def __init__(self, scenes_json, questions_json, image_root, image_transform, vocab_json, question_transform=None, incl_scene=True):
super().__init__()
self.scenes_json = scenes_json
self.questions_json = questions_json
self.image_root = image_root
self.image_transform = image_transform
self.vocab_json = vocab_json
self.question_transform = question_transform
self.incl_scene = incl_scene
logger.info('Loading scenes from: "{}".'.format(self.scenes_json))
self.scenes = io.load_json(self.scenes_json)['scenes']
if isinstance(self.questions_json, (tuple, list)):
self.questions = list()
for filename in self.questions_json:
logger.info('Loading questions from: "{}".'.format(filename))
self.questions.extend(io.load_json(filename)['questions'])
else:
logger.info('Loading questions from: "{}".'.format(self.questions_json))
self.questions = io.load_json(self.questions_json)['questions']
if self.vocab_json is not None:
logger.info('Loading vocab from: "{}".'.format(self.vocab_json))
self.vocab = Vocab.from_json(self.vocab_json)
else:
logger.info('Building the vocab.')
self.vocab = gen_vocab(self)
def main():
scenes = io.load_json(args.scene_json)['scenes']
preds = io.load(args.preds_json)
if isinstance(preds, dict):
preds = list(preds.values())
if False:
preds = [transpose_scene(s) for s in preds]
# flattened_objs = [o for s in scenes for o in s['objects']]
# flattened_preds = {
# k: np.concatenate([np.array(p[k]) for p in preds], axis=0)
# for k in preds[0]
# }
meter = GroupMeters()
'''
for i, scene in tqdm_gofor(scenes, mininterval=0.5):
for j in range(len(scene['objects'])):
test(j, scene['objects'], preds[i], meter)
'''
for i, pred in tqdm_gofor(preds, mininterval=0.5):
scene = scenes[i]
for j in range(len(scene['objects'])):
test(j, scene['objects'], pred, meter)
print(meter.format_simple('Results:', compressed=False))
def __init__(self,
scenes_json,
features_h5=None,
image_root=None,
image_transform=None,
question_per_image=20,
balance_attribute=True,
balance_answer=True):
super().__init__()
self.scenes_json = scenes_json
self.features_h5 = features_h5
self.image_root = image_root
self.image_transform = image_transform
self.question_per_image = question_per_image
self.balance_attribute = balance_attribute
self.balance_answer = balance_answer
self.scenes = io.load_json(self.scenes_json)['anns']
if self.features_h5 is not None:
self.features = io.open_h5(features_h5, 'r')['features']
assert len(self.features) == len(self.scenes)
# TODO(Jiayuan Mao @ 04/10): no vocab for CCG models.
self.vocab = None
self._gen_data()
def main():
scenes = io.load_json(args.scene_json)['scenes']
preds = io.load(args.preds_json)
if isinstance(preds, dict):
preds = list(preds.values())
if False:
preds = [transpose_scene(s) for s in preds]
meter = GroupMeters()
flattened_objs = [o for s in scenes for o in s['objects']]
flattened_preds = {
k: np.concatenate([np.array(p[k]) for p in preds], axis=0)
for k in preds[0]
}
for k, preds in flattened_preds.items():
kk = def_.word2lemma.get(k, k)
for i, o in tqdm_gofor(flattened_objs,
desc='{}(lemma: {})'.format(k, kk),
leave=False):
meter.update(
'acc', (preds[i] > 0) == (kk == o[def_.concept2attribute[kk]]))
meter.update(
f'acc/{k}',
(preds[i] > 0) == (kk == o[def_.concept2attribute[kk]]))
print(meter.format_simple('Results:', compressed=False))
def __init__(
self,
scenes_json,
questions_json,
image_root,
image_transform,
depth_root,
vocab_json,
question_transform=None,
incl_scene=True,
):
super().__init__()
self.scenes_json = scenes_json
self.questions_json = questions_json
self.image_root = image_root
self.image_transform = image_transform
self.depth_root = depth_root
self.vocab_json = vocab_json
self.question_transform = question_transform
self.incl_scene = incl_scene
logger.info('Loading scenes from: "{}".'.format(self.scenes_json))
self.scenes = io.load_json(self.scenes_json)["scenes"]
if isinstance(self.questions_json, (tuple, list)):
self.questions = list()
for filename in self.questions_json:
logger.info('Loading questions from: "{}".'.format(filename))
self.questions.extend(io.load_json(filename)["questions"])
else:
logger.info('Loading questions from: "{}".'.format(self.questions_json))
self.questions = io.load_json(self.questions_json)["questions"]
if self.vocab_json is not None:
logger.info('Loading vocab from: "{}".'.format(self.vocab_json))
self.vocab = Vocab.from_json(self.vocab_json)
else:
logger.info("Building the vocab.")
