Release MobileDet code and model, and require tf_slim installation for OD API. (#8562)

* Merged commit includes the following changes:
311933687  by Sergio Guadarrama:

    Removes spurios use of tf.compat.v2, which results in spurious tf.compat.v1.compat.v2. Adds basic test to nasnet_utils.
    Replaces all remaining import tensorflow as tf with import tensorflow.compat.v1 as tf

--
311766063  by Sergio Guadarrama:

    Removes explicit tf.compat.v1 in all call sites (we already import tf.compat.v1, so this code was  doing tf.compat.v1.compat.v1). The existing code worked in latest version of tensorflow, 2.2, (and 1.15) but not in 1.14 or in 2.0.0a, this CL fixes it.

--
311624958  by Sergio Guadarrama:

    Updates README that doesn't render properly in github documentation

--
310980959  by Sergio Guadarrama:

    Moves research_models/slim off tf.contrib.slim/layers/framework to tf_slim

--
310263156  by Sergio Guadarrama:

    Adds model breakdown for MobilenetV3

--
308640516  by Sergio Guadarrama:

    Internal change

308244396  by Sergio Guadarrama:

    GroupNormalization support for MobilenetV3.

--
307475800  by Sergio Guadarrama:

    Internal change

--
302077708  by Sergio Guadarrama:

    Remove `disable_tf2` behavior from slim py_library targets

--
301208453  by Sergio Guadarrama:

    Automated refactoring to make code Python 3 compatible.

--
300816672  by Sergio Guadarrama:

    Internal change

299433840  by Sergio Guadarrama:

    Internal change

299221609  by Sergio Guadarrama:

    Explicitly disable Tensorflow v2 behaviors for all TF1.x binaries and tests

--
299179617  by Sergio Guadarrama:

    Internal change

299040784  by Sergio Guadarrama:

    Internal change

299036699  by Sergio Guadarrama:

    Internal change

298736510  by Sergio Guadarrama:

    Internal change

298732599  by Sergio Guadarrama:

    Internal change

298729507  by Sergio Guadarrama:

    Internal change

298253328  by Sergio Guadarrama:

    Internal change

297788346  by Sergio Guadarrama:

    Internal change

297785278  by Sergio Guadarrama:

    Internal change

297783127  by Sergio Guadarrama:

    Internal change

297725870  by Sergio Guadarrama:

    Internal change

297721811  by Sergio Guadarrama:

    Internal change

297711347  by Sergio Guadarrama:

    Internal change

297708059  by Sergio Guadarrama:

    Internal change

297701831  by Sergio Guadarrama:

    Internal change

297700038  by Sergio Guadarrama:

    Internal change

297670468  by Sergio Guadarrama:

    Internal change.

--
297350326  by Sergio Guadarrama:

    Explicitly replace "import tensorflow" with "tensorflow.compat.v1" for TF2.x migration

--
297201668  by Sergio Guadarrama:

    Explicitly replace "import tensorflow" with "tensorflow.compat.v1" for TF2.x migration

--
294483372  by Sergio Guadarrama:

    Internal change

PiperOrigin-RevId: 311933687

* Merged commit includes the following changes:
312578615  by Menglong Zhu:

    Modify the LSTM feature extractors to be python 3 compatible.

--
311264357  by Menglong Zhu:

    Removes contrib.slim

--
308957207  by Menglong Zhu:

    Automated refactoring to make code Python 3 compatible.

--
306976470  by yongzhe:

    Internal change

306777559  by Menglong Zhu:

    Internal change

--
299232507  by lzyuan:

    Internal update.

--
299221735  by lzyuan:

    Add small epsilon on max_range for quantize_op to prevent range collapse.

--

PiperOrigin-RevId: 312578615

* Merged commit includes the following changes:
310447280  by lzc:

    Internal changes.

