def create_sequence_dataset_input_fn(dataset_dir, prepare_fn,
record_file_name_format,
meta_data_file_name_format,
bucket_boundaries):
prepare_fn(dataset_dir)
train_data_file = record_file_name_format.format(dataset_dir, Modes.TRAIN)
eval_data_file = record_file_name_format.format(dataset_dir, Modes.EVAL)
meta_data_filename = meta_data_file_name_format.format(dataset_dir)
train_input_fn = create_input_data_fn(
mode=Modes.TRAIN,
pipeline_config=PipelineConfig(module='TFRecordSequencePipeline',
dynamic_pad=True,
bucket_boundaries=bucket_boundaries,
params={
'data_files': train_data_file,
'meta_data_file': meta_data_filename
}))
eval_input_fn = create_input_data_fn(
mode=Modes.EVAL,
pipeline_config=PipelineConfig(module='TFRecordSequencePipeline',
dynamic_pad=True,
bucket_boundaries=bucket_boundaries,
params={
'data_files': eval_data_file,
'meta_data_file': meta_data_filename
}))
return train_input_fn, eval_input_fn
def create_input_fn(dataset_dir):
prepare(dataset_dir)
train_data_file = RECORD_FILE_NAME_FORMAT.format(dataset_dir,
ModeKeys.TRAIN)
eval_data_file = RECORD_FILE_NAME_FORMAT.format(dataset_dir, ModeKeys.EVAL)
meta_data_filename = MEAT_DATA_FILENAME_FORMAT.format(dataset_dir)
train_input_fn = create_input_data_fn(mode=ModeKeys.TRAIN,
pipeline_config=PipelineConfig(
name='TFRecordImagePipeline',
dynamic_pad=False,
params={
'data_files':
train_data_file,
'meta_data_file':
meta_data_filename
}))
eval_input_fn = create_input_data_fn(mode=ModeKeys.EVAL,
pipeline_config=PipelineConfig(
name='TFRecordImagePipeline',
dynamic_pad=False,
params={
'data_files':
eval_data_file,
'meta_data_file':
meta_data_filename
}))
return train_input_fn, eval_input_fn
def create_image_dataset_input_fn(dataset_dir, prepare_fn,
record_file_name_format,
meta_data_file_name_format):
prepare_fn(dataset_dir)
train_data_file = record_file_name_format.format(dataset_dir, Modes.TRAIN)
eval_data_file = record_file_name_format.format(dataset_dir, Modes.EVAL)
meta_data_filename = meta_data_file_name_format.format(dataset_dir)
def get_pipeline_config(mode):
return TFRecordImagePipelineConfig(
dynamic_pad=False,
data_files=train_data_file
if Modes.is_train(mode) else eval_data_file,
meta_data_file=meta_data_filename,
feature_processors=FeatureProcessorsConfig({
'image':
GraphConfig(input_layers=[['image', 0, 0]],
output_layers=[['image_out', 0, 0]],
layers=[
CastConfig(dtype='float32',
name='image_out',
inbound_nodes=[['image', 0, 0]])
])
}))
train_input_fn = create_input_data_fn(mode=Modes.TRAIN,
pipeline_config=get_pipeline_config(
Modes.TRAIN))
eval_input_fn = create_input_data_fn(mode=Modes.EVAL,
pipeline_config=get_pipeline_config(
Modes.EVAL))
return train_input_fn, eval_input_fn
def create_image_dataset_input_fn(dataset_dir, prepare_fn,
record_file_name_format,
meta_data_file_name_format):
prepare_fn(dataset_dir)
train_data_file = record_file_name_format.format(dataset_dir, Modes.TRAIN)
eval_data_file = record_file_name_format.format(dataset_dir, Modes.EVAL)
meta_data_filename = meta_data_file_name_format.format(dataset_dir)
train_input_fn = create_input_data_fn(mode=Modes.TRAIN,
pipeline_config=PipelineConfig(
module='TFRecordImagePipeline',
dynamic_pad=False,
params={
'data_files':
train_data_file,
'meta_data_file':
meta_data_filename
}))
eval_input_fn = create_input_data_fn(mode=Modes.EVAL,
pipeline_config=PipelineConfig(
module='TFRecordImagePipeline',
dynamic_pad=False,
params={
'data_files':
eval_data_file,
'meta_data_file':
meta_data_filename
}))
return train_input_fn, eval_input_fn
def create_image_dataset_input_fn(dataset_dir, prepare_fn, record_file_name_format,
meta_data_file_name_format):
prepare_fn(dataset_dir)
train_data_file = record_file_name_format.format(dataset_dir, Modes.TRAIN)
eval_data_file = record_file_name_format.format(dataset_dir, Modes.EVAL)
meta_data_filename = meta_data_file_name_format.format(dataset_dir)
train_input_fn = create_input_data_fn(
mode=Modes.TRAIN,
pipeline_config=PipelineConfig(module='TFRecordImagePipeline', dynamic_pad=False,
params={'data_files': train_data_file,
