Skip to content

GitLab

Unverified Commit 451906e4 authored by pkulzc's avatar pkulzc Committed by GitHub
Browse files

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: default avatarSergio Guadarrama <sguada@google.com>
Co-authored-by: default avatarMenglong Zhu <menglong@google.com>
parent 73b5be67
Hide whitespace changes
Inline Side-by-side
Showing with 303 additions and 334 deletions
research/slim/eval_image_classifier.py
View file @ 451906e4
...@@ -19,16 +19,15 @@ from __future__ import division ...@@ -19,16 +19,15 @@ from __future__ import division
from __future__ import print_function from __future__ import print_function
import math import math
import tensorflow as tf import tensorflow.compat.v1 as tf
import tf_slim as slim
from tensorflow.contrib import quantize as contrib_quantize from tensorflow.contrib import quantize as contrib_quantize
from tensorflow.contrib import slim as contrib_slim
from datasets import dataset_factory from datasets import dataset_factory
from nets import nets_factory from nets import nets_factory
from preprocessing import preprocessing_factory from preprocessing import preprocessing_factory
slim = contrib_slim
tf.app.flags.DEFINE_integer( tf.app.flags.DEFINE_integer(
'batch_size', 100, 'The number of samples in each batch.') 'batch_size', 100, 'The number of samples in each batch.')
......
research/slim/export_inference_graph.py
View file @ 451906e4
...@@ -57,17 +57,15 @@ from __future__ import division ...@@ -57,17 +57,15 @@ from __future__ import division
from __future__ import print_function from __future__ import print_function
import os import os
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import quantize as contrib_quantize from tensorflow.contrib import quantize as contrib_quantize
from tensorflow.contrib import slim as contrib_slim
from tensorflow.python.platform import gfile from tensorflow.python.platform import gfile
from datasets import dataset_factory from datasets import dataset_factory
from nets import nets_factory from nets import nets_factory
slim = contrib_slim
tf.app.flags.DEFINE_string( tf.app.flags.DEFINE_string(
'model_name', 'inception_v3', 'The name of the architecture to save.') 'model_name', 'inception_v3', 'The name of the architecture to save.')
......
research/slim/export_inference_graph_test.py
View file @ 451906e4
...@@ -22,7 +22,7 @@ from __future__ import print_function ...@@ -22,7 +22,7 @@ from __future__ import print_function
import os import os
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.python.platform import gfile from tensorflow.python.platform import gfile
import export_inference_graph import export_inference_graph
......
research/slim/nets/alexnet.py
View file @ 451906e4
...@@ -36,20 +36,18 @@ from __future__ import absolute_import ...@@ -36,20 +36,18 @@ from __future__ import absolute_import
from __future__ import division from __future__ import division
from __future__ import print_function from __future__ import print_function
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import slim as contrib_slim import tf_slim as slim
slim = contrib_slim
# pylint: disable=g-long-lambda # pylint: disable=g-long-lambda
trunc_normal = lambda stddev: tf.compat.v1.truncated_normal_initializer( trunc_normal = lambda stddev: tf.truncated_normal_initializer(
0.0, stddev) 0.0, stddev)
def alexnet_v2_arg_scope(weight_decay=0.0005): def alexnet_v2_arg_scope(weight_decay=0.0005):
with slim.arg_scope([slim.conv2d, slim.fully_connected], with slim.arg_scope([slim.conv2d, slim.fully_connected],
activation_fn=tf.nn.relu, activation_fn=tf.nn.relu,
biases_initializer=tf.compat.v1.constant_initializer(0.1), biases_initializer=tf.constant_initializer(0.1),
weights_regularizer=slim.l2_regularizer(weight_decay)): weights_regularizer=slim.l2_regularizer(weight_decay)):
with slim.arg_scope([slim.conv2d], padding='SAME'): with slim.arg_scope([slim.conv2d], padding='SAME'):
with slim.arg_scope([slim.max_pool2d], padding='VALID') as arg_sc: with slim.arg_scope([slim.max_pool2d], padding='VALID') as arg_sc:
...@@ -97,7 +95,7 @@ def alexnet_v2(inputs, ...@@ -97,7 +95,7 @@ def alexnet_v2(inputs,
or None). or None).
end_points: a dict of tensors with intermediate activations. end_points: a dict of tensors with intermediate activations.
""" """
with tf.compat.v1.variable_scope(scope, 'alexnet_v2', [inputs]) as sc: with tf.variable_scope(scope, 'alexnet_v2', [inputs]) as sc:
end_points_collection = sc.original_name_scope + '_end_points' end_points_collection = sc.original_name_scope + '_end_points'
# Collect outputs for conv2d, fully_connected and max_pool2d. # Collect outputs for conv2d, fully_connected and max_pool2d.
with slim.arg_scope([slim.conv2d, slim.fully_connected, slim.max_pool2d], with slim.arg_scope([slim.conv2d, slim.fully_connected, slim.max_pool2d],
...@@ -116,7 +114,7 @@ def alexnet_v2(inputs, ...@@ -116,7 +114,7 @@ def alexnet_v2(inputs,
with slim.arg_scope( with slim.arg_scope(
[slim.conv2d], [slim.conv2d],
weights_initializer=trunc_normal(0.005), weights_initializer=trunc_normal(0.005),
biases_initializer=tf.compat.v1.constant_initializer(0.1)): biases_initializer=tf.constant_initializer(0.1)):
net = slim.conv2d(net, 4096, [5, 5], padding='VALID', net = slim.conv2d(net, 4096, [5, 5], padding='VALID',
scope='fc6') scope='fc6')
net = slim.dropout(net, dropout_keep_prob, is_training=is_training, net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
...@@ -137,7 +135,7 @@ def alexnet_v2(inputs, ...@@ -137,7 +135,7 @@ def alexnet_v2(inputs,
num_classes, [1, 1], num_classes, [1, 1],
activation_fn=None, activation_fn=None,
normalizer_fn=None, normalizer_fn=None,
biases_initializer=tf.compat.v1.zeros_initializer(), biases_initializer=tf.zeros_initializer(),
scope='fc8') scope='fc8')
if spatial_squeeze: if spatial_squeeze:
net = tf.squeeze(net, [1, 2], name='fc8/squeezed') net = tf.squeeze(net, [1, 2], name='fc8/squeezed')
......
research/slim/nets/alexnet_test.py
View file @ 451906e4
...@@ -17,13 +17,11 @@ from __future__ import absolute_import ...@@ -17,13 +17,11 @@ from __future__ import absolute_import
from __future__ import division from __future__ import division
from __future__ import print_function from __future__ import print_function
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import slim as contrib_slim import tf_slim as slim
from nets import alexnet from nets import alexnet
slim = contrib_slim
class AlexnetV2Test(tf.test.TestCase): class AlexnetV2Test(tf.test.TestCase):
...@@ -156,7 +154,7 @@ class AlexnetV2Test(tf.test.TestCase): ...@@ -156,7 +154,7 @@ class AlexnetV2Test(tf.test.TestCase):
logits, _ = alexnet.alexnet_v2(train_inputs) logits, _ = alexnet.alexnet_v2(train_inputs)
self.assertListEqual(logits.get_shape().as_list(), self.assertListEqual(logits.get_shape().as_list(),
[train_batch_size, num_classes]) [train_batch_size, num_classes])
tf.compat.v1.get_variable_scope().reuse_variables() tf.get_variable_scope().reuse_variables()
eval_inputs = tf.random.uniform( eval_inputs = tf.random.uniform(
(eval_batch_size, eval_height, eval_width, 3)) (eval_batch_size, eval_height, eval_width, 3))
logits, _ = alexnet.alexnet_v2(eval_inputs, is_training=False, logits, _ = alexnet.alexnet_v2(eval_inputs, is_training=False,
...@@ -173,7 +171,7 @@ class AlexnetV2Test(tf.test.TestCase): ...@@ -173,7 +171,7 @@ class AlexnetV2Test(tf.test.TestCase):
with self.test_session() as sess: with self.test_session() as sess:
inputs = tf.random.uniform((batch_size, height, width, 3)) inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = alexnet.alexnet_v2(inputs) logits, _ = alexnet.alexnet_v2(inputs)
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
output = sess.run(logits) output = sess.run(logits)
self.assertTrue(output.any()) self.assertTrue(output.any())
......
research/slim/nets/cifarnet.py
View file @ 451906e4
...@@ -18,13 +18,11 @@ from __future__ import absolute_import ...@@ -18,13 +18,11 @@ from __future__ import absolute_import
from __future__ import division from __future__ import division
from __future__ import print_function from __future__ import print_function
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import slim as contrib_slim import tf_slim as slim
slim = contrib_slim
# pylint: disable=g-long-lambda # pylint: disable=g-long-lambda
trunc_normal = lambda stddev: tf.compat.v1.truncated_normal_initializer( trunc_normal = lambda stddev: tf.truncated_normal_initializer(
stddev=stddev) stddev=stddev)
...@@ -63,7 +61,7 @@ def cifarnet(images, num_classes=10, is_training=False, ...@@ -63,7 +61,7 @@ def cifarnet(images, num_classes=10, is_training=False,
""" """
end_points = {} end_points = {}
with tf.compat.v1.variable_scope(scope, 'CifarNet', [images]): with tf.variable_scope(scope, 'CifarNet', [images]):
net = slim.conv2d(images, 64, [5, 5], scope='conv1') net = slim.conv2d(images, 64, [5, 5], scope='conv1')
end_points['conv1'] = net end_points['conv1'] = net
net = slim.max_pool2d(net, [2, 2], 2, scope='pool1') net = slim.max_pool2d(net, [2, 2], 2, scope='pool1')
...@@ -87,7 +85,7 @@ def cifarnet(images, num_classes=10, is_training=False, ...@@ -87,7 +85,7 @@ def cifarnet(images, num_classes=10, is_training=False,
logits = slim.fully_connected( logits = slim.fully_connected(
net, net,
num_classes, num_classes,
biases_initializer=tf.compat.v1.zeros_initializer(), biases_initializer=tf.zeros_initializer(),
weights_initializer=trunc_normal(1 / 192.0), weights_initializer=trunc_normal(1 / 192.0),
weights_regularizer=None, weights_regularizer=None,
activation_fn=None, activation_fn=None,
...@@ -111,12 +109,12 @@ def cifarnet_arg_scope(weight_decay=0.004): ...@@ -111,12 +109,12 @@ def cifarnet_arg_scope(weight_decay=0.004):
""" """
with slim.arg_scope( with slim.arg_scope(
[slim.conv2d], [slim.conv2d],
weights_initializer=tf.compat.v1.truncated_normal_initializer( weights_initializer=tf.truncated_normal_initializer(
stddev=5e-2), stddev=5e-2),
activation_fn=tf.nn.relu): activation_fn=tf.nn.relu):
with slim.arg_scope( with slim.arg_scope(
[slim.fully_connected], [slim.fully_connected],
biases_initializer=tf.compat.v1.constant_initializer(0.1), biases_initializer=tf.constant_initializer(0.1),
weights_initializer=trunc_normal(0.04), weights_initializer=trunc_normal(0.04),
weights_regularizer=slim.l2_regularizer(weight_decay), weights_regularizer=slim.l2_regularizer(weight_decay),
activation_fn=tf.nn.relu) as sc: activation_fn=tf.nn.relu) as sc:
......
research/slim/nets/cyclegan.py
View file @ 451906e4
...@@ -19,12 +19,9 @@ from __future__ import print_function ...@@ -19,12 +19,9 @@ from __future__ import print_function
import numpy as np import numpy as np
from six.moves import xrange # pylint: disable=redefined-builtin from six.moves import xrange # pylint: disable=redefined-builtin
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import framework as contrib_framework import tf_slim as slim
from tensorflow.contrib import layers as contrib_layers from tensorflow.python.framework import tensor_util
from tensorflow.contrib import util as contrib_util
layers = contrib_layers
def cyclegan_arg_scope(instance_norm_center=True, def cyclegan_arg_scope(instance_norm_center=True,
...@@ -55,13 +52,13 @@ def cyclegan_arg_scope(instance_norm_center=True, ...@@ -55,13 +52,13 @@ def cyclegan_arg_scope(instance_norm_center=True,
weights_regularizer = None weights_regularizer = None
if weight_decay and weight_decay > 0.0: if weight_decay and weight_decay > 0.0:
weights_regularizer = layers.l2_regularizer(weight_decay) weights_regularizer = slim.l2_regularizer(weight_decay)
with contrib_framework.arg_scope( with slim.arg_scope(
[layers.conv2d], [slim.conv2d],
normalizer_fn=layers.instance_norm, normalizer_fn=slim.instance_norm,
normalizer_params=instance_norm_params, normalizer_params=instance_norm_params,
weights_initializer=tf.compat.v1.random_normal_initializer( weights_initializer=tf.random_normal_initializer(
0, weights_init_stddev), 0, weights_init_stddev),
weights_regularizer=weights_regularizer) as sc: weights_regularizer=weights_regularizer) as sc:
return sc return sc
...@@ -91,7 +88,7 @@ def cyclegan_upsample(net, num_outputs, stride, method='conv2d_transpose', ...@@ -91,7 +88,7 @@ def cyclegan_upsample(net, num_outputs, stride, method='conv2d_transpose',
Raises: Raises:
ValueError: if `method` is not recognized. ValueError: if `method` is not recognized.
