Update slim and fix minor issue in object detection (tensorflow#5354)

* Merged commit includes the following changes:
213899768  by Sergio Guadarrama:

    Fixes tensorflow#3819.

--
213493831  by Sergio Guadarrama:

    Internal change

212057654  by Sergio Guadarrama:

    Internal change

210747685  by Sergio Guadarrama:

    For FPN, when use_depthwise is set to true, use slightly modified mobilenet v1 config.

--
210128931  by Sergio Guadarrama:

    Allow user-defined current_step in NASNet.

--
209092664  by Sergio Guadarrama:

    Add quantized fine-tuning / training / eval and export to slim image classifier binaries.

--
207651347  by Sergio Guadarrama:

    Update mobilenet v1 docs to include revised tflite models.

--
207165245  by Sergio Guadarrama:

    Internal change

207095064  by Sergio Guadarrama:

    Internal change

PiperOrigin-RevId: 213899768

* Update model_lib.py to fix eval_spec name issue.

research/object_detection/model_lib.py

@@ -641,7 +641,7 @@ def create_train_and_eval_specs(train_input_fn,
       input_fn=train_input_fn, max_steps=train_steps)
 
   if eval_spec_names is None:
-    eval_spec_names = range(len(eval_input_fns))
+    eval_spec_names = [ str(i) for i in range(len(eval_input_fns)) ]
 
   eval_specs = []
   for eval_spec_name, eval_input_fn in zip(eval_spec_names, eval_input_fns):

research/slim/BUILD

@@ -80,7 +80,7 @@ py_binary(
     ],
 )
 
-py_binary(
+py_library(
     name = "cifar10",
     srcs = ["datasets/cifar10.py"],
     deps = [
@@ -89,7 +89,7 @@ py_binary(
     ],
 )
 
-py_binary(
+py_library(
     name = "flowers",
     srcs = ["datasets/flowers.py"],
     deps = [
@@ -98,7 +98,7 @@ py_binary(
     ],
 )
 
-py_binary(
+py_library(
     name = "imagenet",
     srcs = ["datasets/imagenet.py"],
     deps = [
@@ -107,7 +107,7 @@ py_binary(
     ],
 )
 
-py_binary(
+py_library(
     name = "mnist",
     srcs = ["datasets/mnist.py"],
     deps = [
@@ -715,8 +715,8 @@ py_binary(
     ],
 )
 
-py_binary(
-    name = "eval_image_classifier",
+py_library(
+    name = "eval_image_classifier_lib",
     srcs = ["eval_image_classifier.py"],
     deps = [
         ":dataset_factory",
@@ -726,6 +726,14 @@ py_binary(
     ],
 )
 
+py_binary(
+    name = "eval_image_classifier",
+    srcs = ["eval_image_classifier.py"],
+    deps = [
+        ":eval_image_classifier_lib",
+    ],
+)
+
 py_binary(
     name = "export_inference_graph",
     srcs = ["export_inference_graph.py"],

research/slim/eval_image_classifier.py

@@ -79,6 +79,9 @@
 tf.app.flags.DEFINE_integer(
     'eval_image_size', None, 'Eval image size')
 
+tf.app.flags.DEFINE_bool(
+    'quantize', False, 'whether to use quantized graph or not.')
+
 FLAGS = tf.app.flags.FLAGS
 
 
@@ -138,6 +141,9 @@ def main(_):
     ####################
     logits, _ = network_fn(images)
 
+    if FLAGS.quantize:
+      tf.contrib.quantize.create_eval_graph()
+
     if FLAGS.moving_average_decay:
       variable_averages = tf.train.ExponentialMovingAverage(
           FLAGS.moving_average_decay, tf_global_step)

research/slim/export_inference_graph.py

@@ -96,6 +96,9 @@ def with the variables inlined as constants using:
 tf.app.flags.DEFINE_string(
     'dataset_dir', '', 'Directory to save intermediate dataset files to')
 
+tf.app.flags.DEFINE_bool(
+    'quantize', False, 'whether to use quantized graph or not.')
+
 FLAGS = tf.app.flags.FLAGS
 
