Merge branch 'panoptic-segmentation' into panoptic-deeplab-modeling

official/vision/detection/modeling/architecture/factory.py → official/legacy/detection/modeling/architecture/factory.py

@@ -18,12 +18,12 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from official.vision.detection.modeling.architecture import fpn
-from official.vision.detection.modeling.architecture import heads
-from official.vision.detection.modeling.architecture import identity
-from official.vision.detection.modeling.architecture import nn_ops
-from official.vision.detection.modeling.architecture import resnet
-from official.vision.detection.modeling.architecture import spinenet
+from official.legacy.detection.modeling.architecture import fpn
+from official.legacy.detection.modeling.architecture import heads
+from official.legacy.detection.modeling.architecture import identity
+from official.legacy.detection.modeling.architecture import nn_ops
+from official.legacy.detection.modeling.architecture import resnet
+from official.legacy.detection.modeling.architecture import spinenet
 
 
 def norm_activation_generator(params):

official/vision/detection/modeling/architecture/fpn.py → official/legacy/detection/modeling/architecture/fpn.py

@@ -28,8 +28,8 @@ import functools
 
 import tensorflow as tf
 
-from official.vision.detection.modeling.architecture import nn_ops
-from official.vision.detection.ops import spatial_transform_ops
+from official.legacy.detection.modeling.architecture import nn_ops
+from official.legacy.detection.ops import spatial_transform_ops
 
 
 class Fpn(object):
@@ -52,6 +52,7 @@ class Fpn(object):
       fpn_feat_dims: `int` number of filters in FPN layers.
       use_separable_conv: `bool`, if True use separable convolution for
         convolution in FPN layers.
+      activation: the activation function.
       use_batch_norm: 'bool', indicating whether batchnorm layers are added.
       norm_activation: an operation that includes a normalization layer
         followed by an optional activation layer.

official/vision/detection/modeling/architecture/heads.py → official/legacy/detection/modeling/architecture/heads.py

@@ -23,8 +23,8 @@ import functools
 import numpy as np
 import tensorflow as tf
 
-from official.vision.detection.modeling.architecture import nn_ops
-from official.vision.detection.ops import spatial_transform_ops
+from official.legacy.detection.modeling.architecture import nn_ops
+from official.legacy.detection.ops import spatial_transform_ops
 
 
 class RpnHead(tf.keras.layers.Layer):

official/vision/detection/modeling/architecture/nn_blocks.py → official/legacy/detection/modeling/architecture/nn_blocks.py

@@ -23,7 +23,6 @@ import tensorflow as tf
 from official.modeling import tf_utils
 
 
-@tf.keras.utils.register_keras_serializable(package='Vision')
 class ResidualBlock(tf.keras.layers.Layer):
   """A residual block."""
 
@@ -163,7 +162,6 @@ class ResidualBlock(tf.keras.layers.Layer):
     return self._activation_fn(x + shortcut)
 
 
-@tf.keras.utils.register_keras_serializable(package='Vision')
 class BottleneckBlock(tf.keras.layers.Layer):
   """A standard bottleneck block."""
 

official/vision/detection/modeling/architecture/nn_ops.py → official/legacy/detection/modeling/architecture/nn_ops.py

@@ -45,11 +45,11 @@ class NormActivation(tf.keras.layers.Layer):
         layer.
       init_zero: `bool` if True, initializes scale parameter of batch
         normalization with 0. If False, initialize it with 1.
-      fused: `bool` fused option in batch normalziation.
-      use_actiation: `bool`, whether to add the optional activation layer after
+      use_activation: `bool`, whether to add the optional activation layer after
         the batch normalization layer.
       activation: 'string', the type of the activation layer. Currently support
         `relu` and `swish`.
+      fused: `bool` fused option in batch normalziation.
       name: `str` name for the operation.
     """
     super(NormActivation, self).__init__(trainable=trainable)

official/vision/detection/modeling/architecture/resnet.py → official/legacy/detection/modeling/architecture/resnet.py

@@ -24,7 +24,7 @@ from __future__ import division
 from __future__ import print_function
 
 import tensorflow as tf
-from official.vision.detection.modeling.architecture import nn_ops
+from official.legacy.detection.modeling.architecture import nn_ops
 
 
 # TODO(b/140112644): Refactor the code with Keras style, i.e. build and call.
@@ -41,6 +41,7 @@ class Resnet(object):
 
     Args:
       resnet_depth: `int` depth of ResNet backbone model.
+      activation: the activation function.
       norm_activation: an operation that includes a normalization layer followed
         by an optional activation layer.
       data_format: `str` either "channels_first" for `[batch, channels, height,

official/vision/detection/modeling/architecture/spinenet.py → official/legacy/detection/modeling/architecture/spinenet.py