--

PiperOrigin-RevId: 310447280
Co-authored-by: Sergio Guadarrama <sguada@google.com>
Co-authored-by: Menglong Zhu <menglong@google.com>

research/object_detection/metrics/oid_challenge_evaluation_utils_test.py

@@ -25,7 +25,7 @@ import numpy as np
 import pandas as pd
 from pycocotools import mask as coco_mask
 import six
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 from object_detection.core import standard_fields
 from object_detection.metrics import oid_challenge_evaluation_utils as utils

research/object_detection/metrics/oid_vrd_challenge_evaluation_utils_test.py

@@ -20,7 +20,7 @@ from __future__ import print_function
 
 import numpy as np
 import pandas as pd
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 from object_detection.core import standard_fields
 from object_detection.metrics import oid_vrd_challenge_evaluation_utils as utils
 from object_detection.utils import vrd_evaluation

research/object_detection/metrics/tf_example_parser_test.py

@@ -16,7 +16,7 @@
 
 import numpy as np
 import numpy.testing as np_testing
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 from object_detection.core import standard_fields as fields
 from object_detection.metrics import tf_example_parser

research/object_detection/model_lib.py

@@ -22,7 +22,9 @@ import copy
 import functools
 import os
 
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
+import tf_slim as slim
+
 
 from object_detection import eval_util
 from object_detection import exporter as exporter_lib
@@ -40,11 +42,8 @@ from object_detection.utils import visualization_utils as vis_utils
 
 # pylint: disable=g-import-not-at-top
 try:
-  from tensorflow.contrib import framework as contrib_framework
-  from tensorflow.contrib import layers as contrib_layers
   from tensorflow.contrib import learn as contrib_learn
   from tensorflow.contrib import tpu as contrib_tpu
-  from tensorflow.contrib import training as contrib_training
 except ImportError:
   # TF 2.0 doesn't ship with contrib.
   pass
@@ -95,6 +94,10 @@ def _prepare_groundtruth_for_eval(detection_model, class_agnostic,
         of groundtruth boxes per image..
       'groundtruth_keypoints': [batch_size, num_boxes, num_keypoints, 2] float32
         tensor of keypoints (if provided in groundtruth).
+      'groundtruth_group_of': [batch_size, num_boxes] bool tensor indicating
+        group_of annotations (if provided in groundtruth).
+      'groundtruth_labeled_classes': [batch_size, num_classes] int64
+        tensor of 1-indexed classes.
     class_agnostic: Boolean indicating whether detections are class agnostic.
   """
   input_data_fields = fields.InputDataFields()
@@ -138,6 +141,29 @@ def _prepare_groundtruth_for_eval(detection_model, class_agnostic,
         detection_model.groundtruth_lists(
             fields.BoxListFields.keypoint_visibilities))
 
+  if detection_model.groundtruth_has_field(fields.BoxListFields.group_of):
+    groundtruth[input_data_fields.groundtruth_group_of] = tf.stack(
+        detection_model.groundtruth_lists(fields.BoxListFields.group_of))
+
+  if detection_model.groundtruth_has_field(
+      fields.InputDataFields.groundtruth_labeled_classes):
+    labeled_classes_list = detection_model.groundtruth_lists(
+        fields.InputDataFields.groundtruth_labeled_classes)
+    labeled_classes = [
+        tf.where(x)[:, 0] + label_id_offset for x in labeled_classes_list
+    ]
+    if len(labeled_classes) > 1:
+      num_classes = labeled_classes_list[0].shape[0]
+      padded_labeled_classes = []
+      for x in labeled_classes:
+        padding = num_classes - tf.shape(x)[0]
+        padded_labeled_classes.append(tf.pad(x, [[0, padding]]))
+      groundtruth[input_data_fields.groundtruth_labeled_classes] = tf.stack(
+          padded_labeled_classes)
+    else:
+      groundtruth[input_data_fields.groundtruth_labeled_classes] = tf.stack(
+          labeled_classes)
+
   groundtruth[input_data_fields.num_groundtruth_boxes] = (
       tf.tile([max_number_of_boxes], multiples=[groundtruth_boxes_shape[0]]))
   return groundtruth
@@ -213,6 +239,7 @@ def unstack_batch(tensor_dict, unpad_groundtruth_tensors=True):
         unpadded_tensor = tf.slice(padded_tensor, slice_begin, slice_size)
         unpadded_tensor_list.append(unpadded_tensor)
       unbatched_unpadded_tensor_dict[key] = unpadded_tensor_list
+
     unbatched_tensor_dict.update(unbatched_unpadded_tensor_dict)
 
   return unbatched_tensor_dict
@@ -252,6 +279,9 @@ def provide_groundtruth(model, labels):
   gt_is_crowd_list = None
   if fields.InputDataFields.groundtruth_is_crowd in labels:
     gt_is_crowd_list = labels[fields.InputDataFields.groundtruth_is_crowd]
+  gt_group_of_list = None
+  if fields.InputDataFields.groundtruth_group_of in labels:
+    gt_group_of_list = labels[fields.InputDataFields.groundtruth_group_of]
   gt_area_list = None
   if fields.InputDataFields.groundtruth_area in labels:
     gt_area_list = labels[fields.InputDataFields.groundtruth_area]
@@ -269,6 +299,7 @@ def provide_groundtruth(model, labels):
       groundtruth_keypoint_visibilities_list=gt_keypoint_visibilities_list,
       groundtruth_weights_list=gt_weights_list,
       groundtruth_is_crowd_list=gt_is_crowd_list,
+      groundtruth_group_of_list=gt_group_of_list,
       groundtruth_area_list=gt_area_list)
 
 
@@ -447,7 +478,7 @@ def create_model_fn(detection_model_fn, configs, hparams, use_tpu=False,
       exclude_variables = (
           train_config.freeze_variables
           if train_config.freeze_variables else None)
-      trainable_variables = contrib_framework.filter_variables(
+      trainable_variables = slim.filter_variables(
           tf.trainable_variables(),
           include_patterns=include_variables,
           exclude_patterns=exclude_variables)
@@ -462,7 +493,7 @@ def create_model_fn(detection_model_fn, configs, hparams, use_tpu=False,
       summaries = [] if use_tpu else None
       if train_config.summarize_gradients:
         summaries = ['gradients', 'gradient_norm', 'global_gradient_norm']
-      train_op = contrib_layers.optimize_loss(
+      train_op = slim.optimizers.optimize_loss(
           loss=total_loss,
           global_step=global_step,
           learning_rate=None,
@@ -826,7 +857,48 @@ def create_train_and_eval_specs(train_input_fn,
   return train_spec, eval_specs
 
 
-def continuous_eval(estimator, model_dir, input_fn, train_steps, name):
+def _evaluate_checkpoint(estimator,
+                         input_fn,
+                         checkpoint_path,
+                         name,
+                         max_retries=0):
+  """Evaluates a checkpoint.
+
+  Args:
+    estimator: Estimator object to use for evaluation.
+    input_fn: Input function to use for evaluation.
+    checkpoint_path: Path of the checkpoint to evaluate.
+    name: Namescope for eval summary.
+    max_retries: Maximum number of times to retry the evaluation on encountering
+      a tf.errors.InvalidArgumentError. If negative, will always retry the
+      evaluation.
+
+  Returns:
+    Estimator evaluation results.
+  """
+  always_retry = True if max_retries < 0 else False
+  retries = 0
+  while always_retry or retries <= max_retries:
+    try:
+      return estimator.evaluate(
+          input_fn=input_fn,
+          steps=None,
+          checkpoint_path=checkpoint_path,
+          name=name)
+    except tf.errors.InvalidArgumentError as e:
+      if always_retry or retries < max_retries:
+        tf.logging.info('Retrying checkpoint evaluation after exception: %s', e)
+        retries += 1
+      else:
+        raise e
+
+
+def continuous_eval(estimator,
+                    model_dir,
+                    input_fn,
+                    train_steps,
+                    name,
+                    max_retries=0):
   """Perform continuous evaluation on checkpoints written to a model directory.
 
   Args:
@@ -836,20 +908,27 @@ def continuous_eval(estimator, model_dir, input_fn, train_steps, name):
     train_steps: Number of training steps. This is used to infer the last
       checkpoint and stop evaluation loop.
     name: Namescope for eval summary.
+    max_retries: Maximum number of times to retry the evaluation on encountering
+      a tf.errors.InvalidArgumentError. If negative, will always retry the
+      evaluation.
   """
 
   def terminate_eval():
     tf.logging.info('Terminating eval after 180 seconds of no checkpoints')
     return True
 
-  for ckpt in contrib_training.checkpoints_iterator(
+  for ckpt in tf.train.checkpoints_iterator(
       model_dir, min_interval_secs=180, timeout=None,
       timeout_fn=terminate_eval):
 
     tf.logging.info('Starting Evaluation.')
     try:
-      eval_results = estimator.evaluate(
-          input_fn=input_fn, steps=None, checkpoint_path=ckpt, name=name)
+      eval_results = _evaluate_checkpoint(
+          estimator=estimator,
+          input_fn=input_fn,
+          checkpoint_path=ckpt,
+          name=name,
+          max_retries=max_retries)
       tf.logging.info('Eval results: %s' % eval_results)
 
       # Terminate eval job when final checkpoint is reached

research/object_detection/model_lib_test.py

@@ -22,7 +22,7 @@ import functools
 import os
 
 import numpy as np
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 from tensorflow.contrib.tpu.python.tpu import tpu_config
 from tensorflow.contrib.tpu.python.tpu import tpu_estimator
@@ -42,11 +42,18 @@ MODEL_NAME_FOR_TEST = 'ssd_inception_v2_pets'
 # Model for testing keypoints.
 MODEL_NAME_FOR_KEYPOINTS_TEST = 'ssd_mobilenet_v1_fpp'
 
+# Model for testing tfSequenceExample inputs.
+MODEL_NAME_FOR_SEQUENCE_EXAMPLE_TEST = 'context_rcnn_camera_trap'
 
-def _get_data_path():
+
+def _get_data_path(model_name):
   """Returns an absolute path to TFRecord file."""
-  return os.path.join(tf.resource_loader.get_data_files_path(), 'test_data',
-                      'pets_examples.record')
+  if model_name == MODEL_NAME_FOR_SEQUENCE_EXAMPLE_TEST:
+    return os.path.join(tf.resource_loader.get_data_files_path(), 'test_data',
+                        'snapshot_serengeti_sequence_examples.record')
+  else:
+    return os.path.join(tf.resource_loader.get_data_files_path(), 'test_data',
+                        'pets_examples.record')
 
 
 def get_pipeline_config_path(model_name):
@@ -54,6 +61,9 @@ def get_pipeline_config_path(model_name):
   if model_name == MODEL_NAME_FOR_KEYPOINTS_TEST:
     return os.path.join(tf.resource_loader.get_data_files_path(), 'test_data',
                         model_name + '.config')
+  elif model_name == MODEL_NAME_FOR_SEQUENCE_EXAMPLE_TEST:
+    return os.path.join(tf.resource_loader.get_data_files_path(), 'test_data',
+                        model_name + '.config')
   else:
     return os.path.join(tf.resource_loader.get_data_files_path(), 'samples',
                         'configs', model_name + '.config')
@@ -71,12 +81,20 @@ def _get_keypoints_labelmap_path():
                       'face_person_with_keypoints_label_map.pbtxt')
 
 
+def _get_sequence_example_labelmap_path():
+  """Returns an absolute path to label map file."""
+  return os.path.join(tf.resource_loader.get_data_files_path(), 'data',
+                      'snapshot_serengeti_label_map.pbtxt')
+
+
 def _get_configs_for_model(model_name):
   """Returns configurations for model."""
   filename = get_pipeline_config_path(model_name)
-  data_path = _get_data_path()
+  data_path = _get_data_path(model_name)
   if model_name == MODEL_NAME_FOR_KEYPOINTS_TEST:
     label_map_path = _get_keypoints_labelmap_path()
+  elif model_name == MODEL_NAME_FOR_SEQUENCE_EXAMPLE_TEST:
+    label_map_path = _get_sequence_example_labelmap_path()
   else:
     label_map_path = _get_labelmap_path()
   configs = config_util.get_configs_from_pipeline_file(filename)
@@ -99,7 +117,7 @@ def _make_initializable_iterator(dataset):
   Returns:
     A `tf.data.Iterator`.
   """
-  iterator = dataset.make_initializable_iterator()
+  iterator = tf.data.make_initializable_iterator(dataset)
   tf.add_to_collection(tf.GraphKeys.TABLE_INITIALIZERS, iterator.initializer)
   return iterator
 
@@ -199,6 +217,11 @@ class ModelLibTest(tf.test.TestCase):
     configs = _get_configs_for_model(MODEL_NAME_FOR_TEST)
     self._assert_model_fn_for_train_eval(configs, 'train')
 
+  def test_model_fn_in_train_mode_sequences(self):
+    """Tests the model function in TRAIN mode."""
+    configs = _get_configs_for_model(MODEL_NAME_FOR_SEQUENCE_EXAMPLE_TEST)
+    self._assert_model_fn_for_train_eval(configs, 'train')
+
   def test_model_fn_in_train_mode_freeze_all_variables(self):
     """Tests model_fn TRAIN mode with all variables frozen."""
     configs = _get_configs_for_model(MODEL_NAME_FOR_TEST)
@@ -229,6 +252,11 @@ class ModelLibTest(tf.test.TestCase):
     configs = _get_configs_for_model(MODEL_NAME_FOR_TEST)
     self._assert_model_fn_for_train_eval(configs, 'eval')
 
+  def test_model_fn_in_eval_mode_sequences(self):
+    """Tests the model function in EVAL mode."""
+    configs = _get_configs_for_model(MODEL_NAME_FOR_SEQUENCE_EXAMPLE_TEST)
+    self._assert_model_fn_for_train_eval(configs, 'eval')
+
   def test_model_fn_in_keypoints_eval_mode(self):
     """Tests the model function in EVAL mode with keypoints config."""
     configs = _get_configs_for_model(MODEL_NAME_FOR_KEYPOINTS_TEST)
@@ -270,6 +298,27 @@ class ModelLibTest(tf.test.TestCase):
     self.assertIn('eval_input_fns', train_and_eval_dict)
     self.assertIn('eval_on_train_input_fn', train_and_eval_dict)
 
+  def test_create_estimator_and_inputs_sequence_example(self):
+    """Tests that Estimator and input function are constructed correctly."""
+    run_config = tf.estimator.RunConfig()
+    hparams = model_hparams.create_hparams(
+        hparams_overrides='load_pretrained=false')
+    pipeline_config_path = get_pipeline_config_path(
+        MODEL_NAME_FOR_SEQUENCE_EXAMPLE_TEST)
+    train_steps = 20
+    train_and_eval_dict = model_lib.create_estimator_and_inputs(
+        run_config,
+        hparams,
+        pipeline_config_path,
+        train_steps=train_steps)
+    estimator = train_and_eval_dict['estimator']
+    train_steps = train_and_eval_dict['train_steps']
+    self.assertIsInstance(estimator, tf.estimator.Estimator)
+    self.assertEqual(20, train_steps)
+    self.assertIn('train_input_fn', train_and_eval_dict)
+    self.assertIn('eval_input_fns', train_and_eval_dict)
+    self.assertIn('eval_on_train_input_fn', train_and_eval_dict)
+
   def test_create_estimator_with_default_train_eval_steps(self):
     """Tests that number of train/eval defaults to config values."""
     run_config = tf.estimator.RunConfig()

research/object_detection/model_lib_v2.py

@@ -22,7 +22,7 @@ import copy
 import os
 import time
 
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 from object_detection import eval_util
 from object_detection import inputs
@@ -101,6 +101,10 @@ def _compute_losses_and_predictions_dicts(
           instance masks for objects.
         labels[fields.InputDataFields.groundtruth_keypoints] is a
           float32 tensor containing keypoints for each box.
+        labels[fields.InputDataFields.groundtruth_group_of] is a tf.bool tensor
+          containing group_of annotations.
+        labels[fields.InputDataFields.groundtruth_labeled_classes] is a float32
+          k-hot tensor of classes.
     add_regularization_loss: Whether or not to include the model's
       regularization loss in the losses dictionary.
 
@@ -199,6 +203,8 @@ def eager_train_step(detection_model,
         labels[fields.InputDataFields.groundtruth_keypoints] is a
           [batch_size, num_boxes, num_keypoints, 2] float32 tensor containing
           keypoints for each box.
+        labels[fields.InputDataFields.groundtruth_labeled_classes] is a float32
+          k-hot tensor of classes.
     unpad_groundtruth_tensors: A parameter passed to unstack_batch.
     optimizer: The training optimizer that will update the variables.
     learning_rate: The learning rate tensor for the current training step.

research/object_detection/model_lib_v2_test.py

@@ -23,7 +23,7 @@ import tempfile
 
 import numpy as np
 import six
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 from object_detection import inputs
 from object_detection import model_hparams

research/object_detection/model_main.py

@@ -20,7 +20,7 @@ from __future__ import print_function
 
 from absl import flags
 
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 from object_detection import model_hparams
 from object_detection import model_lib
@@ -53,6 +53,11 @@ flags.DEFINE_boolean(
     'run_once', False, 'If running in eval-only mode, whether to run just '
     'one round of eval vs running continuously (default).'
 )
+flags.DEFINE_integer(
+    'max_eval_retries', 0, 'If running continuous eval, the maximum number of '
+    'retries upon encountering tf.errors.InvalidArgumentError. If negative, '
+    'will always retry the evaluation.'
+)
 FLAGS = flags.FLAGS
 
 
@@ -91,7 +96,7 @@ def main(unused_argv):
                              FLAGS.checkpoint_dir))
     else:
       model_lib.continuous_eval(estimator, FLAGS.checkpoint_dir, input_fn,
-                                train_steps, name)
+                                train_steps, name, FLAGS.max_eval_retries)
   else:
     train_spec, eval_specs = model_lib.create_train_and_eval_specs(
         train_input_fn,

research/object_detection/model_tpu_main.py

@@ -23,7 +23,7 @@ from __future__ import division
 from __future__ import print_function
 
 from absl import flags
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 
 from object_detection import model_hparams
@@ -85,6 +85,11 @@ flags.DEFINE_string(
     'where event and checkpoint files will be written.')
 flags.DEFINE_string('pipeline_config_path', None, 'Path to pipeline config '
                     'file.')
+flags.DEFINE_integer(
+    'max_eval_retries', 0, 'If running continuous eval, the maximum number of '
+    'retries upon encountering tf.errors.InvalidArgumentError. If negative, '
+    'will always retry the evaluation.'
+)
 
 FLAGS = tf.flags.FLAGS
 
@@ -142,7 +147,7 @@ def main(unused_argv):
       # Currently only a single eval input is allowed.
       input_fn = eval_input_fns[0]
     model_lib.continuous_eval(estimator, FLAGS.model_dir, input_fn, train_steps,
-                              name)
+                              name, FLAGS.max_eval_retries)
 
 
 if __name__ == '__main__':

research/object_detection/models/embedded_ssd_mobilenet_v1_feature_extractor.py

@@ -16,8 +16,8 @@
 
 """Embedded-friendly SSDFeatureExtractor for MobilenetV1 features."""
 
-import tensorflow as tf
-from tensorflow.contrib import slim as contrib_slim
+import tensorflow.compat.v1 as tf
+import tf_slim as slim
 
 from object_detection.meta_architectures import ssd_meta_arch
 from object_detection.models import feature_map_generators
@@ -25,8 +25,6 @@ from object_detection.utils import context_manager
 from object_detection.utils import ops
 from nets import mobilenet_v1
 
-slim = contrib_slim
-
 
 class EmbeddedSSDMobileNetV1FeatureExtractor(ssd_meta_arch.SSDFeatureExtractor):
   """Embedded-friendly SSD Feature Extractor using MobilenetV1 features.

research/object_detection/models/embedded_ssd_mobilenet_v1_feature_extractor_test.py

@@ -15,7 +15,7 @@
 
 """Tests for embedded_ssd_mobilenet_v1_feature_extractor."""
 import numpy as np
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 from object_detection.models import embedded_ssd_mobilenet_v1_feature_extractor
 from object_detection.models import ssd_feature_extractor_test

research/object_detection/models/faster_rcnn_inception_resnet_v2_feature_extractor.py

@@ -22,15 +22,13 @@ as well as
 Huang et al. (https://arxiv.org/abs/1611.10012)
 """
 
-import tensorflow as tf
-from tensorflow.contrib import slim as contrib_slim
+import tensorflow.compat.v1 as tf
+import tf_slim as slim
 
 from object_detection.meta_architectures import faster_rcnn_meta_arch
 from object_detection.utils import variables_helper
 from nets import inception_resnet_v2
 
-slim = contrib_slim
-
 
 class FasterRCNNInceptionResnetV2FeatureExtractor(
     faster_rcnn_meta_arch.FasterRCNNFeatureExtractor):

research/object_detection/models/faster_rcnn_inception_resnet_v2_feature_extractor_test.py

@@ -15,7 +15,7 @@
 
 """Tests for models.faster_rcnn_inception_resnet_v2_feature_extractor."""
 
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 from object_detection.models import faster_rcnn_inception_resnet_v2_feature_extractor as frcnn_inc_res
 

research/object_detection/models/faster_rcnn_inception_resnet_v2_keras_feature_extractor.py

@@ -25,7 +25,7 @@ Huang et al. (https://arxiv.org/abs/1611.10012)
 # Skip pylint for this file because it times out
 # pylint: skip-file
 
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 from object_detection.meta_architectures import faster_rcnn_meta_arch
 from object_detection.models.keras_models import inception_resnet_v2

research/object_detection/models/faster_rcnn_inception_resnet_v2_keras_feature_extractor_test.py

@@ -15,7 +15,7 @@
 
 """Tests for models.faster_rcnn_inception_resnet_v2_keras_feature_extractor."""
 
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 from object_detection.models import faster_rcnn_inception_resnet_v2_keras_feature_extractor as frcnn_inc_res
 

research/object_detection/models/faster_rcnn_inception_v2_feature_extractor.py

@@ -18,14 +18,12 @@
 See "Rethinking the Inception Architecture for Computer Vision"
 https://arxiv.org/abs/1512.00567
 """
-import tensorflow as tf
-from tensorflow.contrib import slim as contrib_slim
+import tensorflow.compat.v1 as tf
+import tf_slim as slim
 
 from object_detection.meta_architectures import faster_rcnn_meta_arch
 from nets import inception_v2
 
-slim = contrib_slim
-
 
 def _batch_norm_arg_scope(list_ops,
                           use_batch_norm=True,

research/object_detection/models/faster_rcnn_inception_v2_feature_extractor_test.py

@@ -16,7 +16,7 @@
 """Tests for faster_rcnn_inception_v2_feature_extractor."""
 
 import numpy as np
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 from object_detection.models import faster_rcnn_inception_v2_feature_extractor as faster_rcnn_inception_v2
 

research/object_detection/models/faster_rcnn_mobilenet_v1_feature_extractor.py

@@ -16,15 +16,13 @@
 """Mobilenet v1 Faster R-CNN implementation."""
 import numpy as np
 
-import tensorflow as tf
-from tensorflow.contrib import slim as contrib_slim
+import tensorflow.compat.v1 as tf
+import tf_slim as slim
 
 from object_detection.meta_architectures import faster_rcnn_meta_arch
 from object_detection.utils import shape_utils
 from nets import mobilenet_v1
 
-slim = contrib_slim
-
 
 def _get_mobilenet_conv_no_last_stride_defs(conv_depth_ratio_in_percentage):
   if conv_depth_ratio_in_percentage not in [25, 50, 75, 100]:

research/object_detection/models/faster_rcnn_mobilenet_v1_feature_extractor_test.py

@@ -16,7 +16,7 @@
 """Tests for faster_rcnn_mobilenet_v1_feature_extractor."""
 
 import numpy as np
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
 
 from object_detection.models import faster_rcnn_mobilenet_v1_feature_extractor as faster_rcnn_mobilenet_v1
 

research/object_detection/models/faster_rcnn_nas_feature_extractor.py

@@ -26,17 +26,15 @@ from __future__ import division
 from __future__ import print_function
 
 from six.moves import range
-import tensorflow as tf
-from tensorflow.contrib import framework as contrib_framework
-from tensorflow.contrib import slim as contrib_slim
+import tensorflow.compat.v1 as tf
+import tf_slim as slim
 
 from object_detection.meta_architectures import faster_rcnn_meta_arch
 from object_detection.utils import variables_helper
 from nets.nasnet import nasnet
 from nets.nasnet import nasnet_utils
 
-arg_scope = contrib_framework.arg_scope
-slim = contrib_slim
+arg_scope = slim.arg_scope
 
 
 def nasnet_large_arg_scope_for_detection(is_batch_norm_training=False):