""" """
with tf.compat.v1.variable_scope('upconv'): with tf.variable_scope('upconv'):
net_shape = tf.shape(input=net) net_shape = tf.shape(input=net)
height = net_shape[1] height = net_shape[1]
width = net_shape[2] width = net_shape[2]
...@@ -106,19 +103,19 @@ def cyclegan_upsample(net, num_outputs, stride, method='conv2d_transpose', ...@@ -106,19 +103,19 @@ def cyclegan_upsample(net, num_outputs, stride, method='conv2d_transpose',
net, [stride[0] * height, stride[1] * width], net, [stride[0] * height, stride[1] * width],
method=tf.image.ResizeMethod.NEAREST_NEIGHBOR) method=tf.image.ResizeMethod.NEAREST_NEIGHBOR)
net = tf.pad(tensor=net, paddings=spatial_pad_1, mode=pad_mode) net = tf.pad(tensor=net, paddings=spatial_pad_1, mode=pad_mode)
net = layers.conv2d(net, num_outputs, kernel_size=[3, 3], padding='valid') net = slim.conv2d(net, num_outputs, kernel_size=[3, 3], padding='valid')
elif method == 'bilinear_upsample_conv': elif method == 'bilinear_upsample_conv':
net = tf.compat.v1.image.resize_bilinear( net = tf.image.resize_bilinear(
net, [stride[0] * height, stride[1] * width], net, [stride[0] * height, stride[1] * width],
align_corners=align_corners) align_corners=align_corners)
net = tf.pad(tensor=net, paddings=spatial_pad_1, mode=pad_mode) net = tf.pad(tensor=net, paddings=spatial_pad_1, mode=pad_mode)
net = layers.conv2d(net, num_outputs, kernel_size=[3, 3], padding='valid') net = slim.conv2d(net, num_outputs, kernel_size=[3, 3], padding='valid')
elif method == 'conv2d_transpose': elif method == 'conv2d_transpose':
# This corrects 1 pixel offset for images with even width and height. # This corrects 1 pixel offset for images with even width and height.
# conv2d is left aligned and conv2d_transpose is right aligned for even # conv2d is left aligned and conv2d_transpose is right aligned for even
# sized images (while doing 'SAME' padding). # sized images (while doing 'SAME' padding).
# Note: This doesn't reflect actual model in paper. # Note: This doesn't reflect actual model in paper.
net = layers.conv2d_transpose( net = slim.conv2d_transpose(
net, num_outputs, kernel_size=[3, 3], stride=stride, padding='valid') net, num_outputs, kernel_size=[3, 3], stride=stride, padding='valid')
net = net[:, 1:, 1:, :] net = net[:, 1:, 1:, :]
else: else:
...@@ -129,7 +126,7 @@ def cyclegan_upsample(net, num_outputs, stride, method='conv2d_transpose', ...@@ -129,7 +126,7 @@ def cyclegan_upsample(net, num_outputs, stride, method='conv2d_transpose',
def _dynamic_or_static_shape(tensor): def _dynamic_or_static_shape(tensor):
shape = tf.shape(input=tensor) shape = tf.shape(input=tensor)
static_shape = contrib_util.constant_value(shape) static_shape = tensor_util.constant_value(shape)
return static_shape if static_shape is not None else shape return static_shape if static_shape is not None else shape
...@@ -201,47 +198,46 @@ def cyclegan_generator_resnet(images, ...@@ -201,47 +198,46 @@ def cyclegan_generator_resnet(images,
dtype=np.int32) dtype=np.int32)
spatial_pad_3 = np.array([[0, 0], [3, 3], [3, 3], [0, 0]]) spatial_pad_3 = np.array([[0, 0], [3, 3], [3, 3], [0, 0]])
with contrib_framework.arg_scope(arg_scope_fn()): with slim.arg_scope(arg_scope_fn()):
########### ###########
# Encoder # # Encoder #
########### ###########
with tf.compat.v1.variable_scope('input'): with tf.variable_scope('input'):
# 7x7 input stage # 7x7 input stage
net = tf.pad(tensor=images, paddings=spatial_pad_3, mode='REFLECT') net = tf.pad(tensor=images, paddings=spatial_pad_3, mode='REFLECT')
net = layers.conv2d(net, num_filters, kernel_size=[7, 7], padding='VALID') net = slim.conv2d(net, num_filters, kernel_size=[7, 7], padding='VALID')
end_points['encoder_0'] = net end_points['encoder_0'] = net
with tf.compat.v1.variable_scope('encoder'): with tf.variable_scope('encoder'):
with contrib_framework.arg_scope([layers.conv2d], with slim.arg_scope([slim.conv2d],
kernel_size=kernel_size, kernel_size=kernel_size,
stride=2, stride=2,
activation_fn=tf.nn.relu, activation_fn=tf.nn.relu,
padding='VALID'): padding='VALID'):
net = tf.pad(tensor=net, paddings=paddings, mode='REFLECT') net = tf.pad(tensor=net, paddings=paddings, mode='REFLECT')
net = layers.conv2d(net, num_filters * 2) net = slim.conv2d(net, num_filters * 2)
end_points['encoder_1'] = net end_points['encoder_1'] = net
net = tf.pad(tensor=net, paddings=paddings, mode='REFLECT') net = tf.pad(tensor=net, paddings=paddings, mode='REFLECT')
net = layers.conv2d(net, num_filters * 4) net = slim.conv2d(net, num_filters * 4)
end_points['encoder_2'] = net end_points['encoder_2'] = net
################### ###################
# Residual Blocks # # Residual Blocks #
################### ###################
with tf.compat.v1.variable_scope('residual_blocks'): with tf.variable_scope('residual_blocks'):
with contrib_framework.arg_scope([layers.conv2d], with slim.arg_scope([slim.conv2d],
kernel_size=kernel_size, kernel_size=kernel_size,
stride=1, stride=1,
activation_fn=tf.nn.relu, activation_fn=tf.nn.relu,
padding='VALID'): padding='VALID'):
for block_id in xrange(num_resnet_blocks): for block_id in xrange(num_resnet_blocks):
with tf.compat.v1.variable_scope('block_{}'.format(block_id)): with tf.variable_scope('block_{}'.format(block_id)):
res_net = tf.pad(tensor=net, paddings=paddings, mode='REFLECT') res_net = tf.pad(tensor=net, paddings=paddings, mode='REFLECT')
res_net = layers.conv2d(res_net, num_filters * 4) res_net = slim.conv2d(res_net, num_filters * 4)
res_net = tf.pad(tensor=res_net, paddings=paddings, mode='REFLECT') res_net = tf.pad(tensor=res_net, paddings=paddings, mode='REFLECT')
res_net = layers.conv2d(res_net, num_filters * 4, res_net = slim.conv2d(res_net, num_filters * 4, activation_fn=None)
activation_fn=None)
net += res_net net += res_net
end_points['resnet_block_%d' % block_id] = net end_points['resnet_block_%d' % block_id] = net
...@@ -249,24 +245,24 @@ def cyclegan_generator_resnet(images, ...@@ -249,24 +245,24 @@ def cyclegan_generator_resnet(images,
########### ###########
# Decoder # # Decoder #
########### ###########
with tf.compat.v1.variable_scope('decoder'): with tf.variable_scope('decoder'):
with contrib_framework.arg_scope([layers.conv2d], with slim.arg_scope([slim.conv2d],
kernel_size=kernel_size, kernel_size=kernel_size,
stride=1, stride=1,
activation_fn=tf.nn.relu): activation_fn=tf.nn.relu):
with tf.compat.v1.variable_scope('decoder1'): with tf.variable_scope('decoder1'):
net = upsample_fn(net, num_outputs=num_filters * 2, stride=[2, 2]) net = upsample_fn(net, num_outputs=num_filters * 2, stride=[2, 2])
end_points['decoder1'] = net end_points['decoder1'] = net
with tf.compat.v1.variable_scope('decoder2'): with tf.variable_scope('decoder2'):
net = upsample_fn(net, num_outputs=num_filters, stride=[2, 2]) net = upsample_fn(net, num_outputs=num_filters, stride=[2, 2])
end_points['decoder2'] = net end_points['decoder2'] = net
with tf.compat.v1.variable_scope('output'): with tf.variable_scope('output'):
net = tf.pad(tensor=net, paddings=spatial_pad_3, mode='REFLECT') net = tf.pad(tensor=net, paddings=spatial_pad_3, mode='REFLECT')
logits = layers.conv2d( logits = slim.conv2d(
net, net,
num_outputs, [7, 7], num_outputs, [7, 7],
activation_fn=None, activation_fn=None,
......
research/slim/nets/cyclegan_test.py
View file @ 451906e4
...@@ -18,7 +18,7 @@ from __future__ import absolute_import ...@@ -18,7 +18,7 @@ from __future__ import absolute_import
from __future__ import division from __future__ import division
from __future__ import print_function from __future__ import print_function
import tensorflow as tf import tensorflow.compat.v1 as tf
from nets import cyclegan from nets import cyclegan
...@@ -31,7 +31,7 @@ class CycleganTest(tf.test.TestCase): ...@@ -31,7 +31,7 @@ class CycleganTest(tf.test.TestCase):
img_batch = tf.zeros([2, 32, 32, 3]) img_batch = tf.zeros([2, 32, 32, 3])
model_output, _ = cyclegan.cyclegan_generator_resnet(img_batch) model_output, _ = cyclegan.cyclegan_generator_resnet(img_batch)
with self.test_session() as sess: with self.test_session() as sess:
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
sess.run(model_output) sess.run(model_output)
def _test_generator_graph_helper(self, shape): def _test_generator_graph_helper(self, shape):
...@@ -50,13 +50,13 @@ class CycleganTest(tf.test.TestCase): ...@@ -50,13 +50,13 @@ class CycleganTest(tf.test.TestCase):
def test_generator_unknown_batch_dim(self): def test_generator_unknown_batch_dim(self):
"""Check that generator can take unknown batch dimension inputs.""" """Check that generator can take unknown batch dimension inputs."""
img = tf.compat.v1.placeholder(tf.float32, shape=[None, 32, None, 3]) img = tf.placeholder(tf.float32, shape=[None, 32, None, 3])
output_imgs, _ = cyclegan.cyclegan_generator_resnet(img) output_imgs, _ = cyclegan.cyclegan_generator_resnet(img)
self.assertAllEqual([None, 32, None, 3], output_imgs.shape.as_list()) self.assertAllEqual([None, 32, None, 3], output_imgs.shape.as_list())
def _input_and_output_same_shape_helper(self, kernel_size): def _input_and_output_same_shape_helper(self, kernel_size):
img_batch = tf.compat.v1.placeholder(tf.float32, shape=[None, 32, 32, 3]) img_batch = tf.placeholder(tf.float32, shape=[None, 32, 32, 3])
output_img_batch, _ = cyclegan.cyclegan_generator_resnet( output_img_batch, _ = cyclegan.cyclegan_generator_resnet(
img_batch, kernel_size=kernel_size) img_batch, kernel_size=kernel_size)
...@@ -79,7 +79,7 @@ class CycleganTest(tf.test.TestCase): ...@@ -79,7 +79,7 @@ class CycleganTest(tf.test.TestCase):
self.assertRaisesRegexp( self.assertRaisesRegexp(
ValueError, 'The input height must be a multiple of 4.', ValueError, 'The input height must be a multiple of 4.',
cyclegan.cyclegan_generator_resnet, cyclegan.cyclegan_generator_resnet,
tf.compat.v1.placeholder(tf.float32, shape=[None, height, 32, 3])) tf.placeholder(tf.float32, shape=[None, height, 32, 3]))
def test_error_if_height_not_multiple_of_four_height29(self): def test_error_if_height_not_multiple_of_four_height29(self):
self._error_if_height_not_multiple_of_four_helper(29) self._error_if_height_not_multiple_of_four_helper(29)
...@@ -94,7 +94,7 @@ class CycleganTest(tf.test.TestCase): ...@@ -94,7 +94,7 @@ class CycleganTest(tf.test.TestCase):
self.assertRaisesRegexp( self.assertRaisesRegexp(
ValueError, 'The input width must be a multiple of 4.', ValueError, 'The input width must be a multiple of 4.',
cyclegan.cyclegan_generator_resnet, cyclegan.cyclegan_generator_resnet,
tf.compat.v1.placeholder(tf.float32, shape=[None, 32, width, 3])) tf.placeholder(tf.float32, shape=[None, 32, width, 3]))
def test_error_if_width_not_multiple_of_four_width29(self): def test_error_if_width_not_multiple_of_four_width29(self):
self._error_if_width_not_multiple_of_four_helper(29) self._error_if_width_not_multiple_of_four_helper(29)
......
research/slim/nets/dcgan.py
View file @ 451906e4
...@@ -20,10 +20,8 @@ from __future__ import print_function ...@@ -20,10 +20,8 @@ from __future__ import print_function
from math import log from math import log
from six.moves import xrange # pylint: disable=redefined-builtin from six.moves import xrange # pylint: disable=redefined-builtin
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import slim as contrib_slim import tf_slim as slim
slim = contrib_slim
def _validate_image_inputs(inputs): def _validate_image_inputs(inputs):
...@@ -82,7 +80,7 @@ def discriminator(inputs, ...@@ -82,7 +80,7 @@ def discriminator(inputs,
inp_shape = inputs.get_shape().as_list()[1] inp_shape = inputs.get_shape().as_list()[1]
end_points = {} end_points = {}
with tf.compat.v1.variable_scope( with tf.variable_scope(
scope, values=[inputs], reuse=reuse) as scope: scope, values=[inputs], reuse=reuse) as scope:
with slim.arg_scope([normalizer_fn], **normalizer_fn_args): with slim.arg_scope([normalizer_fn], **normalizer_fn_args):
with slim.arg_scope([slim.conv2d], with slim.arg_scope([slim.conv2d],
...@@ -157,7 +155,7 @@ def generator(inputs, ...@@ -157,7 +155,7 @@ def generator(inputs,
end_points = {} end_points = {}
num_layers = int(log(final_size, 2)) - 1 num_layers = int(log(final_size, 2)) - 1
with tf.compat.v1.variable_scope( with tf.variable_scope(
scope, values=[inputs], reuse=reuse) as scope: scope, values=[inputs], reuse=reuse) as scope:
with slim.arg_scope([normalizer_fn], **normalizer_fn_args): with slim.arg_scope([normalizer_fn], **normalizer_fn_args):
with slim.arg_scope([slim.conv2d_transpose], with slim.arg_scope([slim.conv2d_transpose],
......
