Python tensorflow.import_graph_def() Examples

def load_graph(frozen_graph_filename):
    # We load the protobuf file from the disk and parse it to retrieve the
    # unserialized graph_def
    with tf.gfile.GFile(frozen_graph_filename, "rb") as f:
        graph_def = tf.GraphDef()
        graph_def.ParseFromString(f.read())

    # Then, we can use again a convenient built-in function to import a graph_def into the
    # current default Graph
    with tf.Graph().as_default() as graph:
        tf.import_graph_def(
            graph_def,
            input_map=None,
            return_elements=None,
            name="",
            op_dict=None,
            producer_op_list=None
        )
    return graph 

def worker(input_q, output_q):
    # Load a (frozen) Tensorflow model into memory.
    detection_graph = tf.Graph()
    with detection_graph.as_default():
        od_graph_def = tf.GraphDef()
        with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
            serialized_graph = fid.read()
            od_graph_def.ParseFromString(serialized_graph)
            tf.import_graph_def(od_graph_def, name='')

        sess = tf.Session(graph=detection_graph)

    fps = FPS().start()
    while True:
        fps.update()
        frame = input_q.get()
        frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
        output_q.put(detect_objects(frame_rgb, sess, detection_graph))

    fps.stop()
    sess.close() 

def load_graph(graph_path,tensorboard=False,**kwargs):
    '''
    :param graph_filename: the path of the pb file
    :return: tensorflow graph
    '''
    with gfile.FastGFile(graph_path,'rb') as f:
        graph_def = graph_pb2.GraphDef()
        graph_def.ParseFromString(f.read())

    with tf.Graph().as_default() as graph:
        tf.import_graph_def(graph_def,name="")

    if tensorboard:
        writer = tf.summary.FileWriter("log/")
        writer.add_graph(graph)

    return graph 

def create_inception_graph():
  """"Creates a graph from saved GraphDef file and returns a Graph object.

  Returns:
    Graph holding the trained Inception network, and various tensors we'll be
    manipulating.
  """
  with tf.Graph().as_default() as graph:
    model_filename = os.path.join(
        FLAGS.model_dir, 'classify_image_graph_def.pb')
    with gfile.FastGFile(model_filename, 'rb') as f:
      graph_def = tf.GraphDef()
      graph_def.ParseFromString(f.read())
      bottleneck_tensor, jpeg_data_tensor, resized_input_tensor = (
          tf.import_graph_def(graph_def, name='', return_elements=[
              BOTTLENECK_TENSOR_NAME, JPEG_DATA_TENSOR_NAME,
              RESIZED_INPUT_TENSOR_NAME]))
  return graph, bottleneck_tensor, jpeg_data_tensor, resized_input_tensor 

def load_model(model, input_map=None):
    # Check if the model is a model directory (containing a metagraph and a checkpoint file)
    #  or if it is a protobuf file with a frozen graph
    model_exp = os.path.expanduser(model)
    if (os.path.isfile(model_exp)):
        print('Model filename: %s' % model_exp)
        with gfile.FastGFile(model_exp,'rb') as f:
            graph_def = tf.GraphDef()
            graph_def.ParseFromString(f.read())
            tf.import_graph_def(graph_def, input_map=input_map, name='')
    else:
        print('Model directory: %s' % model_exp)
        meta_file, ckpt_file = get_model_filenames(model_exp)
        
        print('Metagraph file: %s' % meta_file)
        print('Checkpoint file: %s' % ckpt_file)
      
        saver = tf.train.import_meta_graph(os.path.join(model_exp, meta_file), input_map=input_map)
        saver.restore(tf.get_default_session(), os.path.join(model_exp, ckpt_file)) 

def load_model(model, input_map=None):
    # Check if the model is a model directory (containing a metagraph and a checkpoint file)
    #  or if it is a protobuf file with a frozen graph
    model_exp = os.path.expanduser(model)
    if (os.path.isfile(model_exp)):
        print('Model filename: %s' % model_exp)
        with gfile.FastGFile(model_exp,'rb') as f:
            graph_def = tf.GraphDef()
            graph_def.ParseFromString(f.read())
            tf.import_graph_def(graph_def, input_map=input_map, name='')
    else:
        print('Model directory: %s' % model_exp)
        meta_file, ckpt_file = get_model_filenames(model_exp)
        
        print('Metagraph file: %s' % meta_file)
        print('Checkpoint file: %s' % ckpt_file)
      
