Python tensorflow.python.platform.gfile 模块，IsDirectory() 实例源码

def read_decode_tfrecords(records_path, num_epochs=1020, batch_size=Flags.batch_size, num_threads=2):
    if gfile.IsDirectory(records_path):
        records_path = [os.path.join(records_path, i) for i in os.listdir(records_path)]
    else:
        records_path = [records_path]
    records_path_queue = tf.train.string_input_producer(records_path, seed=123,
                                                        num_epochs=num_epochs,
                                                        name="string_input_producer")
    reader = tf.TFRecordReader()
    _, serialized_example = reader.read(records_path_queue, name="serialized_example")
    features = tf.parse_single_example(serialized=serialized_example,
                                       features={"img_raw": tf.FixedLenFeature([], tf.string),
                                                 "label": tf.FixedLenFeature([], tf.int64),
                                                 "height": tf.FixedLenFeature([], tf.int64),
                                                 "width": tf.FixedLenFeature([], tf.int64),
                                                 "depth": tf.FixedLenFeature([], tf.int64)},
                                       name="parse_single_example")
    image = tf.decode_raw(features["img_raw"], tf.uint8, name="decode_raw")
    image.set_shape([height * width * 3])
    image = tf.cast(image, tf.float32) * (1.0 / 255) - 0.5
    label = tf.cast(features["label"], tf.int32)
    images, labels = tf.train.shuffle_batch([image, label], batch_size=batch_size, num_threads=num_threads,
                                            name="shuffle_bath", capacity=1020, min_after_dequeue=64)
    print("images' shape is :", str(images.shape))
    return images, labels

def read_decode_tfrecords(records_path, num_epochs=1, batch_size=Flags.batch_size, num_threads=1):
    if gfile.IsDirectory(records_path):
        records_path = [os.path.join(records_path, i) for i in os.listdir(records_path)]
    else:
        records_path = [records_path]
    records_path_queue = tf.train.string_input_producer(records_path, seed=123,
                                                        num_epochs=None,
                                                        name="string_input_producer")
    reader = tf.TFRecordReader()
    _, serialized_example = reader.read(records_path_queue, name="serialized_example")
    features = tf.parse_single_example(serialized=serialized_example,
                                       features={"img_raw": tf.FixedLenFeature([], tf.string),
                                                 "label": tf.FixedLenFeature([], tf.int64),
                                                 "height": tf.FixedLenFeature([], tf.int64),
                                                 "width": tf.FixedLenFeature([], tf.int64),
                                                 "depth": tf.FixedLenFeature([], tf.int64)},
                                       name="parse_single_example")
    image = tf.decode_raw(features["img_raw"], tf.uint8, name="decode_raw")
    image.set_shape([IMAGE_PIXELS])
    image = tf.cast(image, tf.float32) * (1.0 / 255) - 0.5
    label = tf.cast(features["label"], tf.int32)
    images, labels = tf.train.shuffle_batch([image, label], batch_size=batch_size, num_threads=num_threads,
                                            name="shuffle_bath", capacity=1020, min_after_dequeue=50)
    return images, labels

def read_decode_tfrecords(records_path, num_epochs=1020, batch_size=Flags.batch_size, num_threads=2):
    if gfile.IsDirectory(records_path):
        records_path = [os.path.join(records_path, i) for i in os.listdir(records_path)]
    else:
        records_path = [records_path]
    records_path_queue = tf.train.string_input_producer(records_path, seed=123,
                                                        # num_epochs=num_epochs,
                                                        name="string_input_producer")
    reader = tf.TFRecordReader()
    _, serialized_example = reader.read(records_path_queue, name="serialized_example")
    features = tf.parse_single_example(serialized=serialized_example,
                                       features={"img_raw": tf.FixedLenFeature([], tf.string),
                                                 "label": tf.FixedLenFeature([], tf.int64),
                                                 "height": tf.FixedLenFeature([], tf.int64),
                                                 "width": tf.FixedLenFeature([], tf.int64),
                                                 "depth": tf.FixedLenFeature([], tf.int64)},
                                       name="parse_single_example")
    image = tf.decode_raw(features["img_raw"], tf.uint8, name="decode_raw")
    image.set_shape([IMAGE_PIXELS])
    image = tf.cast(image, tf.float32) * (1.0 / 255) - 0.5
    label = tf.cast(features["label"], tf.int32)
    # images, labels = tf.train.shuffle_batch([image, label], batch_size=batch_size, num_threads=num_threads,
    #                                         name="shuffle_bath", capacity=1020, min_after_dequeue=64)
    return image, label

def _get_checkpoint_filename(filepattern):
  """Returns checkpoint filename given directory or specific filepattern."""
  if gfile.IsDirectory(filepattern):
    return saver.latest_checkpoint(filepattern)
  return filepattern

def _get_checkpoint_filename(filepattern):
  """Returns checkpoint filename given directory or specific filepattern."""
  if gfile.IsDirectory(filepattern):
    return saver.latest_checkpoint(filepattern)
  return filepattern

def _get_checkpoint_filename(filepattern):
  """Returns checkpoint filename given directory or specific filepattern."""
  if gfile.IsDirectory(filepattern):
    return saver.latest_checkpoint(filepattern)
  return filepattern

def _copy_dir(dir_in, dir_out):
  gfile.MakeDirs(dir_out)
  for name in gfile.ListDirectory(dir_in):
    name_in = os.path.join(dir_in, name)
    name_out = os.path.join(dir_out, name)
    if gfile.IsDirectory(name_in):
      gfile.MakeDirs(name_out)
      _copy_dir(name_in, name_out)
    else:
      gfile.Copy(name_in, name_out, overwrite=True)

def _get_checkpoint_filename(filepattern):
  """Returns checkpoint filename given directory or specific filepattern."""
  if gfile.IsDirectory(filepattern):
    return saver.latest_checkpoint(filepattern)
  return filepattern

def _has_new_checkpoint(path, last):
    new_files = {}
    for f in gfile.ListDirectory(path):
        if not gfile.IsDirectory(f):
            try:
                global_step = extract_global_step(f)
            except Exception:  # pylint: disable=broad-except
                continue
            if global_step and (not last or global_step > last):
                new_files[global_step] = f
    if new_files == {}:
        return last, None
    min_global_step = min(new_files)
    return min_global_step, new_files[min_global_step]

def _get_checkpoint_filename(filepattern):
  """Returns checkpoint filename given directory or specific filepattern."""
  if gfile.IsDirectory(filepattern):
    return saver.latest_checkpoint(filepattern)
  return filepattern