@@ -24,9 +24,8 @@ import math
 
 from absl import logging
 import tensorflow as tf
-
+from official.legacy.detection.modeling.architecture import nn_blocks
 from official.modeling import tf_utils
-from official.vision.detection.modeling.architecture import nn_blocks
 
 layers = tf.keras.layers
 
@@ -113,7 +112,6 @@ def build_block_specs(block_specs=None):
   return [BlockSpec(*b) for b in block_specs]
 
 
-@tf.keras.utils.register_keras_serializable(package='Vision')
 class SpineNet(tf.keras.Model):
   """Class to build SpineNet models."""
 

official/vision/detection/modeling/base_model.py → official/legacy/detection/modeling/base_model.py

@@ -19,13 +19,12 @@ from __future__ import division
 from __future__ import print_function
 
 import abc
-import functools
 import re
 
 import tensorflow as tf
-from official.vision.detection.modeling import checkpoint_utils
-from official.vision.detection.modeling import learning_rates
-from official.vision.detection.modeling import optimizers
+from official.legacy.detection.modeling import checkpoint_utils
+from official.legacy.detection.modeling import learning_rates
+from official.legacy.detection.modeling import optimizers
 
 
 def _make_filter_trainable_variables_fn(frozen_variable_prefix):

official/vision/detection/modeling/checkpoint_utils.py → official/legacy/detection/modeling/checkpoint_utils.py

@@ -33,8 +33,9 @@ def _build_assignment_map(keras_model,
                           prefix='',
                           skip_variables_regex=None,
                           var_to_shape_map=None):
-  """Compute an assignment mapping for loading older checkpoints into a Keras
+  """Builds the variable assignment map.
 
+  Compute an assignment mapping for loading older checkpoints into a Keras
   model. Variable names are remapped from the original TPUEstimator model to
   the new Keras name.
 
@@ -53,10 +54,12 @@ def _build_assignment_map(keras_model,
 
   checkpoint_names = []
   if var_to_shape_map:
+    # pylint: disable=g-long-lambda
     checkpoint_names = list(
         filter(
             lambda x: not x.endswith('Momentum') and not x.endswith(
                 'global_step'), var_to_shape_map.keys()))
+    # pylint: enable=g-long-lambda
 
   logging.info('Number of variables in the checkpoint %d',
                len(checkpoint_names))
@@ -77,7 +80,9 @@ def _build_assignment_map(keras_model,
       continue
 
     # Match name with variables in the checkpoint.
+    # pylint: disable=cell-var-from-loop
     match_names = list(filter(lambda x: x.endswith(var_name), checkpoint_names))
+    # pylint: enable=cell-var-from-loop
     try:
       if match_names:
         assert len(match_names) == 1, 'more then on matches for {}: {}'.format(

official/vision/detection/modeling/factory.py → official/legacy/detection/modeling/factory.py

@@ -15,10 +15,10 @@
 """Factory to build detection model."""
 
 
-from official.vision.detection.modeling import maskrcnn_model
-from official.vision.detection.modeling import olnmask_model
-from official.vision.detection.modeling import retinanet_model
-from official.vision.detection.modeling import shapemask_model
+from official.legacy.detection.modeling import maskrcnn_model
+from official.legacy.detection.modeling import olnmask_model
+from official.legacy.detection.modeling import retinanet_model
+from official.legacy.detection.modeling import shapemask_model
 
 
 def model_generator(params):

official/vision/detection/modeling/learning_rates.py → official/legacy/detection/modeling/learning_rates.py

@@ -18,8 +18,6 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-import functools
-
 import numpy as np
 import tensorflow as tf
 from official.modeling.hyperparams import params_dict

official/vision/detection/modeling/maskrcnn_model.py → official/legacy/detection/modeling/maskrcnn_model.py

@@ -20,17 +20,17 @@ from __future__ import print_function
 
 import tensorflow as tf
 
-from official.vision.detection.dataloader import anchor
-from official.vision.detection.dataloader import mode_keys
-from official.vision.detection.evaluation import factory as eval_factory
-from official.vision.detection.modeling import base_model
-from official.vision.detection.modeling import losses
-from official.vision.detection.modeling.architecture import factory
-from official.vision.detection.ops import postprocess_ops
-from official.vision.detection.ops import roi_ops
-from official.vision.detection.ops import spatial_transform_ops
-from official.vision.detection.ops import target_ops
-from official.vision.detection.utils import box_utils
+from official.legacy.detection.dataloader import anchor
+from official.legacy.detection.dataloader import mode_keys
+from official.legacy.detection.evaluation import factory as eval_factory
+from official.legacy.detection.modeling import base_model
+from official.legacy.detection.modeling import losses
+from official.legacy.detection.modeling.architecture import factory
+from official.legacy.detection.ops import postprocess_ops
+from official.legacy.detection.ops import roi_ops
+from official.legacy.detection.ops import spatial_transform_ops
+from official.legacy.detection.ops import target_ops
+from official.legacy.detection.utils import box_utils
 
 
 class MaskrcnnModel(base_model.Model):

official/vision/detection/modeling/olnmask_model.py → official/legacy/detection/modeling/olnmask_model.py