research/slim/nets/dcgan_test.py
View file @ 451906e4
...@@ -19,7 +19,7 @@ from __future__ import division ...@@ -19,7 +19,7 @@ from __future__ import division
from __future__ import print_function from __future__ import print_function
from six.moves import xrange # pylint: disable=redefined-builtin from six.moves import xrange # pylint: disable=redefined-builtin
import tensorflow as tf import tensorflow.compat.v1 as tf
from nets import dcgan from nets import dcgan
...@@ -27,19 +27,19 @@ from nets import dcgan ...@@ -27,19 +27,19 @@ from nets import dcgan
class DCGANTest(tf.test.TestCase): class DCGANTest(tf.test.TestCase):
def test_generator_run(self): def test_generator_run(self):
tf.compat.v1.set_random_seed(1234) tf.set_random_seed(1234)
noise = tf.random.normal([100, 64]) noise = tf.random.normal([100, 64])
image, _ = dcgan.generator(noise) image, _ = dcgan.generator(noise)
with self.test_session() as sess: with self.test_session() as sess:
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
image.eval() image.eval()
def test_generator_graph(self): def test_generator_graph(self):
tf.compat.v1.set_random_seed(1234) tf.set_random_seed(1234)
# Check graph construction for a number of image size/depths and batch # Check graph construction for a number of image size/depths and batch
# sizes. # sizes.
for i, batch_size in zip(xrange(3, 7), xrange(3, 8)): for i, batch_size in zip(xrange(3, 7), xrange(3, 8)):
tf.compat.v1.reset_default_graph() tf.reset_default_graph()
final_size = 2 ** i final_size = 2 ** i
noise = tf.random.normal([batch_size, 64]) noise = tf.random.normal([batch_size, 64])
image, end_points = dcgan.generator( image, end_points = dcgan.generator(
...@@ -74,14 +74,14 @@ class DCGANTest(tf.test.TestCase): ...@@ -74,14 +74,14 @@ class DCGANTest(tf.test.TestCase):
image = tf.random.uniform([5, 32, 32, 3], -1, 1) image = tf.random.uniform([5, 32, 32, 3], -1, 1)
output, _ = dcgan.discriminator(image) output, _ = dcgan.discriminator(image)
with self.test_session() as sess: with self.test_session() as sess:
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
output.eval() output.eval()
def test_discriminator_graph(self): def test_discriminator_graph(self):
# Check graph construction for a number of image size/depths and batch # Check graph construction for a number of image size/depths and batch
# sizes. # sizes.
for i, batch_size in zip(xrange(1, 6), xrange(3, 8)): for i, batch_size in zip(xrange(1, 6), xrange(3, 8)):
tf.compat.v1.reset_default_graph() tf.reset_default_graph()
img_w = 2 ** i img_w = 2 ** i
image = tf.random.uniform([batch_size, img_w, img_w, 3], -1, 1) image = tf.random.uniform([batch_size, img_w, img_w, 3], -1, 1)
output, end_points = dcgan.discriminator( output, end_points = dcgan.discriminator(
...@@ -103,7 +103,7 @@ class DCGANTest(tf.test.TestCase): ...@@ -103,7 +103,7 @@ class DCGANTest(tf.test.TestCase):
with self.assertRaises(ValueError): with self.assertRaises(ValueError):
dcgan.discriminator(wrong_dim_img) dcgan.discriminator(wrong_dim_img)
spatially_undefined_shape = tf.compat.v1.placeholder( spatially_undefined_shape = tf.placeholder(
tf.float32, [5, 32, None, 3]) tf.float32, [5, 32, None, 3])
with self.assertRaises(ValueError): with self.assertRaises(ValueError):
dcgan.discriminator(spatially_undefined_shape) dcgan.discriminator(spatially_undefined_shape)
......
research/slim/nets/i3d.py
View file @ 451906e4
...@@ -24,16 +24,14 @@ from __future__ import absolute_import ...@@ -24,16 +24,14 @@ from __future__ import absolute_import
from __future__ import division from __future__ import division
from __future__ import print_function from __future__ import print_function
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import slim as contrib_slim import tf_slim as slim
from nets import i3d_utils from nets import i3d_utils
from nets import s3dg from nets import s3dg
slim = contrib_slim
# pylint: disable=g-long-lambda # pylint: disable=g-long-lambda
trunc_normal = lambda stddev: tf.compat.v1.truncated_normal_initializer( trunc_normal = lambda stddev: tf.truncated_normal_initializer(
0.0, stddev) 0.0, stddev)
conv3d_spatiotemporal = i3d_utils.conv3d_spatiotemporal conv3d_spatiotemporal = i3d_utils.conv3d_spatiotemporal
...@@ -152,12 +150,12 @@ def i3d(inputs, ...@@ -152,12 +150,12 @@ def i3d(inputs,
activation. activation.
""" """
# Final pooling and prediction # Final pooling and prediction
with tf.compat.v1.variable_scope( with tf.variable_scope(
scope, 'InceptionV1', [inputs, num_classes], reuse=reuse) as scope: scope, 'InceptionV1', [inputs, num_classes], reuse=reuse) as scope:
with slim.arg_scope( with slim.arg_scope(
[slim.batch_norm, slim.dropout], is_training=is_training): [slim.batch_norm, slim.dropout], is_training=is_training):
net, end_points = i3d_base(inputs, scope=scope) net, end_points = i3d_base(inputs, scope=scope)
with tf.compat.v1.variable_scope('Logits'): with tf.variable_scope('Logits'):
kernel_size = i3d_utils.reduced_kernel_size_3d(net, [2, 7, 7]) kernel_size = i3d_utils.reduced_kernel_size_3d(net, [2, 7, 7])
net = slim.avg_pool3d( net = slim.avg_pool3d(
net, kernel_size, stride=1, scope='AvgPool_0a_7x7') net, kernel_size, stride=1, scope='AvgPool_0a_7x7')
......
research/slim/nets/i3d_test.py
View file @ 451906e4
...@@ -18,7 +18,8 @@ from __future__ import absolute_import ...@@ -18,7 +18,8 @@ from __future__ import absolute_import
from __future__ import division from __future__ import division
from __future__ import print_function from __future__ import print_function
import tensorflow as tf import six
import tensorflow.compat.v1 as tf
from nets import i3d from nets import i3d
...@@ -55,7 +56,7 @@ class I3DTest(tf.test.TestCase): ...@@ -55,7 +56,7 @@ class I3DTest(tf.test.TestCase):
'Mixed_3c', 'MaxPool_4a_3x3', 'Mixed_4b', 'Mixed_4c', 'Mixed_3c', 'MaxPool_4a_3x3', 'Mixed_4b', 'Mixed_4c',
'Mixed_4d', 'Mixed_4e', 'Mixed_4f', 'MaxPool_5a_2x2', 'Mixed_4d', 'Mixed_4e', 'Mixed_4f', 'MaxPool_5a_2x2',
'Mixed_5b', 'Mixed_5c'] 'Mixed_5b', 'Mixed_5c']
self.assertItemsEqual(end_points.keys(), expected_endpoints) self.assertItemsEqual(list(end_points.keys()), expected_endpoints)
def testBuildOnlyUptoFinalEndpoint(self): def testBuildOnlyUptoFinalEndpoint(self):
batch_size = 5 batch_size = 5
...@@ -100,8 +101,9 @@ class I3DTest(tf.test.TestCase): ...@@ -100,8 +101,9 @@ class I3DTest(tf.test.TestCase):
'Mixed_5b': [5, 8, 7, 7, 832], 'Mixed_5b': [5, 8, 7, 7, 832],
'Mixed_5c': [5, 8, 7, 7, 1024]} 'Mixed_5c': [5, 8, 7, 7, 1024]}
self.assertItemsEqual(endpoints_shapes.keys(), end_points.keys()) self.assertItemsEqual(
for endpoint_name, expected_shape in endpoints_shapes.iteritems(): list(endpoints_shapes.keys()), list(end_points.keys()))
for endpoint_name, expected_shape in six.iteritems(endpoints_shapes):
self.assertTrue(endpoint_name in end_points) self.assertTrue(endpoint_name in end_points)
self.assertListEqual(end_points[endpoint_name].get_shape().as_list(), self.assertListEqual(end_points[endpoint_name].get_shape().as_list(),
expected_shape) expected_shape)
...@@ -140,7 +142,7 @@ class I3DTest(tf.test.TestCase): ...@@ -140,7 +142,7 @@ class I3DTest(tf.test.TestCase):
predictions = tf.argmax(input=logits, axis=1) predictions = tf.argmax(input=logits, axis=1)
with self.test_session() as sess: with self.test_session() as sess:
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
output = sess.run(predictions) output = sess.run(predictions)
self.assertEquals(output.shape, (batch_size,)) self.assertEquals(output.shape, (batch_size,))
......
research/slim/nets/i3d_utils.py
View file @ 451906e4
...@@ -19,15 +19,11 @@ from __future__ import division ...@@ -19,15 +19,11 @@ from __future__ import division
from __future__ import print_function from __future__ import print_function
import numpy as np import numpy as np
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import framework as contrib_framework import tf_slim as slim
from tensorflow.contrib import layers as contrib_layers
add_arg_scope = slim.add_arg_scope
# Orignaly, add_arg_scope = slim.add_arg_scope and layers = slim, now switch to layers = slim.layers
# more update-to-date tf.contrib.* API.
add_arg_scope = contrib_framework.add_arg_scope
layers = contrib_layers
def center_initializer(): def center_initializer():
...@@ -228,13 +224,13 @@ def inception_block_v1_3d(inputs, ...@@ -228,13 +224,13 @@ def inception_block_v1_3d(inputs,
""" """
use_gating = self_gating_fn is not None use_gating = self_gating_fn is not None
with tf.compat.v1.variable_scope(scope): with tf.variable_scope(scope):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = layers.conv3d( branch_0 = layers.conv3d(
inputs, num_outputs_0_0a, [1, 1, 1], scope='Conv2d_0a_1x1') inputs, num_outputs_0_0a, [1, 1, 1], scope='Conv2d_0a_1x1')
if use_gating: if use_gating:
branch_0 = self_gating_fn(branch_0, scope='Conv2d_0a_1x1') branch_0 = self_gating_fn(branch_0, scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = layers.conv3d( branch_1 = layers.conv3d(
inputs, num_outputs_1_0a, [1, 1, 1], scope='Conv2d_0a_1x1') inputs, num_outputs_1_0a, [1, 1, 1], scope='Conv2d_0a_1x1')
branch_1 = conv3d_spatiotemporal( branch_1 = conv3d_spatiotemporal(
...@@ -242,7 +238,7 @@ def inception_block_v1_3d(inputs, ...@@ -242,7 +238,7 @@ def inception_block_v1_3d(inputs,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
if use_gating: if use_gating:
branch_1 = self_gating_fn(branch_1, scope='Conv2d_0b_3x3') branch_1 = self_gating_fn(branch_1, scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = layers.conv3d( branch_2 = layers.conv3d(
inputs, num_outputs_2_0a, [1, 1, 1], scope='Conv2d_0a_1x1') inputs, num_outputs_2_0a, [1, 1, 1], scope='Conv2d_0a_1x1')
branch_2 = conv3d_spatiotemporal( branch_2 = conv3d_spatiotemporal(
...@@ -250,7 +246,7 @@ def inception_block_v1_3d(inputs, ...@@ -250,7 +246,7 @@ def inception_block_v1_3d(inputs,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
if use_gating: if use_gating:
branch_2 = self_gating_fn(branch_2, scope='Conv2d_0b_3x3') branch_2 = self_gating_fn(branch_2, scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = layers.max_pool3d(inputs, [3, 3, 3], scope='MaxPool_0a_3x3') branch_3 = layers.max_pool3d(inputs, [3, 3, 3], scope='MaxPool_0a_3x3')
branch_3 = layers.conv3d( branch_3 = layers.conv3d(
branch_3, num_outputs_3_0b, [1, 1, 1], scope='Conv2d_0b_1x1') branch_3, num_outputs_3_0b, [1, 1, 1], scope='Conv2d_0b_1x1')
......
research/slim/nets/inception_resnet_v2.py
View file @ 451906e4
...@@ -25,21 +25,19 @@ from __future__ import division ...@@ -25,21 +25,19 @@ from __future__ import division
from __future__ import print_function from __future__ import print_function
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import slim as contrib_slim import tf_slim as slim
slim = contrib_slim
def block35(net, scale=1.0, activation_fn=tf.nn.relu, scope=None, reuse=None): def block35(net, scale=1.0, activation_fn=tf.nn.relu, scope=None, reuse=None):
"""Builds the 35x35 resnet block.""" """Builds the 35x35 resnet block."""
with tf.compat.v1.variable_scope(scope, 'Block35', [net], reuse=reuse): with tf.variable_scope(scope, 'Block35', [net], reuse=reuse):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
tower_conv = slim.conv2d(net, 32, 1, scope='Conv2d_1x1') tower_conv = slim.conv2d(net, 32, 1, scope='Conv2d_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
tower_conv1_0 = slim.conv2d(net, 32, 1, scope='Conv2d_0a_1x1') tower_conv1_0 = slim.conv2d(net, 32, 1, scope='Conv2d_0a_1x1')
tower_conv1_1 = slim.conv2d(tower_conv1_0, 32, 3, scope='Conv2d_0b_3x3') tower_conv1_1 = slim.conv2d(tower_conv1_0, 32, 3, scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
tower_conv2_0 = slim.conv2d(net, 32, 1, scope='Conv2d_0a_1x1') tower_conv2_0 = slim.conv2d(net, 32, 1, scope='Conv2d_0a_1x1')
tower_conv2_1 = slim.conv2d(tower_conv2_0, 48, 3, scope='Conv2d_0b_3x3') tower_conv2_1 = slim.conv2d(tower_conv2_0, 48, 3, scope='Conv2d_0b_3x3')
tower_conv2_2 = slim.conv2d(tower_conv2_1, 64, 3, scope='Conv2d_0c_3x3') tower_conv2_2 = slim.conv2d(tower_conv2_1, 64, 3, scope='Conv2d_0c_3x3')
...@@ -59,10 +57,10 @@ def block35(net, scale=1.0, activation_fn=tf.nn.relu, scope=None, reuse=None): ...@@ -59,10 +57,10 @@ def block35(net, scale=1.0, activation_fn=tf.nn.relu, scope=None, reuse=None):
def block17(net, scale=1.0, activation_fn=tf.nn.relu, scope=None, reuse=None): def block17(net, scale=1.0, activation_fn=tf.nn.relu, scope=None, reuse=None):
"""Builds the 17x17 resnet block.""" """Builds the 17x17 resnet block."""