        saver = tf.train.import_meta_graph(os.path.join(model_exp, meta_file), input_map=input_map)
        saver.restore(tf.get_default_session(), os.path.join(model_exp, ckpt_file)) 

def load_model(model, input_map=None):
    # Check if the model is a model directory (containing a metagraph and a checkpoint file)
    #  or if it is a protobuf file with a frozen graph
    model_exp = os.path.expanduser(model)
    if (os.path.isfile(model_exp)):
        print('Model filename: %s' % model_exp)
        with gfile.FastGFile(model_exp,'rb') as f:
            graph_def = tf.GraphDef()
            graph_def.ParseFromString(f.read())
            tf.import_graph_def(graph_def, input_map=input_map, name='')
    else:
        print('Model directory: %s' % model_exp)
        meta_file, ckpt_file = get_model_filenames(model_exp)
        
        print('Metagraph file: %s' % meta_file)
        print('Checkpoint file: %s' % ckpt_file)
      
        saver = tf.train.import_meta_graph(os.path.join(model_exp, meta_file), input_map=input_map)
        saver.restore(tf.get_default_session(), os.path.join(model_exp, ckpt_file)) 

def __init__(self, checkpoint_filename, input_name="images",
                 output_name="features"):
        self.session = tf.Session()
        with tf.gfile.GFile(checkpoint_filename, "rb") as file_handle:
            graph_def = tf.GraphDef()
            graph_def.ParseFromString(file_handle.read())
        tf.import_graph_def(graph_def, name="net")
        self.input_var = tf.get_default_graph().get_tensor_by_name(
            "net/%s:0" % input_name)
        self.output_var = tf.get_default_graph().get_tensor_by_name(
            "net/%s:0" % output_name)

        assert len(self.output_var.get_shape()) == 2
        assert len(self.input_var.get_shape()) == 4
        self.feature_dim = self.output_var.get_shape().as_list()[-1]
        self.image_shape = self.input_var.get_shape().as_list()[1:] 

def build_from_pb(self):
        with tf.gfile.FastGFile(self.FLAGS.pbLoad, "rb") as f:
            graph_def = tf.GraphDef()
            graph_def.ParseFromString(f.read())

        tf.import_graph_def(
            graph_def,
            name=""
        )
        with open(self.FLAGS.metaLoad, 'r') as fp:
            self.meta = json.load(fp)
        self.framework = create_framework(self.meta, self.FLAGS)

        # Placeholders
        self.inp = tf.get_default_graph().get_tensor_by_name('input:0')
        self.feed = dict()  # other placeholders
        self.out = tf.get_default_graph().get_tensor_by_name('output:0')

        self.setup_meta_ops() 

def __init__(self, PATH_TO_CKPT):
        """Tensorflow detector
        """

        self.detection_graph = tf.Graph()
        with self.detection_graph.as_default():
            od_graph_def = tf.GraphDef()
            with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
                serialized_graph = fid.read()
                od_graph_def.ParseFromString(serialized_graph)
                tf.import_graph_def(od_graph_def, name='')


        with self.detection_graph.as_default():
            config = tf.ConfigProto()
            config.gpu_options.allow_growth = True
            with tf.Session(graph=self.detection_graph, config=config) as self.sess:

                self.windowNotSet = True 

def __init__(self, config, graph, model_scope, model_dir, model_file):
        self.config = config

        frozen_model = os.path.join(model_dir, model_file)
        with tf.gfile.GFile(frozen_model, "rb") as f:
            graph_def = tf.GraphDef()
            graph_def.ParseFromString(f.read())

        # This model_scope adds a prefix to all the nodes in the graph
        tf.import_graph_def(graph_def, input_map=None, return_elements=None,
                            name="{}/".format(model_scope))

        # Uncomment the two lines below to look for the names of all the operations in the graph
        # for op in graph.get_operations():
        #    print(op.name)

        # Using the lines commented above to look for the tensor name of the input node
        # Or you can figure it out in your original model, if you explicitly named it.
        self.input_tensor = graph.get_tensor_by_name("{}/input_1:0".format(model_scope))
        self.output_tensor = graph.get_tensor_by_name("{}/s1_output0:0".format(model_scope)) 

def _rewrite_graph(self, transform):
        input_map = {k.name: v for k, v in transform.items()}