@@ -20,16 +20,16 @@ from __future__ import print_function
 
 import tensorflow as tf
 
-from official.vision.detection.dataloader import anchor
-from official.vision.detection.dataloader import mode_keys
-from official.vision.detection.modeling import losses
-from official.vision.detection.modeling.architecture import factory
-from official.vision.detection.modeling.maskrcnn_model import MaskrcnnModel
-from official.vision.detection.ops import postprocess_ops
-from official.vision.detection.ops import roi_ops
-from official.vision.detection.ops import spatial_transform_ops
-from official.vision.detection.ops import target_ops
-from official.vision.detection.utils import box_utils
+from official.legacy.detection.dataloader import anchor
+from official.legacy.detection.dataloader import mode_keys
+from official.legacy.detection.modeling import losses
+from official.legacy.detection.modeling.architecture import factory
+from official.legacy.detection.modeling.maskrcnn_model import MaskrcnnModel
+from official.legacy.detection.ops import postprocess_ops
+from official.legacy.detection.ops import roi_ops
+from official.legacy.detection.ops import spatial_transform_ops
+from official.legacy.detection.ops import target_ops
+from official.legacy.detection.utils import box_utils
 
 
 class OlnMaskModel(MaskrcnnModel):

official/vision/detection/modeling/optimizers.py → official/legacy/detection/modeling/optimizers.py

@@ -20,7 +20,6 @@ from __future__ import print_function
 
 import functools
 
-import numpy as np
 import tensorflow as tf
 
 

official/vision/detection/modeling/retinanet_model.py → official/legacy/detection/modeling/retinanet_model.py

@@ -20,12 +20,12 @@ from __future__ import print_function
 
 import tensorflow as tf
 
-from official.vision.detection.dataloader import mode_keys
-from official.vision.detection.evaluation import factory as eval_factory
-from official.vision.detection.modeling import base_model
-from official.vision.detection.modeling import losses
-from official.vision.detection.modeling.architecture import factory
-from official.vision.detection.ops import postprocess_ops
+from official.legacy.detection.dataloader import mode_keys
+from official.legacy.detection.evaluation import factory as eval_factory
+from official.legacy.detection.modeling import base_model
+from official.legacy.detection.modeling import losses
+from official.legacy.detection.modeling.architecture import factory
+from official.legacy.detection.ops import postprocess_ops
 
 
 class RetinanetModel(base_model.Model):
@@ -131,13 +131,13 @@ class RetinanetModel(base_model.Model):
     required_output_fields = ['cls_outputs', 'box_outputs']
     for field in required_output_fields:
       if field not in outputs:
-        raise ValueError('"%s" is missing in outputs, requried %s found %s',
-                         field, required_output_fields, outputs.keys())
+        raise ValueError('"%s" is missing in outputs, requried %s found %s' %
+                         (field, required_output_fields, outputs.keys()))
     required_label_fields = ['image_info', 'groundtruths']
     for field in required_label_fields:
       if field not in labels:
-        raise ValueError('"%s" is missing in outputs, requried %s found %s',
-                         field, required_label_fields, labels.keys())
+        raise ValueError('"%s" is missing in outputs, requried %s found %s' %
+                         (field, required_label_fields, labels.keys()))
     boxes, scores, classes, valid_detections = self._generate_detections_fn(
         outputs['box_outputs'], outputs['cls_outputs'], labels['anchor_boxes'],
         labels['image_info'][:, 1:2, :])

official/vision/detection/modeling/shapemask_model.py → official/legacy/detection/modeling/shapemask_model.py

@@ -20,14 +20,14 @@ from __future__ import print_function
 
 import tensorflow as tf
 
-from official.vision.detection.dataloader import anchor
-from official.vision.detection.dataloader import mode_keys
-from official.vision.detection.evaluation import factory as eval_factory
-from official.vision.detection.modeling import base_model
-from official.vision.detection.modeling import losses
-from official.vision.detection.modeling.architecture import factory
-from official.vision.detection.ops import postprocess_ops
-from official.vision.detection.utils import box_utils
+from official.legacy.detection.dataloader import anchor
+from official.legacy.detection.dataloader import mode_keys
+from official.legacy.detection.evaluation import factory as eval_factory
+from official.legacy.detection.modeling import base_model
+from official.legacy.detection.modeling import losses
+from official.legacy.detection.modeling.architecture import factory
+from official.legacy.detection.ops import postprocess_ops
+from official.legacy.detection.utils import box_utils
 
 
 class ShapeMaskModel(base_model.Model):