with tf.compat.v1.variable_scope(scope, 'Block17', [net], reuse=reuse): with tf.variable_scope(scope, 'Block17', [net], reuse=reuse):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
tower_conv = slim.conv2d(net, 192, 1, scope='Conv2d_1x1') tower_conv = slim.conv2d(net, 192, 1, scope='Conv2d_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
tower_conv1_0 = slim.conv2d(net, 128, 1, scope='Conv2d_0a_1x1') tower_conv1_0 = slim.conv2d(net, 128, 1, scope='Conv2d_0a_1x1')
tower_conv1_1 = slim.conv2d(tower_conv1_0, 160, [1, 7], tower_conv1_1 = slim.conv2d(tower_conv1_0, 160, [1, 7],
scope='Conv2d_0b_1x7') scope='Conv2d_0b_1x7')
...@@ -85,10 +83,10 @@ def block17(net, scale=1.0, activation_fn=tf.nn.relu, scope=None, reuse=None): ...@@ -85,10 +83,10 @@ def block17(net, scale=1.0, activation_fn=tf.nn.relu, scope=None, reuse=None):
def block8(net, scale=1.0, activation_fn=tf.nn.relu, scope=None, reuse=None): def block8(net, scale=1.0, activation_fn=tf.nn.relu, scope=None, reuse=None):
"""Builds the 8x8 resnet block.""" """Builds the 8x8 resnet block."""
with tf.compat.v1.variable_scope(scope, 'Block8', [net], reuse=reuse): with tf.variable_scope(scope, 'Block8', [net], reuse=reuse):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
tower_conv = slim.conv2d(net, 192, 1, scope='Conv2d_1x1') tower_conv = slim.conv2d(net, 192, 1, scope='Conv2d_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
tower_conv1_0 = slim.conv2d(net, 192, 1, scope='Conv2d_0a_1x1') tower_conv1_0 = slim.conv2d(net, 192, 1, scope='Conv2d_0a_1x1')
tower_conv1_1 = slim.conv2d(tower_conv1_0, 224, [1, 3], tower_conv1_1 = slim.conv2d(tower_conv1_0, 224, [1, 3],
scope='Conv2d_0b_1x3') scope='Conv2d_0b_1x3')
...@@ -155,7 +153,7 @@ def inception_resnet_v2_base(inputs, ...@@ -155,7 +153,7 @@ def inception_resnet_v2_base(inputs,
end_points[name] = net end_points[name] = net
return name == final_endpoint return name == final_endpoint
with tf.compat.v1.variable_scope(scope, 'InceptionResnetV2', [inputs]): with tf.variable_scope(scope, 'InceptionResnetV2', [inputs]):
with slim.arg_scope([slim.conv2d, slim.max_pool2d, slim.avg_pool2d], with slim.arg_scope([slim.conv2d, slim.max_pool2d, slim.avg_pool2d],
stride=1, padding='SAME'): stride=1, padding='SAME'):
# 149 x 149 x 32 # 149 x 149 x 32
...@@ -188,20 +186,20 @@ def inception_resnet_v2_base(inputs, ...@@ -188,20 +186,20 @@ def inception_resnet_v2_base(inputs,
if add_and_check_final('MaxPool_5a_3x3', net): return net, end_points if add_and_check_final('MaxPool_5a_3x3', net): return net, end_points
# 35 x 35 x 320 # 35 x 35 x 320
with tf.compat.v1.variable_scope('Mixed_5b'): with tf.variable_scope('Mixed_5b'):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
tower_conv = slim.conv2d(net, 96, 1, scope='Conv2d_1x1') tower_conv = slim.conv2d(net, 96, 1, scope='Conv2d_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
tower_conv1_0 = slim.conv2d(net, 48, 1, scope='Conv2d_0a_1x1') tower_conv1_0 = slim.conv2d(net, 48, 1, scope='Conv2d_0a_1x1')
tower_conv1_1 = slim.conv2d(tower_conv1_0, 64, 5, tower_conv1_1 = slim.conv2d(tower_conv1_0, 64, 5,
scope='Conv2d_0b_5x5') scope='Conv2d_0b_5x5')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
tower_conv2_0 = slim.conv2d(net, 64, 1, scope='Conv2d_0a_1x1') tower_conv2_0 = slim.conv2d(net, 64, 1, scope='Conv2d_0a_1x1')
tower_conv2_1 = slim.conv2d(tower_conv2_0, 96, 3, tower_conv2_1 = slim.conv2d(tower_conv2_0, 96, 3,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
tower_conv2_2 = slim.conv2d(tower_conv2_1, 96, 3, tower_conv2_2 = slim.conv2d(tower_conv2_1, 96, 3,
scope='Conv2d_0c_3x3') scope='Conv2d_0c_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
tower_pool = slim.avg_pool2d(net, 3, stride=1, padding='SAME', tower_pool = slim.avg_pool2d(net, 3, stride=1, padding='SAME',
scope='AvgPool_0a_3x3') scope='AvgPool_0a_3x3')
tower_pool_1 = slim.conv2d(tower_pool, 64, 1, tower_pool_1 = slim.conv2d(tower_pool, 64, 1,
...@@ -218,12 +216,12 @@ def inception_resnet_v2_base(inputs, ...@@ -218,12 +216,12 @@ def inception_resnet_v2_base(inputs,
# 33 x 33 x 1088 if output_stride == 16 # 33 x 33 x 1088 if output_stride == 16
use_atrous = output_stride == 8 use_atrous = output_stride == 8
with tf.compat.v1.variable_scope('Mixed_6a'): with tf.variable_scope('Mixed_6a'):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
tower_conv = slim.conv2d(net, 384, 3, stride=1 if use_atrous else 2, tower_conv = slim.conv2d(net, 384, 3, stride=1 if use_atrous else 2,
padding=padding, padding=padding,
scope='Conv2d_1a_3x3') scope='Conv2d_1a_3x3')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
tower_conv1_0 = slim.conv2d(net, 256, 1, scope='Conv2d_0a_1x1') tower_conv1_0 = slim.conv2d(net, 256, 1, scope='Conv2d_0a_1x1')
tower_conv1_1 = slim.conv2d(tower_conv1_0, 256, 3, tower_conv1_1 = slim.conv2d(tower_conv1_0, 256, 3,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
...@@ -231,7 +229,7 @@ def inception_resnet_v2_base(inputs, ...@@ -231,7 +229,7 @@ def inception_resnet_v2_base(inputs,
stride=1 if use_atrous else 2, stride=1 if use_atrous else 2,
padding=padding, padding=padding,
scope='Conv2d_1a_3x3') scope='Conv2d_1a_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
tower_pool = slim.max_pool2d(net, 3, stride=1 if use_atrous else 2, tower_pool = slim.max_pool2d(net, 3, stride=1 if use_atrous else 2,
padding=padding, padding=padding,
scope='MaxPool_1a_3x3') scope='MaxPool_1a_3x3')
...@@ -251,25 +249,25 @@ def inception_resnet_v2_base(inputs, ...@@ -251,25 +249,25 @@ def inception_resnet_v2_base(inputs,
'PreAuxlogits end_point for now.') 'PreAuxlogits end_point for now.')
# 8 x 8 x 2080 # 8 x 8 x 2080
with tf.compat.v1.variable_scope('Mixed_7a'): with tf.variable_scope('Mixed_7a'):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
tower_conv = slim.conv2d(net, 256, 1, scope='Conv2d_0a_1x1') tower_conv = slim.conv2d(net, 256, 1, scope='Conv2d_0a_1x1')
tower_conv_1 = slim.conv2d(tower_conv, 384, 3, stride=2, tower_conv_1 = slim.conv2d(tower_conv, 384, 3, stride=2,
padding=padding, padding=padding,
scope='Conv2d_1a_3x3') scope='Conv2d_1a_3x3')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
tower_conv1 = slim.conv2d(net, 256, 1, scope='Conv2d_0a_1x1') tower_conv1 = slim.conv2d(net, 256, 1, scope='Conv2d_0a_1x1')
tower_conv1_1 = slim.conv2d(tower_conv1, 288, 3, stride=2, tower_conv1_1 = slim.conv2d(tower_conv1, 288, 3, stride=2,
padding=padding, padding=padding,
scope='Conv2d_1a_3x3') scope='Conv2d_1a_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
tower_conv2 = slim.conv2d(net, 256, 1, scope='Conv2d_0a_1x1') tower_conv2 = slim.conv2d(net, 256, 1, scope='Conv2d_0a_1x1')
tower_conv2_1 = slim.conv2d(tower_conv2, 288, 3, tower_conv2_1 = slim.conv2d(tower_conv2, 288, 3,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
tower_conv2_2 = slim.conv2d(tower_conv2_1, 320, 3, stride=2, tower_conv2_2 = slim.conv2d(tower_conv2_1, 320, 3, stride=2,
padding=padding, padding=padding,
scope='Conv2d_1a_3x3') scope='Conv2d_1a_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
tower_pool = slim.max_pool2d(net, 3, stride=2, tower_pool = slim.max_pool2d(net, 3, stride=2,
padding=padding, padding=padding,
scope='MaxPool_1a_3x3') scope='MaxPool_1a_3x3')
...@@ -320,7 +318,7 @@ def inception_resnet_v2(inputs, num_classes=1001, is_training=True, ...@@ -320,7 +318,7 @@ def inception_resnet_v2(inputs, num_classes=1001, is_training=True,
""" """
end_points = {} end_points = {}
with tf.compat.v1.variable_scope( with tf.variable_scope(
scope, 'InceptionResnetV2', [inputs], reuse=reuse) as scope: scope, 'InceptionResnetV2', [inputs], reuse=reuse) as scope:
with slim.arg_scope([slim.batch_norm, slim.dropout], with slim.arg_scope([slim.batch_norm, slim.dropout],
is_training=is_training): is_training=is_training):
...@@ -329,7 +327,7 @@ def inception_resnet_v2(inputs, num_classes=1001, is_training=True, ...@@ -329,7 +327,7 @@ def inception_resnet_v2(inputs, num_classes=1001, is_training=True,
activation_fn=activation_fn) activation_fn=activation_fn)
if create_aux_logits and num_classes: if create_aux_logits and num_classes:
with tf.compat.v1.variable_scope('AuxLogits'): with tf.variable_scope('AuxLogits'):
aux = end_points['PreAuxLogits'] aux = end_points['PreAuxLogits']
aux = slim.avg_pool2d(aux, 5, stride=3, padding='VALID', aux = slim.avg_pool2d(aux, 5, stride=3, padding='VALID',
scope='Conv2d_1a_3x3') scope='Conv2d_1a_3x3')
...@@ -341,7 +339,7 @@ def inception_resnet_v2(inputs, num_classes=1001, is_training=True, ...@@ -341,7 +339,7 @@ def inception_resnet_v2(inputs, num_classes=1001, is_training=True,
scope='Logits') scope='Logits')
end_points['AuxLogits'] = aux end_points['AuxLogits'] = aux
with tf.compat.v1.variable_scope('Logits'): with tf.variable_scope('Logits'):
# TODO(sguada,arnoegw): Consider adding a parameter global_pool which # TODO(sguada,arnoegw): Consider adding a parameter global_pool which
# can be set to False to disable pooling here (as in resnet_*()). # can be set to False to disable pooling here (as in resnet_*()).
kernel_size = net.get_shape()[1:3] kernel_size = net.get_shape()[1:3]
...@@ -372,7 +370,7 @@ def inception_resnet_v2_arg_scope( ...@@ -372,7 +370,7 @@ def inception_resnet_v2_arg_scope(
batch_norm_decay=0.9997, batch_norm_decay=0.9997,
batch_norm_epsilon=0.001, batch_norm_epsilon=0.001,
activation_fn=tf.nn.relu, activation_fn=tf.nn.relu,
batch_norm_updates_collections=tf.compat.v1.GraphKeys.UPDATE_OPS, batch_norm_updates_collections=tf.GraphKeys.UPDATE_OPS,
batch_norm_scale=False): batch_norm_scale=False):
"""Returns the scope with the default parameters for inception_resnet_v2. """Returns the scope with the default parameters for inception_resnet_v2.
......
research/slim/nets/inception_resnet_v2_test.py
View file @ 451906e4
...@@ -17,8 +17,8 @@ from __future__ import absolute_import ...@@ -17,8 +17,8 @@ from __future__ import absolute_import
from __future__ import division from __future__ import division
from __future__ import print_function from __future__ import print_function
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import slim as contrib_slim import tf_slim as slim
from nets import inception from nets import inception
...@@ -117,7 +117,7 @@ class InceptionTest(tf.test.TestCase): ...@@ -117,7 +117,7 @@ class InceptionTest(tf.test.TestCase):
if endpoint != 'PreAuxLogits': if endpoint != 'PreAuxLogits':
self.assertTrue(out_tensor.op.name.startswith( self.assertTrue(out_tensor.op.name.startswith(
'InceptionResnetV2/' + endpoint)) 'InceptionResnetV2/' + endpoint))
self.assertItemsEqual(endpoints[:index+1], end_points.keys()) self.assertItemsEqual(endpoints[:index + 1], end_points.keys())
def testBuildAndCheckAllEndPointsUptoPreAuxLogits(self): def testBuildAndCheckAllEndPointsUptoPreAuxLogits(self):
batch_size = 5 batch_size = 5
...@@ -204,15 +204,15 @@ class InceptionTest(tf.test.TestCase): ...@@ -204,15 +204,15 @@ class InceptionTest(tf.test.TestCase):
with self.test_session(): with self.test_session():
inputs = tf.random.uniform((batch_size, height, width, 3)) inputs = tf.random.uniform((batch_size, height, width, 3))