        # Modify the input dictionary to replace variables which have been
        # superseded with the use of combinators
        for k, v in self._silently_replace.items():
            input_map[k.name] = self._observed[v]

        with self.session.graph.as_default():
            try:
                tf.import_graph_def(
                        self._model_graph.as_graph_def(),
                        input_map=input_map,
                        name='added',
                )
            except ValueError:
                # Ignore errors that ocour when the input_map tries to
                # rewrite a variable that isn't present in the graph
                pass 

def worker(input_q, output_q):
    # Load a (frozen) Tensorflow model into memory.
    detection_graph = tf.Graph()
    with detection_graph.as_default():
        od_graph_def = tf.GraphDef()
        with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
            serialized_graph = fid.read()
            od_graph_def.ParseFromString(serialized_graph)
            tf.import_graph_def(od_graph_def, name='')

        sess = tf.Session(graph=detection_graph)

    fps = FPS().start()
    while True:
        fps.update()
        frame = input_q.get()
        frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
        output_q.put(detect_objects(frame_rgb, sess, detection_graph))

    fps.stop()
    sess.close() 

def create_inception_graph():
  """"Creates a graph from saved GraphDef file and returns a Graph object.

  Returns:
    Graph holding the trained Inception network, and various tensors we'll be
    manipulating.
  """
  with tf.Session() as sess:
    model_filename = os.path.join(
        FLAGS.model_dir, 'classify_image_graph_def.pb')
    with gfile.FastGFile(model_filename, 'rb') as f:
      graph_def = tf.GraphDef()
      graph_def.ParseFromString(f.read())
      bottleneck_tensor, jpeg_data_tensor, resized_input_tensor = (
          tf.import_graph_def(graph_def, name='', return_elements=[
              BOTTLENECK_TENSOR_NAME, JPEG_DATA_TENSOR_NAME,
              RESIZED_INPUT_TENSOR_NAME]))
  return sess.graph, bottleneck_tensor, jpeg_data_tensor, resized_input_tensor 

def load_graph(graph_path,tensorboard=False,**kwargs):
    '''
    :param graph_filename: the path of the pb file
    :return: tensorflow graph
    '''
    with gfile.FastGFile(graph_path,'rb') as f:
        graph_def = graph_pb2.GraphDef()
        graph_def.ParseFromString(f.read())

    with tf.Graph().as_default() as graph:
        tf.import_graph_def(graph_def,name="")

    if tensorboard:
        writer = tf.summary.FileWriter("log/")
        writer.add_graph(graph)

    return graph 

def fromGraphDef(cls, graph_def, feed_names, fetch_names):
        """
        Construct a TFInputGraph from a tf.GraphDef object.

        :param graph_def: :py:class:`tf.GraphDef`, a serializable object containing the topology and
                           computation units of the TensorFlow graph.
        :param feed_names: list, names of the input tensors.
        :param fetch_names: list, names of the output tensors.
        """
        assert isinstance(graph_def, tf.GraphDef), \
            ('expect tf.GraphDef type but got', type(graph_def))

        graph = tf.Graph()
        with tf.Session(graph=graph) as sess:
            tf.import_graph_def(graph_def, name='')
            return _build_with_feeds_fetches(sess=sess, graph=graph, feed_names=feed_names,
                                             fetch_names=fetch_names) 

def load_inference_graph():
    # load frozen tensorflow model into memory
    print("> ====== loading HAND frozen graph into memory")
    detection_graph = tf.Graph()
    with detection_graph.as_default():
        od_graph_def = tf.GraphDef()
        with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
            serialized_graph = fid.read()
            od_graph_def.ParseFromString(serialized_graph)
            tf.import_graph_def(od_graph_def, name='')
        sess = tf.Session(graph=detection_graph)
    print(">  ====== Hand Inference graph loaded.")
    return detection_graph, sess


# draw the detected bounding boxes on the images
# You can modify this to also draw a label. 

def __init__(self, checkpoint_filename, input_name="images",
                 output_name="features"):
        self.session = tf.Session()
        with tf.gfile.GFile(checkpoint_filename, "rb") as file_handle:
            graph_def = tf.GraphDef()
            graph_def.ParseFromString(file_handle.read())
        tf.import_graph_def(graph_def, name="net")
        self.input_var = tf.get_default_graph().get_tensor_by_name(
            "net/%s:0" % input_name)
        self.output_var = tf.get_default_graph().get_tensor_by_name(
            "net/%s:0" % output_name)

        assert len(self.output_var.get_shape()) == 2
        assert len(self.input_var.get_shape()) == 4
        self.feature_dim = self.output_var.get_shape().as_list()[-1]
        self.image_shape = self.input_var.get_shape().as_list()[1:] 

def load_graph(frozen_graph_file):

