def __call__(self, params):
"""Input function which provides a single batch for train or eval."""
# Retrieves the batch size for the current shard. The # of shards is
# computed according to the input pipeline deployment. See
# `tf.contrib.tpu.RunConfig` for details.
batch_size = params['batch_size']
logging.info('call ToyModelInput() with batch size {}'.format(batch_size))
ds = Dataset.from_tensor_slices((self._images, self._labels)).repeat()
dataset = ds.batch(batch_size, drop_remainder=True).prefetch(2)
return dataset
def _get_dataset(self, x, y, name_scope='data'):
'''
get dataset
:param name_scope:
:param x:
:param y:
:return: train_init_op, train_next_op
'''
batch_size = self._batch_size
with tf.name_scope(name=name_scope):
dataset = Dataset.from_tensor_slices((x, y))
dataset = dataset.shuffle(buffer_size=batch_size * 10) \
.repeat() \
.map(map_func=map_fn, num_parallel_calls=8) \
.batch(batch_size=batch_size) \
.prefetch(buffer_size=batch_size * 10)
init_op = dataset.make_initializable_iterator(shared_name='init_op')
next_op = init_op.get_next(name='next_op')
return init_op, next_op
def __init__(self,
mode,
batch_size,
num_classes,
shuffle=True,
buffer_size=1000):
"""Create a new ImageDataGenerator.
Recieves a path string to a text file, which consists of many lines,
where each line has first a path string to an image and seperated by
a space an integer, referring to the class number. Using this data,
this class will create TensrFlow datasets, that can be used to train
e.g. a convolutional neural network.
Args:
mode: Either 'training' or 'inference'. Depending on this value,
different parsing functions will be used.
batch_size: Number of images per batch.
num_classes: Number of classes in the dataset.
shuffle: Wether or not to shuffle the data in the dataset and the
initial file list.
buffer_size: Number of images used as buffer for TensorFlows
shuffling of the dataset.
Raises:
ValueError: If an invalid mode is passed.
"""
self.num_classes = num_classes
self.duplicated_tag_list = []
fp = open("/media/jiwunghyun/DATA/nus-wide/Concepts81.txt", 'r')
tag_list = []
for i in range(81):
a = fp.readline().split('\n')[0]
tag_list.append(a)
fp.close()
fp2 = open(
"/media/jiwunghyun/DATA/nus-wide/NUS_WID_Tags/TagList1k.txt", 'r')
tag1000_list = []
for i in range(1000):
a = fp2.readline().split('\n')[0]
tag1000_list.append(a)
fp2.close()
for i in range(len(tag1000_list)):
for j in range(len(tag_list)):
if tag1000_list[i] == tag_list[j]:
self.duplicated_tag_list.append(i)
# retrieve the data from the text file
if mode == 'training':
self._read_train_txt_file()
elif mode == 'inference':
self._read_test_txt_file()
# number of samples in the dataset
self.data_size = len(self.labels)
# initial shuffling of the file and label lists (together!)
if shuffle:
self._shuffle_lists()
# convert lists to TF tensor
self.img_paths = convert_to_tensor(self.img_paths, dtype=dtypes.string)
self.labels = convert_to_tensor(self.labels, dtype=dtypes.int32)
self.noisy_tags = convert_to_tensor(self.noisy_tags,
dtype=dtypes.int32)
# create dataset
data = Dataset.from_tensor_slices(
(self.img_paths, self.labels, self.noisy_tags))
# distinguish between train/infer. when calling the parsing functions
if mode == 'training':
data = data.map(self._parse_function_train)
elif mode == 'inference':
data = data.map(self._parse_function_inference)
else:
raise ValueError("Invalid mode '%s'." % (mode))
# shuffle the first `buffer_size` elements of the dataset
if shuffle:
data = data.shuffle(buffer_size=buffer_size)
# create a new dataset with batches of images
data = data.batch(batch_size)
self.data = data