 
@@ -115,6 +118,10 @@ def main(_):
                                  shape=[FLAGS.batch_size, image_size,
                                         image_size, 3])
     network_fn(placeholder)
+
+    if FLAGS.quantize:
+      tf.contrib.quantize.create_eval_graph()
+
     graph_def = graph.as_graph_def()
     with gfile.GFile(FLAGS.output_file, 'wb') as f:
       f.write(graph_def.SerializeToString())

research/slim/nets/mobilenet/mobilenet_v2.py

@@ -91,6 +91,7 @@ def mobilenet(input_tensor,
               finegrain_classification_mode=False,
               min_depth=None,
               divisible_by=None,
+              activation_fn=None,
               **kwargs):
   """Creates mobilenet V2 network.
 
@@ -117,6 +118,8 @@ def mobilenet(input_tensor,
     many channels after application of depth multiplier.
     divisible_by: If provided will ensure that all layers # channels
     will be divisible by this number.
+    activation_fn: Activation function to use, defaults to tf.nn.relu6 if not
+      specified.
     **kwargs: passed directly to mobilenet.mobilenet:
       prediction_fn- what prediction function to use.
       reuse-: whether to reuse variables (if reuse set to true, scope
@@ -136,6 +139,12 @@ def mobilenet(input_tensor,
     conv_defs = copy.deepcopy(conv_defs)
     if depth_multiplier < 1:
       conv_defs['spec'][-1].params['num_outputs'] /= depth_multiplier
+  if activation_fn:
+    conv_defs = copy.deepcopy(conv_defs)
+    defaults = conv_defs['defaults']
+    conv_defaults = (
+        defaults[(slim.conv2d, slim.fully_connected, slim.separable_conv2d)])
+    conv_defaults['activation_fn'] = activation_fn
 
   depth_args = {}
   # NB: do not set depth_args unless they are provided to avoid overriding
@@ -154,6 +163,8 @@ def mobilenet(input_tensor,
         multiplier=depth_multiplier,
         **kwargs)
 
+mobilenet.default_image_size = 224
+
 
 def wrapped_partial(func, *args, **kwargs):
   partial_func = functools.partial(func, *args, **kwargs)

research/slim/nets/mobilenet_v1.py

@@ -120,8 +120,8 @@
 Conv = namedtuple('Conv', ['kernel', 'stride', 'depth'])
 DepthSepConv = namedtuple('DepthSepConv', ['kernel', 'stride', 'depth'])
 
-# _CONV_DEFS specifies the MobileNet body
-_CONV_DEFS = [
+# MOBILENETV1_CONV_DEFS specifies the MobileNet body
+MOBILENETV1_CONV_DEFS = [
     Conv(kernel=[3, 3], stride=2, depth=32),
     DepthSepConv(kernel=[3, 3], stride=1, depth=64),
     DepthSepConv(kernel=[3, 3], stride=2, depth=128),
@@ -221,7 +221,7 @@ def mobilenet_v1_base(inputs,
     raise ValueError('depth_multiplier is not greater than zero.')
 
   if conv_defs is None:
-    conv_defs = _CONV_DEFS
+    conv_defs = MOBILENETV1_CONV_DEFS
 
   if output_stride is not None and output_stride not in [8, 16, 32]:
     raise ValueError('Only allowed output_stride values are 8, 16, 32.')

research/slim/nets/nasnet/nasnet.py

@@ -245,7 +245,7 @@ def _build_aux_head(net, end_points, num_classes, hparams, scope):
       end_points['AuxLogits'] = aux_logits
 
 
-def _imagenet_stem(inputs, hparams, stem_cell):
+def _imagenet_stem(inputs, hparams, stem_cell, current_step=None):
   """Stem used for models trained on ImageNet."""
   num_stem_cells = 2
 
