')
return data