self.vocab = gen_vocab(self)
vocab_file = questions_json.split("/")
vocab_file[-1] = "vocab.json"
vocab_file = "/".join(vocab_file)
self.vocab.dump_json(vocab_file)
def __init__(self, scenes_json, image_root, image_transform, incl_scene=True):
super().__init__()
self.scenes_json = scenes_json
self.image_root = image_root
self.image_transform = image_transform
self.incl_scene = incl_scene
logger.info('Loading scenes from: "{}".'.format(self.scenes_json))
self.scenes = io.load_json(self.scenes_json)['scenes']
def load_files(self):
print('Loading captions and precomputed image embeddings')
self.image_embeddings = load(self.image_embeddings)
self.captions = load_json(self.captions)
assert len(self.captions) == len(self.image_embeddings)
# self.captions = [self.captions[i] for i in non_empty_inds]
# self.image_embeddings = self.image_embeddings[non_empty_inds]
if self.mode is CompletionDatasetMode.SAMPLE:
self.non_empty_inds = [
i for i, c in enumerate(self.captions) if len(c['replace']) > 0
]
else:
if self.mode is CompletionDatasetMode.ALL:
replace = lambda c: c['replace']
elif self.mode is CompletionDatasetMode.NOUN:
replace = lambda c: c['replace_noun']
elif self.mode is CompletionDatasetMode.PREP:
replace = lambda c: c['replace_prep']
self.all_inds = [(i, r) for i, c in enumerate(self.captions)
for r in replace(c)]
def main():
scenes = io.load_json(args.scene_json)['scenes']
preds = io.load(args.preds_json)
if isinstance(preds, dict):
preds = list(preds.values())
if False:
preds = [transpose_scene(s) for s in preds]
scenes = scenes[:1000]
preds = preds[:1000]
flattened_objs = [o for s in scenes for o in s['objects']]
flattened_preds = {
k: np.concatenate([np.array(p[k]) for p in preds], axis=0)
for k in preds[0]
}
meter = GroupMeters()
for i, obj in tqdm_gofor(flattened_objs, mininterval=0.5):
test(i, flattened_objs, flattened_preds, meter)
print(meter.format_simple('Results:', compressed=False))
def __init__(self, scenes_json, image_root, depth_root, incl_scene=True):
super().__init__()
self.scenes_json = scenes_json
self.image_root = image_root
self.depth_root = depth_root
self.incl_scene = incl_scene
logger.info('Loading scenes from: "{}".'.format(self.scenes_json))
self.scenes = io.load_json(self.scenes_json)["scenes"]
import jactorch.transforms.bbox as T
self.image_transform = T.Compose(
[
# T.NormalizeBbox(),
# T.Resize(configs.data.image_size),
# T.DenormalizeBbox(),
T.ToTensor(),
T.Lambda(lambda x, y: (x - 0.5, y))
# T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
]
)
def check_json_consistency(self, json_file):
rhs = io.load_json(json_file)
for k, v in self.word2idx.items():
if not (k in rhs and rhs[k] == v):
return False
return True
def from_json(cls, json_file):
return cls(io.load_json(json_file))