@@ -266,7 +266,8 @@ def _imagenet_stem(inputs, hparams, stem_cell):
         filter_scaling=filter_scaling,
         stride=2,
         prev_layer=cell_outputs[-2],
-        cell_num=cell_num)
+        cell_num=cell_num,
+        current_step=current_step)
     cell_outputs.append(net)
     filter_scaling *= hparams.filter_scaling_rate
   return net, cell_outputs
@@ -286,7 +287,8 @@ def _cifar_stem(inputs, hparams):
 
 def build_nasnet_cifar(images, num_classes,
                        is_training=True,
-                       config=None):
+                       config=None,
+                       current_step=None):
   """Build NASNet model for the Cifar Dataset."""
   hparams = cifar_config() if config is None else copy.deepcopy(config)
   _update_hparams(hparams, is_training)
@@ -326,14 +328,16 @@ def build_nasnet_cifar(images, num_classes,
                                 num_classes=num_classes,
                                 hparams=hparams,
                                 is_training=is_training,
-                                stem_type='cifar')
+                                stem_type='cifar',
+                                current_step=current_step)
 build_nasnet_cifar.default_image_size = 32
 
 
 def build_nasnet_mobile(images, num_classes,
                         is_training=True,
                         final_endpoint=None,
-                        config=None):
+                        config=None,
+                        current_step=None):
   """Build NASNet Mobile model for the ImageNet Dataset."""
   hparams = (mobile_imagenet_config() if config is None
              else copy.deepcopy(config))
@@ -377,14 +381,16 @@ def build_nasnet_mobile(images, num_classes,
                                 hparams=hparams,
                                 is_training=is_training,
                                 stem_type='imagenet',
-                                final_endpoint=final_endpoint)
+                                final_endpoint=final_endpoint,
+                                current_step=current_step)
 build_nasnet_mobile.default_image_size = 224
 
 
 def build_nasnet_large(images, num_classes,
                        is_training=True,
                        final_endpoint=None,
-                       config=None):
+                       config=None,
+                       current_step=None):
   """Build NASNet Large model for the ImageNet Dataset."""
   hparams = (large_imagenet_config() if config is None
              else copy.deepcopy(config))
@@ -428,7 +434,8 @@ def build_nasnet_large(images, num_classes,
                                 hparams=hparams,
                                 is_training=is_training,
                                 stem_type='imagenet',
-                                final_endpoint=final_endpoint)
+                                final_endpoint=final_endpoint,
+                                current_step=current_step)
 build_nasnet_large.default_image_size = 331
 
 
@@ -439,7 +446,8 @@ def _build_nasnet_base(images,
                        hparams,
                        is_training,
                        stem_type,
-                       final_endpoint=None):
+                       final_endpoint=None,
+                       current_step=None):
   """Constructs a NASNet image model."""
 
   end_points = {}
@@ -482,7 +490,8 @@ def add_and_check_endpoint(endpoint_name, net):
           filter_scaling=filter_scaling,
           stride=2,
           prev_layer=cell_outputs[-2],
-          cell_num=true_cell_num)
+          cell_num=true_cell_num,
+          current_step=current_step)
       if add_and_check_endpoint(
           'Reduction_Cell_{}'.format(reduction_indices.index(cell_num)), net):
         return net, end_points
@@ -496,7 +505,8 @@ def add_and_check_endpoint(endpoint_name, net):
         filter_scaling=filter_scaling,
         stride=stride,
         prev_layer=prev_layer,
-        cell_num=true_cell_num)
+        cell_num=true_cell_num,
+        current_step=current_step)
 
     if add_and_check_endpoint('Cell_{}'.format(cell_num), net):
       return net, end_points

research/slim/nets/nasnet/nasnet_test.py

@@ -371,6 +371,24 @@ def testOverrideHParamsLargeModel(self):
     self.assertListEqual(
         end_points['Stem'].shape.as_list(), [batch_size, 336, 42, 42])
 