# Force all Variables to reside on the device. # Force all Variables to reside on the device.
with tf.compat.v1.variable_scope('on_cpu'), tf.device('/cpu:0'): with tf.variable_scope('on_cpu'), tf.device('/cpu:0'):
inception.inception_resnet_v2(inputs, num_classes) inception.inception_resnet_v2(inputs, num_classes)
with tf.compat.v1.variable_scope('on_gpu'), tf.device('/gpu:0'): with tf.variable_scope('on_gpu'), tf.device('/gpu:0'):
inception.inception_resnet_v2(inputs, num_classes) inception.inception_resnet_v2(inputs, num_classes)
for v in tf.compat.v1.get_collection( for v in tf.get_collection(
tf.compat.v1.GraphKeys.GLOBAL_VARIABLES, scope='on_cpu'): tf.GraphKeys.GLOBAL_VARIABLES, scope='on_cpu'):
self.assertDeviceEqual(v.device, '/cpu:0') self.assertDeviceEqual(v.device, '/cpu:0')
for v in tf.compat.v1.get_collection( for v in tf.get_collection(
tf.compat.v1.GraphKeys.GLOBAL_VARIABLES, scope='on_gpu'): tf.GraphKeys.GLOBAL_VARIABLES, scope='on_gpu'):
self.assertDeviceEqual(v.device, '/gpu:0') self.assertDeviceEqual(v.device, '/gpu:0')
def testHalfSizeImages(self): def testHalfSizeImages(self):
...@@ -248,7 +248,7 @@ class InceptionTest(tf.test.TestCase): ...@@ -248,7 +248,7 @@ class InceptionTest(tf.test.TestCase):
height, width = 330, 400 height, width = 330, 400
num_classes = 1000 num_classes = 1000
with self.test_session() as sess: with self.test_session() as sess:
inputs = tf.compat.v1.placeholder(tf.float32, (batch_size, None, None, 3)) inputs = tf.placeholder(tf.float32, (batch_size, None, None, 3))
logits, end_points = inception.inception_resnet_v2( logits, end_points = inception.inception_resnet_v2(
inputs, num_classes, create_aux_logits=False) inputs, num_classes, create_aux_logits=False)
self.assertTrue(logits.op.name.startswith('InceptionResnetV2/Logits')) self.assertTrue(logits.op.name.startswith('InceptionResnetV2/Logits'))
...@@ -256,7 +256,7 @@ class InceptionTest(tf.test.TestCase): ...@@ -256,7 +256,7 @@ class InceptionTest(tf.test.TestCase):
[batch_size, num_classes]) [batch_size, num_classes])
pre_pool = end_points['Conv2d_7b_1x1'] pre_pool = end_points['Conv2d_7b_1x1']
images = tf.random.uniform((batch_size, height, width, 3)) images = tf.random.uniform((batch_size, height, width, 3))
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
logits_out, pre_pool_out = sess.run([logits, pre_pool], logits_out, pre_pool_out = sess.run([logits, pre_pool],
{inputs: images.eval()}) {inputs: images.eval()})
self.assertTupleEqual(logits_out.shape, (batch_size, num_classes)) self.assertTupleEqual(logits_out.shape, (batch_size, num_classes))
...@@ -267,13 +267,13 @@ class InceptionTest(tf.test.TestCase): ...@@ -267,13 +267,13 @@ class InceptionTest(tf.test.TestCase):
height, width = 299, 299 height, width = 299, 299
num_classes = 1000 num_classes = 1000
with self.test_session() as sess: with self.test_session() as sess:
inputs = tf.compat.v1.placeholder(tf.float32, (None, height, width, 3)) inputs = tf.placeholder(tf.float32, (None, height, width, 3))
logits, _ = inception.inception_resnet_v2(inputs, num_classes) logits, _ = inception.inception_resnet_v2(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionResnetV2/Logits')) self.assertTrue(logits.op.name.startswith('InceptionResnetV2/Logits'))
self.assertListEqual(logits.get_shape().as_list(), self.assertListEqual(logits.get_shape().as_list(),
[None, num_classes]) [None, num_classes])
images = tf.random.uniform((batch_size, height, width, 3)) images = tf.random.uniform((batch_size, height, width, 3))
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
output = sess.run(logits, {inputs: images.eval()}) output = sess.run(logits, {inputs: images.eval()})
self.assertEquals(output.shape, (batch_size, num_classes)) self.assertEquals(output.shape, (batch_size, num_classes))
...@@ -287,7 +287,7 @@ class InceptionTest(tf.test.TestCase): ...@@ -287,7 +287,7 @@ class InceptionTest(tf.test.TestCase):
num_classes, num_classes,
is_training=False) is_training=False)
predictions = tf.argmax(input=logits, axis=1) predictions = tf.argmax(input=logits, axis=1)
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
output = sess.run(predictions) output = sess.run(predictions)
self.assertEquals(output.shape, (batch_size,)) self.assertEquals(output.shape, (batch_size,))
...@@ -305,32 +305,32 @@ class InceptionTest(tf.test.TestCase): ...@@ -305,32 +305,32 @@ class InceptionTest(tf.test.TestCase):
is_training=False, is_training=False,
reuse=True) reuse=True)
predictions = tf.argmax(input=logits, axis=1) predictions = tf.argmax(input=logits, axis=1)
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
output = sess.run(predictions) output = sess.run(predictions)
self.assertEquals(output.shape, (eval_batch_size,)) self.assertEquals(output.shape, (eval_batch_size,))
def testNoBatchNormScaleByDefault(self): def testNoBatchNormScaleByDefault(self):
height, width = 299, 299 height, width = 299, 299
num_classes = 1000 num_classes = 1000
inputs = tf.compat.v1.placeholder(tf.float32, (1, height, width, 3)) inputs = tf.placeholder(tf.float32, (1, height, width, 3))
with contrib_slim.arg_scope(inception.inception_resnet_v2_arg_scope()): with slim.arg_scope(inception.inception_resnet_v2_arg_scope()):
inception.inception_resnet_v2(inputs, num_classes, is_training=False) inception.inception_resnet_v2(inputs, num_classes, is_training=False)
self.assertEqual(tf.compat.v1.global_variables('.*/BatchNorm/gamma:0$'), []) self.assertEqual(tf.global_variables('.*/BatchNorm/gamma:0$'), [])
def testBatchNormScale(self): def testBatchNormScale(self):
height, width = 299, 299 height, width = 299, 299
num_classes = 1000 num_classes = 1000
inputs = tf.compat.v1.placeholder(tf.float32, (1, height, width, 3)) inputs = tf.placeholder(tf.float32, (1, height, width, 3))
with contrib_slim.arg_scope( with slim.arg_scope(
inception.inception_resnet_v2_arg_scope(batch_norm_scale=True)): inception.inception_resnet_v2_arg_scope(batch_norm_scale=True)):
inception.inception_resnet_v2(inputs, num_classes, is_training=False) inception.inception_resnet_v2(inputs, num_classes, is_training=False)
gamma_names = set( gamma_names = set(
v.op.name v.op.name
for v in tf.compat.v1.global_variables('.*/BatchNorm/gamma:0$')) for v in tf.global_variables('.*/BatchNorm/gamma:0$'))
self.assertGreater(len(gamma_names), 0) self.assertGreater(len(gamma_names), 0)
for v in tf.compat.v1.global_variables('.*/BatchNorm/moving_mean:0$'): for v in tf.global_variables('.*/BatchNorm/moving_mean:0$'):
self.assertIn(v.op.name[:-len('moving_mean')] + 'gamma', gamma_names) self.assertIn(v.op.name[:-len('moving_mean')] + 'gamma', gamma_names)
......
research/slim/nets/inception_utils.py
View file @ 451906e4
...@@ -24,10 +24,8 @@ from __future__ import absolute_import ...@@ -24,10 +24,8 @@ from __future__ import absolute_import
from __future__ import division from __future__ import division
from __future__ import print_function from __future__ import print_function
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import slim as contrib_slim import tf_slim as slim
slim = contrib_slim
def inception_arg_scope( def inception_arg_scope(
...@@ -36,7 +34,7 @@ def inception_arg_scope( ...@@ -36,7 +34,7 @@ def inception_arg_scope(
batch_norm_decay=0.9997, batch_norm_decay=0.9997,
batch_norm_epsilon=0.001, batch_norm_epsilon=0.001,
activation_fn=tf.nn.relu, activation_fn=tf.nn.relu,
batch_norm_updates_collections=tf.compat.v1.GraphKeys.UPDATE_OPS, batch_norm_updates_collections=tf.GraphKeys.UPDATE_OPS,
batch_norm_scale=False): batch_norm_scale=False):
"""Defines the default arg scope for inception models. """Defines the default arg scope for inception models.
......
research/slim/nets/inception_v1.py
View file @ 451906e4
...@@ -18,15 +18,13 @@ from __future__ import absolute_import ...@@ -18,15 +18,13 @@ from __future__ import absolute_import
from __future__ import division from __future__ import division
from __future__ import print_function from __future__ import print_function
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import slim as contrib_slim import tf_slim as slim
from nets import inception_utils from nets import inception_utils
slim = contrib_slim
# pylint: disable=g-long-lambda # pylint: disable=g-long-lambda
trunc_normal = lambda stddev: tf.compat.v1.truncated_normal_initializer( trunc_normal = lambda stddev: tf.truncated_normal_initializer(
0.0, stddev) 0.0, stddev)
...@@ -62,7 +60,7 @@ def inception_v1_base(inputs, ...@@ -62,7 +60,7 @@ def inception_v1_base(inputs,
ValueError: if final_endpoint is not set to one of the predefined values. ValueError: if final_endpoint is not set to one of the predefined values.
""" """
end_points = {} end_points = {}
with tf.compat.v1.variable_scope(scope, 'InceptionV1', [inputs]): with tf.variable_scope(scope, 'InceptionV1', [inputs]):
with slim.arg_scope( with slim.arg_scope(
[slim.conv2d, slim.fully_connected], [slim.conv2d, slim.fully_connected],
weights_initializer=trunc_normal(0.01)): weights_initializer=trunc_normal(0.01)):
...@@ -97,16 +95,16 @@ def inception_v1_base(inputs, ...@@ -97,16 +95,16 @@ def inception_v1_base(inputs,
return net, end_points return net, end_points
end_point = 'Mixed_3b' end_point = 'Mixed_3b'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, 64, [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, 64, [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d(net, 96, [1, 1], scope='Conv2d_0a_1x1') branch_1 = slim.conv2d(net, 96, [1, 1], scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, 128, [3, 3], scope='Conv2d_0b_3x3') branch_1 = slim.conv2d(branch_1, 128, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d(net, 16, [1, 1], scope='Conv2d_0a_1x1') branch_2 = slim.conv2d(net, 16, [1, 1], scope='Conv2d_0a_1x1')
branch_2 = slim.conv2d(branch_2, 32, [3, 3], scope='Conv2d_0b_3x3') branch_2 = slim.conv2d(branch_2, 32, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3') branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3')
branch_3 = slim.conv2d(branch_3, 32, [1, 1], scope='Conv2d_0b_1x1') branch_3 = slim.conv2d(branch_3, 32, [1, 1], scope='Conv2d_0b_1x1')
net = tf.concat( net = tf.concat(
...@@ -115,16 +113,16 @@ def inception_v1_base(inputs, ...@@ -115,16 +113,16 @@ def inception_v1_base(inputs,
if final_endpoint == end_point: return net, end_points if final_endpoint == end_point: return net, end_points
end_point = 'Mixed_3c' end_point = 'Mixed_3c'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, 128, [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, 128, [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d(net, 128, [1, 1], scope='Conv2d_0a_1x1') branch_1 = slim.conv2d(net, 128, [1, 1], scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, 192, [3, 3], scope='Conv2d_0b_3x3') branch_1 = slim.conv2d(branch_1, 192, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d(net, 32, [1, 1], scope='Conv2d_0a_1x1') branch_2 = slim.conv2d(net, 32, [1, 1], scope='Conv2d_0a_1x1')
branch_2 = slim.conv2d(branch_2, 96, [3, 3], scope='Conv2d_0b_3x3') branch_2 = slim.conv2d(branch_2, 96, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3') branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3')
branch_3 = slim.conv2d(branch_3, 64, [1, 1], scope='Conv2d_0b_1x1') branch_3 = slim.conv2d(branch_3, 64, [1, 1], scope='Conv2d_0b_1x1')
net = tf.concat( net = tf.concat(
...@@ -138,16 +136,16 @@ def inception_v1_base(inputs, ...@@ -138,16 +136,16 @@ def inception_v1_base(inputs,
if final_endpoint == end_point: return net, end_points if final_endpoint == end_point: return net, end_points
end_point = 'Mixed_4b' end_point = 'Mixed_4b'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, 192, [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, 192, [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d(net, 96, [1, 1], scope='Conv2d_0a_1x1') branch_1 = slim.conv2d(net, 96, [1, 1], scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, 208, [3, 3], scope='Conv2d_0b_3x3') branch_1 = slim.conv2d(branch_1, 208, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d(net, 16, [1, 1], scope='Conv2d_0a_1x1') branch_2 = slim.conv2d(net, 16, [1, 1], scope='Conv2d_0a_1x1')
branch_2 = slim.conv2d(branch_2, 48, [3, 3], scope='Conv2d_0b_3x3') branch_2 = slim.conv2d(branch_2, 48, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3') branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3')
branch_3 = slim.conv2d(branch_3, 64, [1, 1], scope='Conv2d_0b_1x1') branch_3 = slim.conv2d(branch_3, 64, [1, 1], scope='Conv2d_0b_1x1')
net = tf.concat( net = tf.concat(
...@@ -156,16 +154,16 @@ def inception_v1_base(inputs, ...@@ -156,16 +154,16 @@ def inception_v1_base(inputs,
if final_endpoint == end_point: return net, end_points if final_endpoint == end_point: return net, end_points
end_point = 'Mixed_4c' end_point = 'Mixed_4c'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, 160, [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, 160, [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d(net, 112, [1, 1], scope='Conv2d_0a_1x1') branch_1 = slim.conv2d(net, 112, [1, 1], scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, 224, [3, 3], scope='Conv2d_0b_3x3') branch_1 = slim.conv2d(branch_1, 224, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d(net, 24, [1, 1], scope='Conv2d_0a_1x1') branch_2 = slim.conv2d(net, 24, [1, 1], scope='Conv2d_0a_1x1')
branch_2 = slim.conv2d(branch_2, 64, [3, 3], scope='Conv2d_0b_3x3') branch_2 = slim.conv2d(branch_2, 64, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3') branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3')
branch_3 = slim.conv2d(branch_3, 64, [1, 1], scope='Conv2d_0b_1x1') branch_3 = slim.conv2d(branch_3, 64, [1, 1], scope='Conv2d_0b_1x1')
net = tf.concat( net = tf.concat(
...@@ -174,16 +172,16 @@ def inception_v1_base(inputs, ...@@ -174,16 +172,16 @@ def inception_v1_base(inputs,
if final_endpoint == end_point: return net, end_points if final_endpoint == end_point: return net, end_points
end_point = 'Mixed_4d' end_point = 'Mixed_4d'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, 128, [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, 128, [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d(net, 128, [1, 1], scope='Conv2d_0a_1x1') branch_1 = slim.conv2d(net, 128, [1, 1], scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, 256, [3, 3], scope='Conv2d_0b_3x3') branch_1 = slim.conv2d(branch_1, 256, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d(net, 24, [1, 1], scope='Conv2d_0a_1x1') branch_2 = slim.conv2d(net, 24, [1, 1], scope='Conv2d_0a_1x1')