    # we parse the graph_def file
    with tf.gfile.GFile(frozen_graph_file, 'rb') as f:
        graph_def = tf.GraphDef()
        graph_def.ParseFromString(f.read())

    # we load the graph_def in the default graph

    with tf.Graph().as_default() as graph:
        tf.import_graph_def(graph_def,
                            input_map=None,
                            return_elements=None,
                            name="",
                            op_dict=None,
                            producer_op_list=None)
    return graph 

def inference():
  graph = tf.Graph()

  with graph.as_default():
    with tf.gfile.FastGFile(FLAGS.input, 'rb') as f:
      image_data = f.read()
      input_image = tf.image.decode_jpeg(image_data, channels=3)
      input_image = tf.image.resize_images(input_image, size=(FLAGS.image_size, FLAGS.image_size))
      input_image = utils.convert2float(input_image)
      input_image.set_shape([FLAGS.image_size, FLAGS.image_size, 3])

    with tf.gfile.FastGFile(FLAGS.model, 'rb') as model_file:
      graph_def = tf.GraphDef()
      graph_def.ParseFromString(model_file.read())
    [output_image] = tf.import_graph_def(graph_def,
                          input_map={'input_image': input_image},
                          return_elements=['output_image:0'],
                          name='output')

  with tf.Session(graph=graph) as sess:
    generated = output_image.eval()
    with open(FLAGS.output, 'wb') as f:
      f.write(generated) 

def _import_graph_and_run_inference(self, tflite_graph_file, num_channels=3):
    """Imports a tflite graph, runs single inference and returns outputs."""
    graph = tf.Graph()
    with graph.as_default():
      graph_def = tf.GraphDef()
      with tf.gfile.Open(tflite_graph_file) as f:
        graph_def.ParseFromString(f.read())
      tf.import_graph_def(graph_def, name='')
      input_tensor = graph.get_tensor_by_name('normalized_input_image_tensor:0')
      box_encodings = graph.get_tensor_by_name('raw_outputs/box_encodings:0')
      class_predictions = graph.get_tensor_by_name(
          'raw_outputs/class_predictions:0')
      with self.test_session(graph) as sess:
        [box_encodings_np, class_predictions_np] = sess.run(
            [box_encodings, class_predictions],
            feed_dict={input_tensor: np.random.rand(1, 10, 10, num_channels)})
    return box_encodings_np, class_predictions_np 

def initialize():
    print('Loading model...',end=''),
    with open(filename, 'rb') as f:
        graph_def.ParseFromString(f.read())
        tf.import_graph_def(graph_def, name='')

    # Retrieving 'network_input_size' from shape of 'input_node'
    with tf.compat.v1.Session() as sess:
        input_tensor_shape = sess.graph.get_tensor_by_name(input_node).shape.as_list()
        
    assert len(input_tensor_shape) == 4
    assert input_tensor_shape[1] == input_tensor_shape[2]

    global network_input_size
    network_input_size = input_tensor_shape[1]
   
    print('Success!')
    print('Loading labels...', end='')
    with open(labels_filename, 'rt') as lf:
        global labels
        labels = [l.strip() for l in lf.readlines()]
    print(len(labels), 'found. Success!') 

def build_from_pb(self):
		with tf.gfile.FastGFile(self.FLAGS.pbLoad, "rb") as f:
			graph_def = tf.GraphDef()
			graph_def.ParseFromString(f.read())
		
		tf.import_graph_def(
			graph_def,
			name=""
		)
		with open(self.FLAGS.metaLoad, 'r') as fp:
			self.meta = json.load(fp)
		self.framework = create_framework(self.meta, self.FLAGS)

		# Placeholders
		self.inp = tf.get_default_graph().get_tensor_by_name('input:0')
		self.feed = dict() # other placeholders
		self.out = tf.get_default_graph().get_tensor_by_name('output:0')
		
		self.setup_meta_ops() 

def initialize():
    print('Loading model...',end=''),
    with open(filename, 'rb') as f:
        graph_def.ParseFromString(f.read())
        tf.import_graph_def(graph_def, name='')

    # Retrieving 'network_input_size' from shape of 'input_node'
    with tf.compat.v1.Session() as sess:
        input_tensor_shape = sess.graph.get_tensor_by_name(input_node).shape.as_list()
        
    assert len(input_tensor_shape) == 4
    assert input_tensor_shape[1] == input_tensor_shape[2]

    global network_input_size
    network_input_size = input_tensor_shape[1]
   
    print('Success!')
    print('Loading labels...', end='')
    with open(labels_filename, 'rt') as lf:
        global labels
        labels = [l.strip() for l in lf.readlines()]
    print(len(labels), 'found. Success!') 

def create_inception_graph():
  """"Creates a graph from saved GraphDef file and returns a Graph object.