+  def testCurrentStepCifarModel(self):
+    batch_size = 5
+    height, width = 32, 32
+    num_classes = 10
+    inputs = tf.random_uniform((batch_size, height, width, 3))
+    global_step = tf.train.create_global_step()
+    with slim.arg_scope(nasnet.nasnet_cifar_arg_scope()):
+      logits, end_points = nasnet.build_nasnet_cifar(inputs,
+                                                     num_classes,
+                                                     current_step=global_step)
+    auxlogits = end_points['AuxLogits']
+    predictions = end_points['Predictions']
+    self.assertListEqual(auxlogits.get_shape().as_list(),
+                         [batch_size, num_classes])
+    self.assertListEqual(logits.get_shape().as_list(),
+                         [batch_size, num_classes])
+    self.assertListEqual(predictions.get_shape().as_list(),
+                         [batch_size, num_classes])
 
 if __name__ == '__main__':
   tf.test.main()

research/slim/nets/nasnet/nasnet_utils.py

@@ -300,7 +300,7 @@ def _cell_base(self, net, prev_layer):
     return net
 
   def __call__(self, net, scope=None, filter_scaling=1, stride=1,
-               prev_layer=None, cell_num=-1):
+               prev_layer=None, cell_num=-1, current_step=None):
     """Runs the conv cell."""
     self._cell_num = cell_num
     self._filter_scaling = filter_scaling
@@ -325,10 +325,12 @@ def __call__(self, net, scope=None, filter_scaling=1, stride=1,
           # Apply conv operations
           with tf.variable_scope('left'):
             h1 = self._apply_conv_operation(h1, operation_left,
-                                            stride, original_input_left)
+                                            stride, original_input_left,
+                                            current_step)
           with tf.variable_scope('right'):
             h2 = self._apply_conv_operation(h2, operation_right,
-                                            stride, original_input_right)
+                                            stride, original_input_right,
+                                            current_step)
 
           # Combine hidden states using 'add'.
           with tf.variable_scope('combine'):
@@ -343,7 +345,7 @@ def __call__(self, net, scope=None, filter_scaling=1, stride=1,
       return net
 
   def _apply_conv_operation(self, net, operation,
-                            stride, is_from_original_input):
+                            stride, is_from_original_input, current_step):
     """Applies the predicted conv operation to net."""
     # Dont stride if this is not one of the original hiddenstates
     if stride > 1 and not is_from_original_input:
@@ -367,7 +369,7 @@ def _apply_conv_operation(self, net, operation,
       raise ValueError('Unimplemented operation', operation)
 
     if operation != 'none':
-      net = self._apply_drop_path(net)
+      net = self._apply_drop_path(net, current_step=current_step)
     return net
 
   def _combine_unused_states(self, net):
@@ -433,9 +435,9 @@ def _apply_drop_path(self, net, current_step=None,
         drop_path_keep_prob = 1 - layer_ratio * (1 - drop_path_keep_prob)
       if drop_connect_version in ['v1', 'v3']:
         # Decrease the keep probability over time
-        if not current_step:
-          current_step = tf.cast(tf.train.get_or_create_global_step(),
-                                 tf.float32)
+        if current_step is None:
+          current_step = tf.train.get_or_create_global_step()
+        current_step = tf.cast(current_step, tf.float32)
         drop_path_burn_in_steps = self._total_training_steps
         current_ratio = current_step / drop_path_burn_in_steps
         current_ratio = tf.minimum(1.0, current_ratio)

research/slim/train_image_classifier.py

@@ -124,6 +124,11 @@
 
 tf.app.flags.DEFINE_float('rmsprop_decay', 0.9, 'Decay term for RMSProp.')
 
+tf.app.flags.DEFINE_integer(
+    'quantize_delay', -1,
+    'Number of steps to start quantized training. Set to -1 would disable '
+    'quantized training.')
+
 #######################
 # Learning Rate Flags #
 #######################
@@ -511,6 +516,10 @@ def clone_fn(batch_queue):
     else:
       moving_average_variables, variable_averages = None, None
 
+    if FLAGS.quantize_delay >= 0:
+      tf.contrib.quantize.create_training_graph(
+          quant_delay=FLAGS.quantize_delay)
+
     #########################################
     # Configure the optimization procedure. #
     #########################################