branch_2 = slim.conv2d(branch_2, 64, [3, 3], scope='Conv2d_0b_3x3') branch_2 = slim.conv2d(branch_2, 64, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3') branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3')
branch_3 = slim.conv2d(branch_3, 64, [1, 1], scope='Conv2d_0b_1x1') branch_3 = slim.conv2d(branch_3, 64, [1, 1], scope='Conv2d_0b_1x1')
net = tf.concat( net = tf.concat(
...@@ -192,16 +190,16 @@ def inception_v1_base(inputs, ...@@ -192,16 +190,16 @@ def inception_v1_base(inputs,
if final_endpoint == end_point: return net, end_points if final_endpoint == end_point: return net, end_points
end_point = 'Mixed_4e' end_point = 'Mixed_4e'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, 112, [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, 112, [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d(net, 144, [1, 1], scope='Conv2d_0a_1x1') branch_1 = slim.conv2d(net, 144, [1, 1], scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, 288, [3, 3], scope='Conv2d_0b_3x3') branch_1 = slim.conv2d(branch_1, 288, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d(net, 32, [1, 1], scope='Conv2d_0a_1x1') branch_2 = slim.conv2d(net, 32, [1, 1], scope='Conv2d_0a_1x1')
branch_2 = slim.conv2d(branch_2, 64, [3, 3], scope='Conv2d_0b_3x3') branch_2 = slim.conv2d(branch_2, 64, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3') branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3')
branch_3 = slim.conv2d(branch_3, 64, [1, 1], scope='Conv2d_0b_1x1') branch_3 = slim.conv2d(branch_3, 64, [1, 1], scope='Conv2d_0b_1x1')
net = tf.concat( net = tf.concat(
...@@ -210,16 +208,16 @@ def inception_v1_base(inputs, ...@@ -210,16 +208,16 @@ def inception_v1_base(inputs,
if final_endpoint == end_point: return net, end_points if final_endpoint == end_point: return net, end_points
end_point = 'Mixed_4f' end_point = 'Mixed_4f'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, 256, [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, 256, [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d(net, 160, [1, 1], scope='Conv2d_0a_1x1') branch_1 = slim.conv2d(net, 160, [1, 1], scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, 320, [3, 3], scope='Conv2d_0b_3x3') branch_1 = slim.conv2d(branch_1, 320, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d(net, 32, [1, 1], scope='Conv2d_0a_1x1') branch_2 = slim.conv2d(net, 32, [1, 1], scope='Conv2d_0a_1x1')
branch_2 = slim.conv2d(branch_2, 128, [3, 3], scope='Conv2d_0b_3x3') branch_2 = slim.conv2d(branch_2, 128, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3') branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3')
branch_3 = slim.conv2d(branch_3, 128, [1, 1], scope='Conv2d_0b_1x1') branch_3 = slim.conv2d(branch_3, 128, [1, 1], scope='Conv2d_0b_1x1')
net = tf.concat( net = tf.concat(
...@@ -233,16 +231,16 @@ def inception_v1_base(inputs, ...@@ -233,16 +231,16 @@ def inception_v1_base(inputs,
if final_endpoint == end_point: return net, end_points if final_endpoint == end_point: return net, end_points
end_point = 'Mixed_5b' end_point = 'Mixed_5b'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, 256, [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, 256, [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d(net, 160, [1, 1], scope='Conv2d_0a_1x1') branch_1 = slim.conv2d(net, 160, [1, 1], scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, 320, [3, 3], scope='Conv2d_0b_3x3') branch_1 = slim.conv2d(branch_1, 320, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d(net, 32, [1, 1], scope='Conv2d_0a_1x1') branch_2 = slim.conv2d(net, 32, [1, 1], scope='Conv2d_0a_1x1')
branch_2 = slim.conv2d(branch_2, 128, [3, 3], scope='Conv2d_0a_3x3') branch_2 = slim.conv2d(branch_2, 128, [3, 3], scope='Conv2d_0a_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3') branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3')
branch_3 = slim.conv2d(branch_3, 128, [1, 1], scope='Conv2d_0b_1x1') branch_3 = slim.conv2d(branch_3, 128, [1, 1], scope='Conv2d_0b_1x1')
net = tf.concat( net = tf.concat(
...@@ -251,16 +249,16 @@ def inception_v1_base(inputs, ...@@ -251,16 +249,16 @@ def inception_v1_base(inputs,
if final_endpoint == end_point: return net, end_points if final_endpoint == end_point: return net, end_points
end_point = 'Mixed_5c' end_point = 'Mixed_5c'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, 384, [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, 384, [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d(net, 192, [1, 1], scope='Conv2d_0a_1x1') branch_1 = slim.conv2d(net, 192, [1, 1], scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, 384, [3, 3], scope='Conv2d_0b_3x3') branch_1 = slim.conv2d(branch_1, 384, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d(net, 48, [1, 1], scope='Conv2d_0a_1x1') branch_2 = slim.conv2d(net, 48, [1, 1], scope='Conv2d_0a_1x1')
branch_2 = slim.conv2d(branch_2, 128, [3, 3], scope='Conv2d_0b_3x3') branch_2 = slim.conv2d(branch_2, 128, [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3') branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3')
branch_3 = slim.conv2d(branch_3, 128, [1, 1], scope='Conv2d_0b_1x1') branch_3 = slim.conv2d(branch_3, 128, [1, 1], scope='Conv2d_0b_1x1')
net = tf.concat( net = tf.concat(
...@@ -316,12 +314,12 @@ def inception_v1(inputs, ...@@ -316,12 +314,12 @@ def inception_v1(inputs,
activation. activation.
""" """
# Final pooling and prediction # Final pooling and prediction
with tf.compat.v1.variable_scope( with tf.variable_scope(
scope, 'InceptionV1', [inputs], reuse=reuse) as scope: scope, 'InceptionV1', [inputs], reuse=reuse) as scope:
with slim.arg_scope([slim.batch_norm, slim.dropout], with slim.arg_scope([slim.batch_norm, slim.dropout],
is_training=is_training): is_training=is_training):
net, end_points = inception_v1_base(inputs, scope=scope) net, end_points = inception_v1_base(inputs, scope=scope)
with tf.compat.v1.variable_scope('Logits'): with tf.variable_scope('Logits'):
if global_pool: if global_pool:
# Global average pooling. # Global average pooling.
net = tf.reduce_mean( net = tf.reduce_mean(
......
research/slim/nets/inception_v1_test.py
View file @ 451906e4
...@@ -19,13 +19,11 @@ from __future__ import division ...@@ -19,13 +19,11 @@ from __future__ import division
from __future__ import print_function from __future__ import print_function
import numpy as np import numpy as np
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import slim as contrib_slim import tf_slim as slim
from nets import inception from nets import inception
slim = contrib_slim
class InceptionV1Test(tf.test.TestCase): class InceptionV1Test(tf.test.TestCase):
...@@ -172,13 +170,13 @@ class InceptionV1Test(tf.test.TestCase): ...@@ -172,13 +170,13 @@ class InceptionV1Test(tf.test.TestCase):
expected_shape) expected_shape)
def testUnknownImageShape(self): def testUnknownImageShape(self):
tf.compat.v1.reset_default_graph() tf.reset_default_graph()
batch_size = 2 batch_size = 2
height, width = 224, 224 height, width = 224, 224
num_classes = 1000 num_classes = 1000
input_np = np.random.uniform(0, 1, (batch_size, height, width, 3)) input_np = np.random.uniform(0, 1, (batch_size, height, width, 3))
with self.test_session() as sess: with self.test_session() as sess:
inputs = tf.compat.v1.placeholder( inputs = tf.placeholder(
tf.float32, shape=(batch_size, None, None, 3)) tf.float32, shape=(batch_size, None, None, 3))
logits, end_points = inception.inception_v1(inputs, num_classes) logits, end_points = inception.inception_v1(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV1/Logits')) self.assertTrue(logits.op.name.startswith('InceptionV1/Logits'))
...@@ -186,18 +184,18 @@ class InceptionV1Test(tf.test.TestCase): ...@@ -186,18 +184,18 @@ class InceptionV1Test(tf.test.TestCase):
[batch_size, num_classes]) [batch_size, num_classes])
pre_pool = end_points['Mixed_5c'] pre_pool = end_points['Mixed_5c']
feed_dict = {inputs: input_np} feed_dict = {inputs: input_np}
tf.compat.v1.global_variables_initializer().run() tf.global_variables_initializer().run()
pre_pool_out = sess.run(pre_pool, feed_dict=feed_dict) pre_pool_out = sess.run(pre_pool, feed_dict=feed_dict)
self.assertListEqual(list(pre_pool_out.shape), [batch_size, 7, 7, 1024]) self.assertListEqual(list(pre_pool_out.shape), [batch_size, 7, 7, 1024])
def testGlobalPoolUnknownImageShape(self): def testGlobalPoolUnknownImageShape(self):
tf.compat.v1.reset_default_graph() tf.reset_default_graph()
batch_size = 1 batch_size = 1
height, width = 250, 300 height, width = 250, 300
num_classes = 1000 num_classes = 1000
input_np = np.random.uniform(0, 1, (batch_size, height, width, 3)) input_np = np.random.uniform(0, 1, (batch_size, height, width, 3))
with self.test_session() as sess: with self.test_session() as sess:
inputs = tf.compat.v1.placeholder( inputs = tf.placeholder(
tf.float32, shape=(batch_size, None, None, 3)) tf.float32, shape=(batch_size, None, None, 3))
logits, end_points = inception.inception_v1(inputs, num_classes, logits, end_points = inception.inception_v1(inputs, num_classes,
global_pool=True) global_pool=True)
...@@ -206,7 +204,7 @@ class InceptionV1Test(tf.test.TestCase): ...@@ -206,7 +204,7 @@ class InceptionV1Test(tf.test.TestCase):
[batch_size, num_classes]) [batch_size, num_classes])
pre_pool = end_points['Mixed_5c'] pre_pool = end_points['Mixed_5c']
feed_dict = {inputs: input_np} feed_dict = {inputs: input_np}
tf.compat.v1.global_variables_initializer().run() tf.global_variables_initializer().run()
pre_pool_out = sess.run(pre_pool, feed_dict=feed_dict) pre_pool_out = sess.run(pre_pool, feed_dict=feed_dict)
self.assertListEqual(list(pre_pool_out.shape), [batch_size, 8, 10, 1024]) self.assertListEqual(list(pre_pool_out.shape), [batch_size, 8, 10, 1024])
...@@ -215,7 +213,7 @@ class InceptionV1Test(tf.test.TestCase): ...@@ -215,7 +213,7 @@ class InceptionV1Test(tf.test.TestCase):
height, width = 224, 224 height, width = 224, 224
num_classes = 1000 num_classes = 1000
inputs = tf.compat.v1.placeholder(tf.float32, (None, height, width, 3)) inputs = tf.placeholder(tf.float32, (None, height, width, 3))
logits, _ = inception.inception_v1(inputs, num_classes) logits, _ = inception.inception_v1(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV1/Logits')) self.assertTrue(logits.op.name.startswith('InceptionV1/Logits'))
self.assertListEqual(logits.get_shape().as_list(), self.assertListEqual(logits.get_shape().as_list(),
...@@ -223,7 +221,7 @@ class InceptionV1Test(tf.test.TestCase): ...@@ -223,7 +221,7 @@ class InceptionV1Test(tf.test.TestCase):
images = tf.random.uniform((batch_size, height, width, 3)) images = tf.random.uniform((batch_size, height, width, 3))
with self.test_session() as sess: with self.test_session() as sess:
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
output = sess.run(logits, {inputs: images.eval()}) output = sess.run(logits, {inputs: images.eval()})
self.assertEquals(output.shape, (batch_size, num_classes)) self.assertEquals(output.shape, (batch_size, num_classes))
...@@ -238,7 +236,7 @@ class InceptionV1Test(tf.test.TestCase): ...@@ -238,7 +236,7 @@ class InceptionV1Test(tf.test.TestCase):
predictions = tf.argmax(input=logits, axis=1) predictions = tf.argmax(input=logits, axis=1)
with self.test_session() as sess: with self.test_session() as sess:
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
output = sess.run(predictions) output = sess.run(predictions)
self.assertEquals(output.shape, (batch_size,)) self.assertEquals(output.shape, (batch_size,))
...@@ -255,7 +253,7 @@ class InceptionV1Test(tf.test.TestCase): ...@@ -255,7 +253,7 @@ class InceptionV1Test(tf.test.TestCase):
predictions = tf.argmax(input=logits, axis=1) predictions = tf.argmax(input=logits, axis=1)
with self.test_session() as sess: with self.test_session() as sess:
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
output = sess.run(predictions) output = sess.run(predictions)
self.assertEquals(output.shape, (eval_batch_size,)) self.assertEquals(output.shape, (eval_batch_size,))
...@@ -267,32 +265,32 @@ class InceptionV1Test(tf.test.TestCase): ...@@ -267,32 +265,32 @@ class InceptionV1Test(tf.test.TestCase):
spatial_squeeze=False) spatial_squeeze=False)
with self.test_session() as sess: with self.test_session() as sess:
tf.compat.v1.global_variables_initializer().run() tf.global_variables_initializer().run()
logits_out = sess.run(logits) logits_out = sess.run(logits)
self.assertListEqual(list(logits_out.shape), [1, 1, 1, num_classes]) self.assertListEqual(list(logits_out.shape), [1, 1, 1, num_classes])
def testNoBatchNormScaleByDefault(self): def testNoBatchNormScaleByDefault(self):
height, width = 224, 224 height, width = 224, 224
num_classes = 1000 num_classes = 1000
inputs = tf.compat.v1.placeholder(tf.float32, (1, height, width, 3)) inputs = tf.placeholder(tf.float32, (1, height, width, 3))
with slim.arg_scope(inception.inception_v1_arg_scope()): with slim.arg_scope(inception.inception_v1_arg_scope()):
inception.inception_v1(inputs, num_classes, is_training=False) inception.inception_v1(inputs, num_classes, is_training=False)
self.assertEqual(tf.compat.v1.global_variables('.*/BatchNorm/gamma:0$'), []) self.assertEqual(tf.global_variables('.*/BatchNorm/gamma:0$'), [])
def testBatchNormScale(self): def testBatchNormScale(self):
height, width = 224, 224 height, width = 224, 224
num_classes = 1000 num_classes = 1000
inputs = tf.compat.v1.placeholder(tf.float32, (1, height, width, 3)) inputs = tf.placeholder(tf.float32, (1, height, width, 3))
with slim.arg_scope( with slim.arg_scope(
inception.inception_v1_arg_scope(batch_norm_scale=True)): inception.inception_v1_arg_scope(batch_norm_scale=True)):
inception.inception_v1(inputs, num_classes, is_training=False) inception.inception_v1(inputs, num_classes, is_training=False)
gamma_names = set( gamma_names = set(
v.op.name v.op.name
for v in tf.compat.v1.global_variables('.*/BatchNorm/gamma:0$')) for v in tf.global_variables('.*/BatchNorm/gamma:0$'))
self.assertGreater(len(gamma_names), 0) self.assertGreater(len(gamma_names), 0)
for v in tf.compat.v1.global_variables('.*/BatchNorm/moving_mean:0$'): for v in tf.global_variables('.*/BatchNorm/moving_mean:0$'):
self.assertIn(v.op.name[:-len('moving_mean')] + 'gamma', gamma_names) self.assertIn(v.op.name[:-len('moving_mean')] + 'gamma', gamma_names)
......