'meta_data_file': meta_data_filename})
)
eval_input_fn = create_input_data_fn(
mode=Modes.EVAL,
pipeline_config=PipelineConfig(module='TFRecordImagePipeline', dynamic_pad=False,
params={'data_files': eval_data_file,
'meta_data_file': meta_data_filename})
)
return train_input_fn, eval_input_fn
def create_sequence_dataset_input_fn(dataset_dir, prepare_fn, record_file_name_format,
meta_data_file_name_format, bucket_boundaries):
prepare_fn(dataset_dir)
train_data_file = record_file_name_format.format(dataset_dir, Modes.TRAIN)
eval_data_file = record_file_name_format.format(dataset_dir, Modes.EVAL)
meta_data_filename = meta_data_file_name_format.format(dataset_dir)
train_input_fn = create_input_data_fn(
mode=Modes.TRAIN,
pipeline_config=PipelineConfig(module='TFRecordSequencePipeline', dynamic_pad=True,
bucket_boundaries=bucket_boundaries,
params={'data_files': train_data_file,
'meta_data_file': meta_data_filename})
)
eval_input_fn = create_input_data_fn(
mode=Modes.EVAL,
pipeline_config=PipelineConfig(module='TFRecordSequencePipeline', dynamic_pad=True,
bucket_boundaries=bucket_boundaries,
params={'data_files': eval_data_file,
'meta_data_file': meta_data_filename})
)
return train_input_fn, eval_input_fn
def create_dataset_predict_input_fn(dataset_dir, prepare_fn, record_file_name_format,
meta_data_file_name_format):
prepare_fn(dataset_dir)
test_data_file = record_file_name_format.format(dataset_dir, Modes.PREDICT)
meta_data_filename = meta_data_file_name_format.format(dataset_dir)
test_input_fn = create_input_data_fn(
mode=Modes.PREDICT,
pipeline_config=PipelineConfig(module='TFRecordImagePipeline', dynamic_pad=False,
num_epochs=1,
params={'data_files': test_data_file,
'meta_data_file': meta_data_filename})
)
return test_input_fn
def create_image_dataset_predict_input_fn(dataset_dir, prepare_fn, record_file_name_format,
meta_data_file_name_format):
prepare_fn(dataset_dir)
test_data_file = record_file_name_format.format(dataset_dir, Modes.PREDICT)
meta_data_filename = meta_data_file_name_format.format(dataset_dir)
test_input_fn = create_input_data_fn(
mode=Modes.PREDICT,
pipeline_config=PipelineConfig(module='TFRecordImagePipeline', dynamic_pad=False,
num_epochs=1,
params={'data_files': test_data_file,
'meta_data_file': meta_data_filename})
)
return test_input_fn
def create_image_dataset_predict_input_fn(dataset_dir, prepare_fn,
record_file_name_format,
meta_data_file_name_format):
prepare_fn(dataset_dir)
test_data_file = record_file_name_format.format(dataset_dir, Modes.PREDICT)
meta_data_filename = meta_data_file_name_format.format(dataset_dir)
test_input_fn = create_input_data_fn(
mode=Modes.PREDICT,
pipeline_config=TFRecordImagePipelineConfig(
dynamic_pad=False,
num_epochs=1,
data_files=test_data_file,
meta_data_file=meta_data_filename))
return test_input_fn
def create_sequence_dataset_predict_input_fn(dataset_dir, prepare_fn, record_file_name_format,
meta_data_file_name_format, bucket_boundaries):
prepare_fn(dataset_dir)
test_data_file = record_file_name_format.format(dataset_dir, Modes.PREDICT)
meta_data_filename = meta_data_file_name_format.format(dataset_dir)
test_input_fn = create_input_data_fn(
mode=Modes.PREDICT,
pipeline_config=PipelineConfig(module='TFRecordSequencePipeline', dynamic_pad=True,
num_epochs=1, batch_size=4, min_after_dequeue=0,
bucket_boundaries=bucket_boundaries,
params={'data_files': test_data_file,
'meta_data_file': meta_data_filename})
)
return test_input_fn
def create_sequence_dataset_predict_input_fn(dataset_dir, prepare_fn,
record_file_name_format,
meta_data_file_name_format,
bucket_boundaries):
prepare_fn(dataset_dir)
test_data_file = record_file_name_format.format(dataset_dir, Modes.PREDICT)
meta_data_filename = meta_data_file_name_format.format(dataset_dir)
test_input_fn = create_input_data_fn(
mode=Modes.PREDICT,
pipeline_config=TFRecordSequencePipelineConfig(
dynamic_pad=True,
num_epochs=1,
batch_size=4,
min_after_dequeue=0,
bucket_boundaries=bucket_boundaries,
data_files=test_data_file,
meta_data_file=meta_data_filename))
return test_input_fn