  Returns:
    Graph holding the trained Inception network, and various tensors we'll be
    manipulating.
  """
  with tf.Graph().as_default() as graph:
    model_filename = os.path.join(
        FLAGS.model_dir, 'classify_image_graph_def.pb')
    with gfile.FastGFile(model_filename, 'rb') as f:
      graph_def = tf.GraphDef()
      graph_def.ParseFromString(f.read())
      bottleneck_tensor, jpeg_data_tensor, resized_input_tensor = (
          tf.import_graph_def(graph_def, name='', return_elements=[
              BOTTLENECK_TENSOR_NAME, JPEG_DATA_TENSOR_NAME,
              RESIZED_INPUT_TENSOR_NAME]))
  return graph, bottleneck_tensor, jpeg_data_tensor, resized_input_tensor 

def _check_output(gin, tf_input, expected):
    """
    Takes a TFInputGraph object (assumed to have the input and outputs of the given
    names above) and compares the outcome against some expected outcome.
    """
    graph = tf.Graph()
    graph_def = gin.graph_def
    with tf.Session(graph=graph) as sess:
        tf.import_graph_def(graph_def, name="")
        tgt_feed = tfx.get_tensor(_tensor_input_name, graph)
        tgt_fetch = tfx.get_tensor(_tensor_output_name, graph)
        # Run on the testing target
        tgt_out = sess.run(tgt_fetch, feed_dict={tgt_feed: tf_input})
        # Working on integers, the calculation should be exact
        assert np.all(tgt_out == expected), (tgt_out, expected)


# TODO: we could factorize with _check_output, but this is not worth the time doing it. 

def _check_output_2(gin, tf_input1, tf_input2, expected):
    """
    Takes a TFInputGraph object (assumed to have the input and outputs of the given
    names above) and compares the outcome against some expected outcome.
    """
    graph = tf.Graph()
    graph_def = gin.graph_def
    with tf.Session(graph=graph) as sess:
        tf.import_graph_def(graph_def, name="")
        tgt_feed1 = tfx.get_tensor(_tensor_input_name, graph)
        tgt_feed2 = tfx.get_tensor(_tensor_input_name_2, graph)
        tgt_fetch = tfx.get_tensor(_tensor_output_name, graph)
        # Run on the testing target
        tgt_out = sess.run(tgt_fetch, feed_dict={tgt_feed1: tf_input1, tgt_feed2: tf_input2})
        # Working on integers, the calculation should be exact
        assert np.all(tgt_out == expected), (tgt_out, expected) 

def build_from_pb(self):
		with tf.gfile.FastGFile(self.FLAGS.pbLoad, "rb") as f:
			graph_def = tf.GraphDef()
			graph_def.ParseFromString(f.read())
		
		tf.import_graph_def(
			graph_def,
			name=""
		)
		with open(self.FLAGS.metaLoad, 'r') as fp:
			self.meta = json.load(fp)
		self.framework = create_framework(self.meta, self.FLAGS)

		# Placeholders
		self.inp = tf.get_default_graph().get_tensor_by_name('input:0')
		self.feed = dict() # other placeholders
		self.out = tf.get_default_graph().get_tensor_by_name('output:0')
		
		self.setup_meta_ops() 

def load_graph(filename):
  """Unpersists graph from file as default graph."""
  with tf.gfile.FastGFile(filename, 'rb') as f:
    graph_def = tf.GraphDef()
    graph_def.ParseFromString(f.read())
    tf.import_graph_def(graph_def, name='') 

def _get_placeholder_types(self, user_graph_def):
        """ Returns a list of placeholder type enums for the input nodes """
        user_graph = tf.Graph()
        with user_graph.as_default():
            # Load user-specified graph into memory, then get the data type of each input node
            tf.import_graph_def(user_graph_def, name="")
            res = []
            for _, tensor_name in self.getInputMapping():
                placeholder_type = tfx.get_tensor(tensor_name, user_graph).dtype.as_datatype_enum
                res.append(placeholder_type)
        return res