research/slim/nets/inception_v2.py
View file @ 451906e4
...@@ -18,15 +18,13 @@ from __future__ import absolute_import ...@@ -18,15 +18,13 @@ from __future__ import absolute_import
from __future__ import division from __future__ import division
from __future__ import print_function from __future__ import print_function
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import slim as contrib_slim import tf_slim as slim
from nets import inception_utils from nets import inception_utils
slim = contrib_slim
# pylint: disable=g-long-lambda # pylint: disable=g-long-lambda
trunc_normal = lambda stddev: tf.compat.v1.truncated_normal_initializer( trunc_normal = lambda stddev: tf.truncated_normal_initializer(
0.0, stddev) 0.0, stddev)
...@@ -95,7 +93,7 @@ def inception_v2_base(inputs, ...@@ -95,7 +93,7 @@ def inception_v2_base(inputs,
) )
concat_dim = 3 if data_format == 'NHWC' else 1 concat_dim = 3 if data_format == 'NHWC' else 1
with tf.compat.v1.variable_scope(scope, 'InceptionV2', [inputs]): with tf.variable_scope(scope, 'InceptionV2', [inputs]):
with slim.arg_scope( with slim.arg_scope(
[slim.conv2d, slim.max_pool2d, slim.avg_pool2d], [slim.conv2d, slim.max_pool2d, slim.avg_pool2d],
stride=1, stride=1,
...@@ -170,17 +168,17 @@ def inception_v2_base(inputs, ...@@ -170,17 +168,17 @@ def inception_v2_base(inputs,
# 28 x 28 x 192 # 28 x 28 x 192
# Inception module. # Inception module.
end_point = 'Mixed_3b' end_point = 'Mixed_3b'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, depth(64), [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, depth(64), [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d( branch_1 = slim.conv2d(
net, depth(64), [1, 1], net, depth(64), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
scope='Conv2d_0a_1x1') scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, depth(64), [3, 3], branch_1 = slim.conv2d(branch_1, depth(64), [3, 3],
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d( branch_2 = slim.conv2d(
net, depth(64), [1, 1], net, depth(64), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
...@@ -189,7 +187,7 @@ def inception_v2_base(inputs, ...@@ -189,7 +187,7 @@ def inception_v2_base(inputs,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
branch_2 = slim.conv2d(branch_2, depth(96), [3, 3], branch_2 = slim.conv2d(branch_2, depth(96), [3, 3],
scope='Conv2d_0c_3x3') scope='Conv2d_0c_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3') branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3')
branch_3 = slim.conv2d( branch_3 = slim.conv2d(
branch_3, depth(32), [1, 1], branch_3, depth(32), [1, 1],
...@@ -201,17 +199,17 @@ def inception_v2_base(inputs, ...@@ -201,17 +199,17 @@ def inception_v2_base(inputs,
if end_point == final_endpoint: return net, end_points if end_point == final_endpoint: return net, end_points
# 28 x 28 x 256 # 28 x 28 x 256
end_point = 'Mixed_3c' end_point = 'Mixed_3c'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, depth(64), [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, depth(64), [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d( branch_1 = slim.conv2d(
net, depth(64), [1, 1], net, depth(64), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
scope='Conv2d_0a_1x1') scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, depth(96), [3, 3], branch_1 = slim.conv2d(branch_1, depth(96), [3, 3],
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d( branch_2 = slim.conv2d(
net, depth(64), [1, 1], net, depth(64), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
...@@ -220,7 +218,7 @@ def inception_v2_base(inputs, ...@@ -220,7 +218,7 @@ def inception_v2_base(inputs,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
branch_2 = slim.conv2d(branch_2, depth(96), [3, 3], branch_2 = slim.conv2d(branch_2, depth(96), [3, 3],
scope='Conv2d_0c_3x3') scope='Conv2d_0c_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3') branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3')
branch_3 = slim.conv2d( branch_3 = slim.conv2d(
branch_3, depth(64), [1, 1], branch_3, depth(64), [1, 1],
...@@ -232,15 +230,15 @@ def inception_v2_base(inputs, ...@@ -232,15 +230,15 @@ def inception_v2_base(inputs,
if end_point == final_endpoint: return net, end_points if end_point == final_endpoint: return net, end_points
# 28 x 28 x 320 # 28 x 28 x 320
end_point = 'Mixed_4a' end_point = 'Mixed_4a'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d( branch_0 = slim.conv2d(
net, depth(128), [1, 1], net, depth(128), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
scope='Conv2d_0a_1x1') scope='Conv2d_0a_1x1')
branch_0 = slim.conv2d(branch_0, depth(160), [3, 3], stride=2, branch_0 = slim.conv2d(branch_0, depth(160), [3, 3], stride=2,
scope='Conv2d_1a_3x3') scope='Conv2d_1a_3x3')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d( branch_1 = slim.conv2d(
net, depth(64), [1, 1], net, depth(64), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
...@@ -249,7 +247,7 @@ def inception_v2_base(inputs, ...@@ -249,7 +247,7 @@ def inception_v2_base(inputs,
branch_1, depth(96), [3, 3], scope='Conv2d_0b_3x3') branch_1, depth(96), [3, 3], scope='Conv2d_0b_3x3')
branch_1 = slim.conv2d( branch_1 = slim.conv2d(
branch_1, depth(96), [3, 3], stride=2, scope='Conv2d_1a_3x3') branch_1, depth(96), [3, 3], stride=2, scope='Conv2d_1a_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.max_pool2d( branch_2 = slim.max_pool2d(
net, [3, 3], stride=2, scope='MaxPool_1a_3x3') net, [3, 3], stride=2, scope='MaxPool_1a_3x3')
net = tf.concat(axis=concat_dim, values=[branch_0, branch_1, branch_2]) net = tf.concat(axis=concat_dim, values=[branch_0, branch_1, branch_2])
...@@ -257,17 +255,17 @@ def inception_v2_base(inputs, ...@@ -257,17 +255,17 @@ def inception_v2_base(inputs,
if end_point == final_endpoint: return net, end_points if end_point == final_endpoint: return net, end_points
# 14 x 14 x 576 # 14 x 14 x 576
end_point = 'Mixed_4b' end_point = 'Mixed_4b'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, depth(224), [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, depth(224), [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d( branch_1 = slim.conv2d(
net, depth(64), [1, 1], net, depth(64), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
scope='Conv2d_0a_1x1') scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d( branch_1 = slim.conv2d(
branch_1, depth(96), [3, 3], scope='Conv2d_0b_3x3') branch_1, depth(96), [3, 3], scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d( branch_2 = slim.conv2d(
net, depth(96), [1, 1], net, depth(96), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
...@@ -276,7 +274,7 @@ def inception_v2_base(inputs, ...@@ -276,7 +274,7 @@ def inception_v2_base(inputs,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
branch_2 = slim.conv2d(branch_2, depth(128), [3, 3], branch_2 = slim.conv2d(branch_2, depth(128), [3, 3],
scope='Conv2d_0c_3x3') scope='Conv2d_0c_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3') branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3')
branch_3 = slim.conv2d( branch_3 = slim.conv2d(
branch_3, depth(128), [1, 1], branch_3, depth(128), [1, 1],
...@@ -288,17 +286,17 @@ def inception_v2_base(inputs, ...@@ -288,17 +286,17 @@ def inception_v2_base(inputs,
if end_point == final_endpoint: return net, end_points if end_point == final_endpoint: return net, end_points
# 14 x 14 x 576 # 14 x 14 x 576
end_point = 'Mixed_4c' end_point = 'Mixed_4c'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, depth(192), [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, depth(192), [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d( branch_1 = slim.conv2d(
net, depth(96), [1, 1], net, depth(96), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
scope='Conv2d_0a_1x1') scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, depth(128), [3, 3], branch_1 = slim.conv2d(branch_1, depth(128), [3, 3],
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d( branch_2 = slim.conv2d(
net, depth(96), [1, 1], net, depth(96), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
...@@ -307,7 +305,7 @@ def inception_v2_base(inputs, ...@@ -307,7 +305,7 @@ def inception_v2_base(inputs,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
branch_2 = slim.conv2d(branch_2, depth(128), [3, 3], branch_2 = slim.conv2d(branch_2, depth(128), [3, 3],
scope='Conv2d_0c_3x3') scope='Conv2d_0c_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3') branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3')
branch_3 = slim.conv2d( branch_3 = slim.conv2d(
branch_3, depth(128), [1, 1], branch_3, depth(128), [1, 1],
...@@ -319,17 +317,17 @@ def inception_v2_base(inputs, ...@@ -319,17 +317,17 @@ def inception_v2_base(inputs,
if end_point == final_endpoint: return net, end_points if end_point == final_endpoint: return net, end_points
# 14 x 14 x 576 # 14 x 14 x 576
end_point = 'Mixed_4d' end_point = 'Mixed_4d'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, depth(160), [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, depth(160), [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d( branch_1 = slim.conv2d(
net, depth(128), [1, 1], net, depth(128), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
scope='Conv2d_0a_1x1') scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, depth(160), [3, 3], branch_1 = slim.conv2d(branch_1, depth(160), [3, 3],
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d( branch_2 = slim.conv2d(
net, depth(128), [1, 1], net, depth(128), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
...@@ -338,7 +336,7 @@ def inception_v2_base(inputs, ...@@ -338,7 +336,7 @@ def inception_v2_base(inputs,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
branch_2 = slim.conv2d(branch_2, depth(160), [3, 3], branch_2 = slim.conv2d(branch_2, depth(160), [3, 3],
scope='Conv2d_0c_3x3') scope='Conv2d_0c_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3') branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3')
branch_3 = slim.conv2d( branch_3 = slim.conv2d(
branch_3, depth(96), [1, 1], branch_3, depth(96), [1, 1],
...@@ -350,17 +348,17 @@ def inception_v2_base(inputs, ...@@ -350,17 +348,17 @@ def inception_v2_base(inputs,
if end_point == final_endpoint: return net, end_points if end_point == final_endpoint: return net, end_points
# 14 x 14 x 576 # 14 x 14 x 576
end_point = 'Mixed_4e' end_point = 'Mixed_4e'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, depth(96), [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, depth(96), [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d( branch_1 = slim.conv2d(
net, depth(128), [1, 1], net, depth(128), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
scope='Conv2d_0a_1x1') scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, depth(192), [3, 3], branch_1 = slim.conv2d(branch_1, depth(192), [3, 3],
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d( branch_2 = slim.conv2d(
net, depth(160), [1, 1], net, depth(160), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
...@@ -369,7 +367,7 @@ def inception_v2_base(inputs, ...@@ -369,7 +367,7 @@ def inception_v2_base(inputs,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
branch_2 = slim.conv2d(branch_2, depth(192), [3, 3], branch_2 = slim.conv2d(branch_2, depth(192), [3, 3],
scope='Conv2d_0c_3x3') scope='Conv2d_0c_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3') branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3')
branch_3 = slim.conv2d( branch_3 = slim.conv2d(
branch_3, depth(96), [1, 1], branch_3, depth(96), [1, 1],
...@@ -381,15 +379,15 @@ def inception_v2_base(inputs, ...@@ -381,15 +379,15 @@ def inception_v2_base(inputs,
if end_point == final_endpoint: return net, end_points if end_point == final_endpoint: return net, end_points
# 14 x 14 x 576 # 14 x 14 x 576
end_point = 'Mixed_5a' end_point = 'Mixed_5a'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d( branch_0 = slim.conv2d(
net, depth(128), [1, 1], net, depth(128), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
scope='Conv2d_0a_1x1') scope='Conv2d_0a_1x1')
branch_0 = slim.conv2d(branch_0, depth(192), [3, 3], stride=2, branch_0 = slim.conv2d(branch_0, depth(192), [3, 3], stride=2,
scope='Conv2d_1a_3x3') scope='Conv2d_1a_3x3')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d( branch_1 = slim.conv2d(
net, depth(192), [1, 1], net, depth(192), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
...@@ -398,7 +396,7 @@ def inception_v2_base(inputs, ...@@ -398,7 +396,7 @@ def inception_v2_base(inputs,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
branch_1 = slim.conv2d(branch_1, depth(256), [3, 3], stride=2, branch_1 = slim.conv2d(branch_1, depth(256), [3, 3], stride=2,
scope='Conv2d_1a_3x3') scope='Conv2d_1a_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.max_pool2d(net, [3, 3], stride=2, branch_2 = slim.max_pool2d(net, [3, 3], stride=2,
scope='MaxPool_1a_3x3') scope='MaxPool_1a_3x3')
net = tf.concat( net = tf.concat(
...@@ -407,17 +405,17 @@ def inception_v2_base(inputs, ...@@ -407,17 +405,17 @@ def inception_v2_base(inputs,
if end_point == final_endpoint: return net, end_points if end_point == final_endpoint: return net, end_points
# 7 x 7 x 1024 # 7 x 7 x 1024
end_point = 'Mixed_5b' end_point = 'Mixed_5b'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, depth(352), [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, depth(352), [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d( branch_1 = slim.conv2d(
net, depth(192), [1, 1], net, depth(192), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
scope='Conv2d_0a_1x1') scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, depth(320), [3, 3], branch_1 = slim.conv2d(branch_1, depth(320), [3, 3],
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d( branch_2 = slim.conv2d(
net, depth(160), [1, 1], net, depth(160), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
...@@ -426,7 +424,7 @@ def inception_v2_base(inputs, ...@@ -426,7 +424,7 @@ def inception_v2_base(inputs,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
branch_2 = slim.conv2d(branch_2, depth(224), [3, 3], branch_2 = slim.conv2d(branch_2, depth(224), [3, 3],
scope='Conv2d_0c_3x3') scope='Conv2d_0c_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3') branch_3 = slim.avg_pool2d(net, [3, 3], scope='AvgPool_0a_3x3')
branch_3 = slim.conv2d( branch_3 = slim.conv2d(
branch_3, depth(128), [1, 1], branch_3, depth(128), [1, 1],
...@@ -438,17 +436,17 @@ def inception_v2_base(inputs, ...@@ -438,17 +436,17 @@ def inception_v2_base(inputs,
if end_point == final_endpoint: return net, end_points if end_point == final_endpoint: return net, end_points
# 7 x 7 x 1024 # 7 x 7 x 1024
end_point = 'Mixed_5c' end_point = 'Mixed_5c'
with tf.compat.v1.variable_scope(end_point): with tf.variable_scope(end_point):
with tf.compat.v1.variable_scope('Branch_0'): with tf.variable_scope('Branch_0'):
branch_0 = slim.conv2d(net, depth(352), [1, 1], scope='Conv2d_0a_1x1') branch_0 = slim.conv2d(net, depth(352), [1, 1], scope='Conv2d_0a_1x1')
with tf.compat.v1.variable_scope('Branch_1'): with tf.variable_scope('Branch_1'):
branch_1 = slim.conv2d( branch_1 = slim.conv2d(
net, depth(192), [1, 1], net, depth(192), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
scope='Conv2d_0a_1x1') scope='Conv2d_0a_1x1')
branch_1 = slim.conv2d(branch_1, depth(320), [3, 3], branch_1 = slim.conv2d(branch_1, depth(320), [3, 3],
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
with tf.compat.v1.variable_scope('Branch_2'): with tf.variable_scope('Branch_2'):
branch_2 = slim.conv2d( branch_2 = slim.conv2d(
net, depth(192), [1, 1], net, depth(192), [1, 1],
weights_initializer=trunc_normal(0.09), weights_initializer=trunc_normal(0.09),
...@@ -457,7 +455,7 @@ def inception_v2_base(inputs, ...@@ -457,7 +455,7 @@ def inception_v2_base(inputs,
scope='Conv2d_0b_3x3') scope='Conv2d_0b_3x3')
branch_2 = slim.conv2d(branch_2, depth(224), [3, 3], branch_2 = slim.conv2d(branch_2, depth(224), [3, 3],
scope='Conv2d_0c_3x3') scope='Conv2d_0c_3x3')
with tf.compat.v1.variable_scope('Branch_3'): with tf.variable_scope('Branch_3'):
branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3') branch_3 = slim.max_pool2d(net, [3, 3], scope='MaxPool_0a_3x3')
branch_3 = slim.conv2d( branch_3 = slim.conv2d(
branch_3, depth(128), [1, 1], branch_3, depth(128), [1, 1],
...@@ -528,14 +526,14 @@ def inception_v2(inputs, ...@@ -528,14 +526,14 @@ def inception_v2(inputs,
raise ValueError('depth_multiplier is not greater than zero.') raise ValueError('depth_multiplier is not greater than zero.')
