Exporter updates.

object_detection/exporter.py

@@ -113,14 +113,19 @@ def freeze_graph_with_def_protos(
 
 
 def _image_tensor_input_placeholder():
-  """Returns input node that accepts a batch of uint8 images."""
-  return tf.placeholder(dtype=tf.uint8,
-                        shape=(None, None, None, 3),
-                        name='image_tensor')
+  """Returns placeholder and input node that accepts a batch of uint8 images."""
+  input_tensor = tf.placeholder(dtype=tf.uint8,
+                                shape=(None, None, None, 3),
+                                name='image_tensor')
+  return input_tensor, input_tensor
 
 
 def _tf_example_input_placeholder():
-  """Returns input node that accepts a batch of strings with tf examples."""
+  """Returns input that accepts a batch of strings with tf examples.
+
+  Returns:
+    a tuple of placeholder and input nodes that output decoded images.
+  """
   batch_tf_example_placeholder = tf.placeholder(
       tf.string, shape=[None], name='tf_example')
   def decode(tf_example_string_tensor):
@@ -128,15 +133,20 @@ def _tf_example_input_placeholder():
         tf_example_string_tensor)
     image_tensor = tensor_dict[fields.InputDataFields.image]
     return image_tensor
-  return tf.map_fn(decode,
-                   elems=batch_tf_example_placeholder,
-                   dtype=tf.uint8,
-                   parallel_iterations=32,
-                   back_prop=False)
+  return (batch_tf_example_placeholder,
+          tf.map_fn(decode,
+                    elems=batch_tf_example_placeholder,
+                    dtype=tf.uint8,
+                    parallel_iterations=32,
+                    back_prop=False))
 
 
 def _encoded_image_string_tensor_input_placeholder():
-  """Returns input node that accepts a batch of PNG or JPEG strings."""
+  """Returns input that accepts a batch of PNG or JPEG strings.
+
+  Returns:
+    a tuple of placeholder and input nodes that output decoded images.
+  """
   batch_image_str_placeholder = tf.placeholder(
       dtype=tf.string,
       shape=[None],
@@ -146,11 +156,13 @@ def _encoded_image_string_tensor_input_placeholder():
                                          channels=3)
     image_tensor.set_shape((None, None, 3))
     return image_tensor
-  return tf.map_fn(decode,
-                   elems=batch_image_str_placeholder,
-                   dtype=tf.uint8,
-                   parallel_iterations=32,
-                   back_prop=False)
+  return (batch_image_str_placeholder,
+          tf.map_fn(
+              decode,
+              elems=batch_image_str_placeholder,
+              dtype=tf.uint8,
+              parallel_iterations=32,
+              back_prop=False))
 
 
 input_placeholder_fn_map = {
@@ -262,7 +274,7 @@ def _write_saved_model(saved_model_path,
       builder.add_meta_graph_and_variables(
           sess, [tf.saved_model.tag_constants.SERVING],
           signature_def_map={
-              'signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY':
+              signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY:
                   detection_signature,
           },
       )
@@ -300,7 +312,8 @@ def _export_inference_graph(input_type,
 
   if input_type not in input_placeholder_fn_map:
     raise ValueError('Unknown input type: {}'.format(input_type))
-  inputs = tf.to_float(input_placeholder_fn_map[input_type]())
+  placeholder_tensor, input_tensors = input_placeholder_fn_map[input_type]()
+  inputs = tf.to_float(input_tensors)
   preprocessed_inputs = detection_model.preprocess(inputs)
   output_tensors = detection_model.predict(preprocessed_inputs)
   postprocessed_tensors = detection_model.postprocess(output_tensors)
@@ -333,7 +346,8 @@ def _export_inference_graph(input_type,
       optimize_graph=optimize_graph,
       initializer_nodes='')
   _write_frozen_graph(frozen_graph_path, frozen_graph_def)
-  _write_saved_model(saved_model_path, frozen_graph_def, inputs, outputs)
+  _write_saved_model(saved_model_path, frozen_graph_def, placeholder_tensor,
+                     outputs)
 
 
 def export_inference_graph(input_type,