def predict_input_fn():
source = tf.data.TFRecordDataset(list(test_source_files))
target = tf.data.TFRecordDataset(list(test_target_files))
dataset = dataset_factory(source, target, hparams)
batched = dataset.prepare_and_zip().group_by_batch(
batch_size=1).move_mel_to_source()
return batched.dataset
def eval_input_fn():
source_and_target_files = list(zip(eval_source_files, eval_target_files))
shuffle(source_and_target_files)
source = tf.data.TFRecordDataset([s for s, _ in source_and_target_files])
target = tf.data.TFRecordDataset([t for _, t in source_and_target_files])
dataset = dataset_factory(source, target, hparams)
zipped = dataset.prepare_and_zip()
dataset = zipped.filter_by_max_output_length().repeat().group_by_batch(batch_size=1)
return dataset.dataset
def _config_ds_iterator(self):
with tf.device('/cpu:0'):
# split dataset by worker
data_sources = dataset_utils.get_tfrecord_files(
self._num_workers, FLAGS.file_pattern)
# select the preprocessing func
dataset_obj_initializer = dataset_factory(
FLAGS.dataset_name).get_dataset_obj()
self._dataset_obj = dataset_obj_initializer(
data_sources, None, None)
self._ds_iterator = self._dataset_obj.get_dataset()
def _get_dataset(tfrecords_dir='tfrecords/', preprocessing_type='caffe', dataset_class='pascalvoc', coco_year='2017',
pascalvoc_year='2007', pascalvoc_class='trainval', num_pascalvoc_tfrecords=2,
data_root_path=None):
if dataset_class == 'pascalvoc':
file_pattern = 'pascalvoc_{}_{}_%02d.tfrecord'.format(pascalvoc_year, pascalvoc_class)
file_names = [os.path.join(tfrecords_dir, file_pattern % i) for i in range(num_pascalvoc_tfrecords)]
# print(file_names)
dataset_configs = { 'tf_record_list': file_names,
'min_size': configs['image_min_size'],
'max_size': configs['image_max_size'],
'preprocessing_type': configs['preprocessing_type'],
'caffe_pixel_means': configs['bgr_pixel_means'],
'data_argumentation': True
}
# def dataset_factory(dataset_class, mode, configs):
dataset = dataset_factory('pascalvoc', 'trainval', dataset_configs)
return dataset