# Final pooling and prediction # Final pooling and prediction
with tf.compat.v1.variable_scope( with tf.variable_scope(
scope, 'InceptionV2', [inputs], reuse=reuse) as scope: scope, 'InceptionV2', [inputs], reuse=reuse) as scope:
with slim.arg_scope([slim.batch_norm, slim.dropout], with slim.arg_scope([slim.batch_norm, slim.dropout],
is_training=is_training): is_training=is_training):
net, end_points = inception_v2_base( net, end_points = inception_v2_base(
inputs, scope=scope, min_depth=min_depth, inputs, scope=scope, min_depth=min_depth,
depth_multiplier=depth_multiplier) depth_multiplier=depth_multiplier)
with tf.compat.v1.variable_scope('Logits'): with tf.variable_scope('Logits'):
if global_pool: if global_pool:
# Global average pooling. # Global average pooling.
net = tf.reduce_mean( net = tf.reduce_mean(
......
research/slim/nets/inception_v2_test.py
View file @ 451906e4
...@@ -19,13 +19,11 @@ from __future__ import division ...@@ -19,13 +19,11 @@ from __future__ import division
from __future__ import print_function from __future__ import print_function
import numpy as np import numpy as np
import tensorflow as tf import tensorflow.compat.v1 as tf
from tensorflow.contrib import slim as contrib_slim import tf_slim as slim
from nets import inception from nets import inception
slim = contrib_slim
class InceptionV2Test(tf.test.TestCase): class InceptionV2Test(tf.test.TestCase):
...@@ -284,13 +282,13 @@ class InceptionV2Test(tf.test.TestCase): ...@@ -284,13 +282,13 @@ class InceptionV2Test(tf.test.TestCase):
expected_shape) expected_shape)
def testUnknownImageShape(self): def testUnknownImageShape(self):
tf.compat.v1.reset_default_graph() tf.reset_default_graph()
batch_size = 2 batch_size = 2
height, width = 224, 224 height, width = 224, 224
num_classes = 1000 num_classes = 1000
input_np = np.random.uniform(0, 1, (batch_size, height, width, 3)) input_np = np.random.uniform(0, 1, (batch_size, height, width, 3))
with self.test_session() as sess: with self.test_session() as sess:
inputs = tf.compat.v1.placeholder( inputs = tf.placeholder(
tf.float32, shape=(batch_size, None, None, 3)) tf.float32, shape=(batch_size, None, None, 3))
logits, end_points = inception.inception_v2(inputs, num_classes) logits, end_points = inception.inception_v2(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV2/Logits')) self.assertTrue(logits.op.name.startswith('InceptionV2/Logits'))
...@@ -298,18 +296,18 @@ class InceptionV2Test(tf.test.TestCase): ...@@ -298,18 +296,18 @@ class InceptionV2Test(tf.test.TestCase):
[batch_size, num_classes]) [batch_size, num_classes])
pre_pool = end_points['Mixed_5c'] pre_pool = end_points['Mixed_5c']
feed_dict = {inputs: input_np} feed_dict = {inputs: input_np}
tf.compat.v1.global_variables_initializer().run() tf.global_variables_initializer().run()
pre_pool_out = sess.run(pre_pool, feed_dict=feed_dict) pre_pool_out = sess.run(pre_pool, feed_dict=feed_dict)
self.assertListEqual(list(pre_pool_out.shape), [batch_size, 7, 7, 1024]) self.assertListEqual(list(pre_pool_out.shape), [batch_size, 7, 7, 1024])
def testGlobalPoolUnknownImageShape(self): def testGlobalPoolUnknownImageShape(self):
tf.compat.v1.reset_default_graph() tf.reset_default_graph()
batch_size = 1 batch_size = 1
height, width = 250, 300 height, width = 250, 300
num_classes = 1000 num_classes = 1000
input_np = np.random.uniform(0, 1, (batch_size, height, width, 3)) input_np = np.random.uniform(0, 1, (batch_size, height, width, 3))
with self.test_session() as sess: with self.test_session() as sess:
inputs = tf.compat.v1.placeholder( inputs = tf.placeholder(
tf.float32, shape=(batch_size, None, None, 3)) tf.float32, shape=(batch_size, None, None, 3))
logits, end_points = inception.inception_v2(inputs, num_classes, logits, end_points = inception.inception_v2(inputs, num_classes,
global_pool=True) global_pool=True)
...@@ -318,7 +316,7 @@ class InceptionV2Test(tf.test.TestCase): ...@@ -318,7 +316,7 @@ class InceptionV2Test(tf.test.TestCase):
[batch_size, num_classes]) [batch_size, num_classes])
pre_pool = end_points['Mixed_5c'] pre_pool = end_points['Mixed_5c']
feed_dict = {inputs: input_np} feed_dict = {inputs: input_np}
tf.compat.v1.global_variables_initializer().run() tf.global_variables_initializer().run()
pre_pool_out = sess.run(pre_pool, feed_dict=feed_dict) pre_pool_out = sess.run(pre_pool, feed_dict=feed_dict)
self.assertListEqual(list(pre_pool_out.shape), [batch_size, 8, 10, 1024]) self.assertListEqual(list(pre_pool_out.shape), [batch_size, 8, 10, 1024])
...@@ -327,7 +325,7 @@ class InceptionV2Test(tf.test.TestCase): ...@@ -327,7 +325,7 @@ class InceptionV2Test(tf.test.TestCase):
height, width = 224, 224 height, width = 224, 224
num_classes = 1000 num_classes = 1000
inputs = tf.compat.v1.placeholder(tf.float32, (None, height, width, 3)) inputs = tf.placeholder(tf.float32, (None, height, width, 3))
logits, _ = inception.inception_v2(inputs, num_classes) logits, _ = inception.inception_v2(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV2/Logits')) self.assertTrue(logits.op.name.startswith('InceptionV2/Logits'))
self.assertListEqual(logits.get_shape().as_list(), self.assertListEqual(logits.get_shape().as_list(),
...@@ -335,7 +333,7 @@ class InceptionV2Test(tf.test.TestCase): ...@@ -335,7 +333,7 @@ class InceptionV2Test(tf.test.TestCase):
images = tf.random.uniform((batch_size, height, width, 3)) images = tf.random.uniform((batch_size, height, width, 3))
with self.test_session() as sess: with self.test_session() as sess:
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
output = sess.run(logits, {inputs: images.eval()}) output = sess.run(logits, {inputs: images.eval()})
self.assertEquals(output.shape, (batch_size, num_classes)) self.assertEquals(output.shape, (batch_size, num_classes))
...@@ -350,7 +348,7 @@ class InceptionV2Test(tf.test.TestCase): ...@@ -350,7 +348,7 @@ class InceptionV2Test(tf.test.TestCase):
predictions = tf.argmax(input=logits, axis=1) predictions = tf.argmax(input=logits, axis=1)
with self.test_session() as sess: with self.test_session() as sess:
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
output = sess.run(predictions) output = sess.run(predictions)
self.assertEquals(output.shape, (batch_size,)) self.assertEquals(output.shape, (batch_size,))
...@@ -367,7 +365,7 @@ class InceptionV2Test(tf.test.TestCase): ...@@ -367,7 +365,7 @@ class InceptionV2Test(tf.test.TestCase):
predictions = tf.argmax(input=logits, axis=1) predictions = tf.argmax(input=logits, axis=1)
with self.test_session() as sess: with self.test_session() as sess:
sess.run(tf.compat.v1.global_variables_initializer()) sess.run(tf.global_variables_initializer())
output = sess.run(predictions) output = sess.run(predictions)
self.assertEquals(output.shape, (eval_batch_size,)) self.assertEquals(output.shape, (eval_batch_size,))
...@@ -379,32 +377,32 @@ class InceptionV2Test(tf.test.TestCase): ...@@ -379,32 +377,32 @@ class InceptionV2Test(tf.test.TestCase):
spatial_squeeze=False) spatial_squeeze=False)
with self.test_session() as sess: with self.test_session() as sess:
tf.compat.v1.global_variables_initializer().run() tf.global_variables_initializer().run()
logits_out = sess.run(logits) logits_out = sess.run(logits)
self.assertListEqual(list(logits_out.shape), [1, 1, 1, num_classes]) self.assertListEqual(list(logits_out.shape), [1, 1, 1, num_classes])
def testNoBatchNormScaleByDefault(self): def testNoBatchNormScaleByDefault(self):
height, width = 224, 224 height, width = 224, 224
num_classes = 1000 num_classes = 1000
inputs = tf.compat.v1.placeholder(tf.float32, (1, height, width, 3)) inputs = tf.placeholder(tf.float32, (1, height, width, 3))
with slim.arg_scope(inception.inception_v2_arg_scope()): with slim.arg_scope(inception.inception_v2_arg_scope()):
inception.inception_v2(inputs, num_classes, is_training=False) inception.inception_v2(inputs, num_classes, is_training=False)
self.assertEqual(tf.compat.v1.global_variables('.*/BatchNorm/gamma:0$'), []) self.assertEqual(tf.global_variables('.*/BatchNorm/gamma:0$'), [])
def testBatchNormScale(self): def testBatchNormScale(self):
height, width = 224, 224 height, width = 224, 224
num_classes = 1000 num_classes = 1000
inputs = tf.compat.v1.placeholder(tf.float32, (1, height, width, 3)) inputs = tf.placeholder(tf.float32, (1, height, width, 3))
with slim.arg_scope( with slim.arg_scope(
inception.inception_v2_arg_scope(batch_norm_scale=True)): inception.inception_v2_arg_scope(batch_norm_scale=True)):
inception.inception_v2(inputs, num_classes, is_training=False) inception.inception_v2(inputs, num_classes, is_training=False)
gamma_names = set( gamma_names = set(
v.op.name v.op.name
for v in tf.compat.v1.global_variables('.*/BatchNorm/gamma:0$')) for v in tf.global_variables('.*/BatchNorm/gamma:0$'))
self.assertGreater(len(gamma_names), 0) self.assertGreater(len(gamma_names), 0)
for v in tf.compat.v1.global_variables('.*/BatchNorm/moving_mean:0$'): for v in tf.global_variables('.*/BatchNorm/moving_mean:0$'):
self.assertIn(v.op.name[:-len('moving_mean')] + 'gamma', gamma_names) self.assertIn(v.op.name[:-len('moving_mean')] + 'gamma', gamma_names)
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment