Skip to content

GitLab

"references/vscode:/vscode.git/clone" did not exist on "d4e5aa21afadb3608f6244d15e7bcd2a9e7d6081"
Unverified Commit 57014e4c authored by gunan's avatar gunan Committed by GitHub
Browse files

Merge pull request #2627 from sguada/nasnet 

Bring tensorflow/models slim up to date.
parents c46caa56 cbb62479
Hide whitespace changes
Inline Side-by-side
Showing with 383 additions and 56 deletions
research/slim/nets/resnet_v2_test.py
View file @ 57014e4c
...@@ -285,6 +285,31 @@ class ResnetCompleteNetworkTest(tf.test.TestCase): ...@@ -285,6 +285,31 @@ class ResnetCompleteNetworkTest(tf.test.TestCase):
self.assertTrue('predictions' in end_points) self.assertTrue('predictions' in end_points)
self.assertListEqual(end_points['predictions'].get_shape().as_list(), self.assertListEqual(end_points['predictions'].get_shape().as_list(),
[2, 1, 1, num_classes]) [2, 1, 1, num_classes])
self.assertTrue('global_pool' in end_points)
self.assertListEqual(end_points['global_pool'].get_shape().as_list(),
[2, 1, 1, 32])
def testEndpointNames(self):
# Like ResnetUtilsTest.testEndPointsV2(), but for the public API.
global_pool = True
num_classes = 10
inputs = create_test_input(2, 224, 224, 3)
with slim.arg_scope(resnet_utils.resnet_arg_scope()):
_, end_points = self._resnet_small(inputs, num_classes,
global_pool=global_pool,
scope='resnet')
expected = ['resnet/conv1']
for block in range(1, 5):
for unit in range(1, 4 if block < 4 else 3):
for conv in range(1, 4):
expected.append('resnet/block%d/unit_%d/bottleneck_v2/conv%d' %
(block, unit, conv))
expected.append('resnet/block%d/unit_%d/bottleneck_v2' % (block, unit))
expected.append('resnet/block%d/unit_1/bottleneck_v2/shortcut' % block)
expected.append('resnet/block%d' % block)
expected.extend(['global_pool', 'resnet/logits', 'resnet/spatial_squeeze',
'predictions'])
self.assertItemsEqual(end_points.keys(), expected)
def testClassificationShapes(self): def testClassificationShapes(self):
global_pool = True global_pool = True
......
research/slim/nets/vgg.py
View file @ 57014e4c
...@@ -69,7 +69,8 @@ def vgg_a(inputs, ...@@ -69,7 +69,8 @@ def vgg_a(inputs,
dropout_keep_prob=0.5, dropout_keep_prob=0.5,
spatial_squeeze=True, spatial_squeeze=True,
scope='vgg_a', scope='vgg_a',
fc_conv_padding='VALID'): fc_conv_padding='VALID',
global_pool=False):
"""Oxford Net VGG 11-Layers version A Example. """Oxford Net VGG 11-Layers version A Example.
Note: All the fully_connected layers have been transformed to conv2d layers. Note: All the fully_connected layers have been transformed to conv2d layers.
...@@ -77,7 +78,8 @@ def vgg_a(inputs, ...@@ -77,7 +78,8 @@ def vgg_a(inputs,
Args: Args:
inputs: a tensor of size [batch_size, height, width, channels]. inputs: a tensor of size [batch_size, height, width, channels].
num_classes: number of predicted classes. num_classes: number of predicted classes. If 0 or None, the logits layer is
omitted and the input features to the logits layer are returned instead.
is_training: whether or not the model is being trained. is_training: whether or not the model is being trained.
dropout_keep_prob: the probability that activations are kept in the dropout dropout_keep_prob: the probability that activations are kept in the dropout
layers during training. layers during training.
...@@ -90,12 +92,17 @@ def vgg_a(inputs, ...@@ -90,12 +92,17 @@ def vgg_a(inputs,
get a prediction map downsampled by a factor of 32 as an output. get a prediction map downsampled by a factor of 32 as an output.
Otherwise, the output prediction map will be (input / 32) - 6 in case of Otherwise, the output prediction map will be (input / 32) - 6 in case of
'VALID' padding. 'VALID' padding.
global_pool: Optional boolean flag. If True, the input to the classification
layer is avgpooled to size 1x1, for any input size. (This is not part
of the original VGG architecture.)
Returns: Returns:
the last op containing the log predictions and end_points dict. net: the output of the logits layer (if num_classes is a non-zero integer),
or the input to the logits layer (if num_classes is 0 or None).
end_points: a dict of tensors with intermediate activations.
""" """
with tf.variable_scope(scope, 'vgg_a', [inputs]) as sc: with tf.variable_scope(scope, 'vgg_a', [inputs]) as sc:
end_points_collection = sc.name + '_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.max_pool2d], with slim.arg_scope([slim.conv2d, slim.max_pool2d],
outputs_collections=end_points_collection): outputs_collections=end_points_collection):
...@@ -109,21 +116,26 @@ def vgg_a(inputs, ...@@ -109,21 +116,26 @@ def vgg_a(inputs,
net = slim.max_pool2d(net, [2, 2], scope='pool4') net = slim.max_pool2d(net, [2, 2], scope='pool4')
net = slim.repeat(net, 2, slim.conv2d, 512, [3, 3], scope='conv5') net = slim.repeat(net, 2, slim.conv2d, 512, [3, 3], scope='conv5')
net = slim.max_pool2d(net, [2, 2], scope='pool5') net = slim.max_pool2d(net, [2, 2], scope='pool5')
# Use conv2d instead of fully_connected layers. # Use conv2d instead of fully_connected layers.
net = slim.conv2d(net, 4096, [7, 7], padding=fc_conv_padding, scope='fc6') net = slim.conv2d(net, 4096, [7, 7], padding=fc_conv_padding, scope='fc6')
net = slim.dropout(net, dropout_keep_prob, is_training=is_training, net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
scope='dropout6') scope='dropout6')
net = slim.conv2d(net, 4096, [1, 1], scope='fc7') net = slim.conv2d(net, 4096, [1, 1], scope='fc7')
net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
scope='dropout7')
net = slim.conv2d(net, num_classes, [1, 1],
activation_fn=None,
normalizer_fn=None,
scope='fc8')
# Convert end_points_collection into a end_point dict. # Convert end_points_collection into a end_point dict.
end_points = slim.utils.convert_collection_to_dict(end_points_collection) end_points = slim.utils.convert_collection_to_dict(end_points_collection)
if spatial_squeeze: if global_pool:
net = tf.squeeze(net, [1, 2], name='fc8/squeezed') net = tf.reduce_mean(net, [1, 2], keep_dims=True, name='global_pool')
end_points['global_pool'] = net
if num_classes:
net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
scope='dropout7')
net = slim.conv2d(net, num_classes, [1, 1],
activation_fn=None,
normalizer_fn=None,
scope='fc8')
if spatial_squeeze:
net = tf.squeeze(net, [1, 2], name='fc8/squeezed')
end_points[sc.name + '/fc8'] = net end_points[sc.name + '/fc8'] = net
return net, end_points return net, end_points
vgg_a.default_image_size = 224 vgg_a.default_image_size = 224
...@@ -135,7 +147,8 @@ def vgg_16(inputs, ...@@ -135,7 +147,8 @@ def vgg_16(inputs,
dropout_keep_prob=0.5, dropout_keep_prob=0.5,
spatial_squeeze=True, spatial_squeeze=True,
scope='vgg_16', scope='vgg_16',
fc_conv_padding='VALID'): fc_conv_padding='VALID',
global_pool=False):
"""Oxford Net VGG 16-Layers version D Example. """Oxford Net VGG 16-Layers version D Example.
Note: All the fully_connected layers have been transformed to conv2d layers. Note: All the fully_connected layers have been transformed to conv2d layers.
...@@ -143,7 +156,8 @@ def vgg_16(inputs, ...@@ -143,7 +156,8 @@ def vgg_16(inputs,
Args: Args:
inputs: a tensor of size [batch_size, height, width, channels]. inputs: a tensor of size [batch_size, height, width, channels].
num_classes: number of predicted classes. num_classes: number of predicted classes. If 0 or None, the logits layer is
omitted and the input features to the logits layer are returned instead.
is_training: whether or not the model is being trained. is_training: whether or not the model is being trained.
dropout_keep_prob: the probability that activations are kept in the dropout dropout_keep_prob: the probability that activations are kept in the dropout
layers during training. layers during training.
...@@ -156,12 +170,17 @@ def vgg_16(inputs, ...@@ -156,12 +170,17 @@ def vgg_16(inputs,
get a prediction map downsampled by a factor of 32 as an output. get a prediction map downsampled by a factor of 32 as an output.
Otherwise, the output prediction map will be (input / 32) - 6 in case of Otherwise, the output prediction map will be (input / 32) - 6 in case of
'VALID' padding. 'VALID' padding.
global_pool: Optional boolean flag. If True, the input to the classification
layer is avgpooled to size 1x1, for any input size. (This is not part
of the original VGG architecture.)
Returns: Returns:
the last op containing the log predictions and end_points dict. net: the output of the logits layer (if num_classes is a non-zero integer),
or the input to the logits layer (if num_classes is 0 or None).
end_points: a dict of tensors with intermediate activations.
""" """
with tf.variable_scope(scope, 'vgg_16', [inputs]) as sc: with tf.variable_scope(scope, 'vgg_16', [inputs]) as sc:
end_points_collection = sc.name + '_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],
outputs_collections=end_points_collection): outputs_collections=end_points_collection):
...@@ -175,21 +194,26 @@ def vgg_16(inputs, ...@@ -175,21 +194,26 @@ def vgg_16(inputs,
net = slim.max_pool2d(net, [2, 2], scope='pool4') net = slim.max_pool2d(net, [2, 2], scope='pool4')
net = slim.repeat(net, 3, slim.conv2d, 512, [3, 3], scope='conv5') net = slim.repeat(net, 3, slim.conv2d, 512, [3, 3], scope='conv5')
net = slim.max_pool2d(net, [2, 2], scope='pool5') net = slim.max_pool2d(net, [2, 2], scope='pool5')
# Use conv2d instead of fully_connected layers. # Use conv2d instead of fully_connected layers.
net = slim.conv2d(net, 4096, [7, 7], padding=fc_conv_padding, scope='fc6') net = slim.conv2d(net, 4096, [7, 7], padding=fc_conv_padding, scope='fc6')
net = slim.dropout(net, dropout_keep_prob, is_training=is_training, net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
scope='dropout6') scope='dropout6')
net = slim.conv2d(net, 4096, [1, 1], scope='fc7') net = slim.conv2d(net, 4096, [1, 1], scope='fc7')
net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
scope='dropout7')
net = slim.conv2d(net, num_classes, [1, 1],
activation_fn=None,
normalizer_fn=None,
scope='fc8')
# Convert end_points_collection into a end_point dict. # Convert end_points_collection into a end_point dict.
end_points = slim.utils.convert_collection_to_dict(end_points_collection) end_points = slim.utils.convert_collection_to_dict(end_points_collection)
if spatial_squeeze: if global_pool:
net = tf.squeeze(net, [1, 2], name='fc8/squeezed') net = tf.reduce_mean(net, [1, 2], keep_dims=True, name='global_pool')
end_points['global_pool'] = net
if num_classes:
net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
scope='dropout7')
net = slim.conv2d(net, num_classes, [1, 1],
activation_fn=None,
normalizer_fn=None,
scope='fc8')
if spatial_squeeze and num_classes is not None:
net = tf.squeeze(net, [1, 2], name='fc8/squeezed')
end_points[sc.name + '/fc8'] = net end_points[sc.name + '/fc8'] = net
return net, end_points return net, end_points
vgg_16.default_image_size = 224 vgg_16.default_image_size = 224
...@@ -201,7 +225,8 @@ def vgg_19(inputs, ...@@ -201,7 +225,8 @@ def vgg_19(inputs,
dropout_keep_prob=0.5, dropout_keep_prob=0.5,
spatial_squeeze=True, spatial_squeeze=True,
scope='vgg_19', scope='vgg_19',
fc_conv_padding='VALID'): fc_conv_padding='VALID',
global_pool=False):
"""Oxford Net VGG 19-Layers version E Example. """Oxford Net VGG 19-Layers version E Example.
Note: All the fully_connected layers have been transformed to conv2d layers. Note: All the fully_connected layers have been transformed to conv2d layers.
...@@ -209,7 +234,8 @@ def vgg_19(inputs, ...@@ -209,7 +234,8 @@ def vgg_19(inputs,
Args: Args:
inputs: a tensor of size [batch_size, height, width, channels]. inputs: a tensor of size [batch_size, height, width, channels].
num_classes: number of predicted classes. num_classes: number of predicted classes. If 0 or None, the logits layer is
omitted and the input features to the logits layer are returned instead.
is_training: whether or not the model is being trained. is_training: whether or not the model is being trained.
dropout_keep_prob: the probability that activations are kept in the dropout dropout_keep_prob: the probability that activations are kept in the dropout
layers during training. layers during training.
...@@ -222,13 +248,18 @@ def vgg_19(inputs, ...@@ -222,13 +248,18 @@ def vgg_19(inputs,
get a prediction map downsampled by a factor of 32 as an output. get a prediction map downsampled by a factor of 32 as an output.
Otherwise, the output prediction map will be (input / 32) - 6 in case of Otherwise, the output prediction map will be (input / 32) - 6 in case of
'VALID' padding. 'VALID' padding.
global_pool: Optional boolean flag. If True, the input to the classification
layer is avgpooled to size 1x1, for any input size. (This is not part
of the original VGG architecture.)
Returns: Returns:
the last op containing the log predictions and end_points dict. net: the output of the logits layer (if num_classes is a non-zero integer),
or the non-dropped-out input to the logits layer (if num_classes is 0 or
None).
end_points: a dict of tensors with intermediate activations.
""" """
with tf.variable_scope(scope, 'vgg_19', [inputs]) as sc: with tf.variable_scope(scope, 'vgg_19', [inputs]) as sc:
end_points_collection = sc.name + '_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],
outputs_collections=end_points_collection): outputs_collections=end_points_collection):
...@@ -242,21 +273,26 @@ def vgg_19(inputs, ...@@ -242,21 +273,26 @@ def vgg_19(inputs,
net = slim.max_pool2d(net, [2, 2], scope='pool4') net = slim.max_pool2d(net, [2, 2], scope='pool4')
net = slim.repeat(net, 4, slim.conv2d, 512, [3, 3], scope='conv5') net = slim.repeat(net, 4, slim.conv2d, 512, [3, 3], scope='conv5')
net = slim.max_pool2d(net, [2, 2], scope='pool5') net = slim.max_pool2d(net, [2, 2], scope='pool5')
# Use conv2d instead of fully_connected layers. # Use conv2d instead of fully_connected layers.
net = slim.conv2d(net, 4096, [7, 7], padding=fc_conv_padding, scope='fc6') net = slim.conv2d(net, 4096, [7, 7], padding=fc_conv_padding, scope='fc6')
net = slim.dropout(net, dropout_keep_prob, is_training=is_training, net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
scope='dropout6') scope='dropout6')
net = slim.conv2d(net, 4096, [1, 1], scope='fc7') net = slim.conv2d(net, 4096, [1, 1], scope='fc7')
net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
scope='dropout7')
net = slim.conv2d(net, num_classes, [1, 1],
activation_fn=None,
normalizer_fn=None,
scope='fc8')
# Convert end_points_collection into a end_point dict. # Convert end_points_collection into a end_point dict.
end_points = slim.utils.convert_collection_to_dict(end_points_collection) end_points = slim.utils.convert_collection_to_dict(end_points_collection)
if spatial_squeeze: if global_pool:
net = tf.squeeze(net, [1, 2], name='fc8/squeezed') net = tf.reduce_mean(net, [1, 2], keep_dims=True, name='global_pool')
end_points['global_pool'] = net
if num_classes:
net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
scope='dropout7')
net = slim.conv2d(net, num_classes, [1, 1],
activation_fn=None,
normalizer_fn=None,
scope='fc8')
if spatial_squeeze:
net = tf.squeeze(net, [1, 2], name='fc8/squeezed')
end_points[sc.name + '/fc8'] = net end_points[sc.name + '/fc8'] = net
return net, end_points return net, end_points
vgg_19.default_image_size = 224 vgg_19.default_image_size = 224
......
research/slim/nets/vgg_test.py
View file @ 57014e4c
...@@ -48,6 +48,18 @@ class VGGATest(tf.test.TestCase): ...@@ -48,6 +48,18 @@ class VGGATest(tf.test.TestCase):
self.assertListEqual(logits.get_shape().as_list(), self.assertListEqual(logits.get_shape().as_list(),
[batch_size, 2, 2, num_classes]) [batch_size, 2, 2, num_classes])
def testGlobalPool(self):
batch_size = 1
height, width = 256, 256
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_a(inputs, num_classes, spatial_squeeze=False,
global_pool=True)
self.assertEquals(logits.op.name, 'vgg_a/fc8/BiasAdd')
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, 1, 1, num_classes])
def testEndPoints(self): def testEndPoints(self):
batch_size = 5 batch_size = 5
height, width = 224, 224 height, width = 224, 224
...@@ -74,6 +86,32 @@ class VGGATest(tf.test.TestCase): ...@@ -74,6 +86,32 @@ class VGGATest(tf.test.TestCase):
] ]
self.assertSetEqual(set(end_points.keys()), set(expected_names)) self.assertSetEqual(set(end_points.keys()), set(expected_names))
def testNoClasses(self):
batch_size = 5
height, width = 224, 224
num_classes = None
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
net, end_points = vgg.vgg_a(inputs, num_classes)
expected_names = ['vgg_a/conv1/conv1_1',
'vgg_a/pool1',
'vgg_a/conv2/conv2_1',
'vgg_a/pool2',
'vgg_a/conv3/conv3_1',
'vgg_a/conv3/conv3_2',
'vgg_a/pool3',
'vgg_a/conv4/conv4_1',
'vgg_a/conv4/conv4_2',
'vgg_a/pool4',
'vgg_a/conv5/conv5_1',
'vgg_a/conv5/conv5_2',
'vgg_a/pool5',
'vgg_a/fc6',
'vgg_a/fc7',
]
self.assertSetEqual(set(end_points.keys()), set(expected_names))
self.assertTrue(net.op.name.startswith('vgg_a/fc7'))
def testModelVariables(self): def testModelVariables(self):
batch_size = 5 batch_size = 5
height, width = 224, 224 height, width = 224, 224
...@@ -177,6 +215,18 @@ class VGG16Test(tf.test.TestCase): ...@@ -177,6 +215,18 @@ class VGG16Test(tf.test.TestCase):
self.assertListEqual(logits.get_shape().as_list(), self.assertListEqual(logits.get_shape().as_list(),
[batch_size, 2, 2, num_classes]) [batch_size, 2, 2, num_classes])
def testGlobalPool(self):
batch_size = 1
height, width = 256, 256
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_16(inputs, num_classes, spatial_squeeze=False,
global_pool=True)
self.assertEquals(logits.op.name, 'vgg_16/fc8/BiasAdd')
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, 1, 1, num_classes])
def testEndPoints(self): def testEndPoints(self):
batch_size = 5 batch_size = 5
height, width = 224, 224 height, width = 224, 224
...@@ -208,6 +258,37 @@ class VGG16Test(tf.test.TestCase): ...@@ -208,6 +258,37 @@ class VGG16Test(tf.test.TestCase):
] ]
self.assertSetEqual(set(end_points.keys()), set(expected_names)) self.assertSetEqual(set(end_points.keys()), set(expected_names))
def testNoClasses(self):
batch_size = 5
height, width = 224, 224
num_classes = None
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
net, end_points = vgg.vgg_16(inputs, num_classes)
expected_names = ['vgg_16/conv1/conv1_1',
'vgg_16/conv1/conv1_2',
'vgg_16/pool1',
'vgg_16/conv2/conv2_1',
'vgg_16/conv2/conv2_2',
'vgg_16/pool2',
'vgg_16/conv3/conv3_1',
'vgg_16/conv3/conv3_2',
'vgg_16/conv3/conv3_3',
'vgg_16/pool3',
'vgg_16/conv4/conv4_1',
'vgg_16/conv4/conv4_2',
'vgg_16/conv4/conv4_3',
'vgg_16/pool4',
'vgg_16/conv5/conv5_1',
'vgg_16/conv5/conv5_2',
'vgg_16/conv5/conv5_3',
'vgg_16/pool5',
'vgg_16/fc6',
'vgg_16/fc7',
]
self.assertSetEqual(set(end_points.keys()), set(expected_names))
self.assertTrue(net.op.name.startswith('vgg_16/fc7'))
def testModelVariables(self): def testModelVariables(self):
batch_size = 5 batch_size = 5
height, width = 224, 224 height, width = 224, 224
...@@ -321,6 +402,18 @@ class VGG19Test(tf.test.TestCase): ...@@ -321,6 +402,18 @@ class VGG19Test(tf.test.TestCase):
self.assertListEqual(logits.get_shape().as_list(), self.assertListEqual(logits.get_shape().as_list(),
[batch_size, 2, 2, num_classes]) [batch_size, 2, 2, num_classes])
def testGlobalPool(self):
batch_size = 1
height, width = 256, 256
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_19(inputs, num_classes, spatial_squeeze=False,
global_pool=True)
self.assertEquals(logits.op.name, 'vgg_19/fc8/BiasAdd')
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, 1, 1, num_classes])
def testEndPoints(self): def testEndPoints(self):
batch_size = 5 batch_size = 5
height, width = 224, 224 height, width = 224, 224
...@@ -356,6 +449,41 @@ class VGG19Test(tf.test.TestCase): ...@@ -356,6 +449,41 @@ class VGG19Test(tf.test.TestCase):
] ]
self.assertSetEqual(set(end_points.keys()), set(expected_names)) self.assertSetEqual(set(end_points.keys()), set(expected_names))
def testNoClasses(self):
batch_size = 5
height, width = 224, 224
num_classes = None
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
net, end_points = vgg.vgg_19(inputs, num_classes)
expected_names = [
'vgg_19/conv1/conv1_1',
'vgg_19/conv1/conv1_2',
'vgg_19/pool1',
'vgg_19/conv2/conv2_1',
'vgg_19/conv2/conv2_2',
'vgg_19/pool2',
'vgg_19/conv3/conv3_1',
'vgg_19/conv3/conv3_2',
'vgg_19/conv3/conv3_3',
'vgg_19/conv3/conv3_4',
'vgg_19/pool3',
'vgg_19/conv4/conv4_1',
'vgg_19/conv4/conv4_2',
'vgg_19/conv4/conv4_3',
'vgg_19/conv4/conv4_4',
'vgg_19/pool4',
'vgg_19/conv5/conv5_1',
'vgg_19/conv5/conv5_2',
'vgg_19/conv5/conv5_3',
'vgg_19/conv5/conv5_4',
'vgg_19/pool5',
'vgg_19/fc6',
'vgg_19/fc7',
]
self.assertSetEqual(set(end_points.keys()), set(expected_names))
self.assertTrue(net.op.name.startswith('vgg_19/fc7'))
def testModelVariables(self): def testModelVariables(self):
batch_size = 5 batch_size = 5
height, width = 224, 224 height, width = 224, 224
......
research/slim/preprocessing/cifarnet_preprocessing.py
View file @ 57014e4c
...@@ -30,7 +30,8 @@ slim = tf.contrib.slim ...@@ -30,7 +30,8 @@ slim = tf.contrib.slim
def preprocess_for_train(image, def preprocess_for_train(image,
output_height, output_height,
output_width, output_width,
padding=_PADDING): padding=_PADDING,
add_image_summaries=True):
"""Preprocesses the given image for training. """Preprocesses the given image for training.
Note that the actual resizing scale is sampled from Note that the actual resizing scale is sampled from
...@@ -41,11 +42,13 @@ def preprocess_for_train(image, ...@@ -41,11 +42,13 @@ def preprocess_for_train(image,
output_height: The height of the image after preprocessing. output_height: The height of the image after preprocessing.
output_width: The width of the image after preprocessing. output_width: The width of the image after preprocessing.
padding: The amound of padding before and after each dimension of the image. padding: The amound of padding before and after each dimension of the image.
add_image_summaries: Enable image summaries.
Returns: Returns:
A preprocessed image. A preprocessed image.
""" """
tf.summary.image('image', tf.expand_dims(image, 0)) if add_image_summaries:
tf.summary.image('image', tf.expand_dims(image, 0))
# Transform the image to floats. # Transform the image to floats.
image = tf.to_float(image) image = tf.to_float(image)
...@@ -58,7 +61,8 @@ def preprocess_for_train(image, ...@@ -58,7 +61,8 @@ def preprocess_for_train(image,
# Randomly flip the image horizontally. # Randomly flip the image horizontally.
distorted_image = tf.image.random_flip_left_right(distorted_image) distorted_image = tf.image.random_flip_left_right(distorted_image)
tf.summary.image('distorted_image', tf.expand_dims(distorted_image, 0)) if add_image_summaries:
tf.summary.image('distorted_image', tf.expand_dims(distorted_image, 0))
# Because these operations are not commutative, consider randomizing # Because these operations are not commutative, consider randomizing
# the order their operation. # the order their operation.
...@@ -70,18 +74,21 @@ def preprocess_for_train(image, ...@@ -70,18 +74,21 @@ def preprocess_for_train(image,
return tf.image.per_image_standardization(distorted_image) return tf.image.per_image_standardization(distorted_image)
def preprocess_for_eval(image, output_height, output_width): def preprocess_for_eval(image, output_height, output_width,
add_image_summaries=True):
"""Preprocesses the given image for evaluation. """Preprocesses the given image for evaluation.
Args: Args:
image: A `Tensor` representing an image of arbitrary size. image: A `Tensor` representing an image of arbitrary size.
output_height: The height of the image after preprocessing. output_height: The height of the image after preprocessing.
output_width: The width of the image after preprocessing. output_width: The width of the image after preprocessing.
add_image_summaries: Enable image summaries.
Returns: Returns:
A preprocessed image. A preprocessed image.
""" """
tf.summary.image('image', tf.expand_dims(image, 0)) if add_image_summaries:
tf.summary.image('image', tf.expand_dims(image, 0))
# Transform the image to floats. # Transform the image to floats.
image = tf.to_float(image) image = tf.to_float(image)
...@@ -89,13 +96,15 @@ def preprocess_for_eval(image, output_height, output_width): ...@@ -89,13 +96,15 @@ def preprocess_for_eval(image, output_height, output_width):
resized_image = tf.image.resize_image_with_crop_or_pad(image, resized_image = tf.image.resize_image_with_crop_or_pad(image,
output_width, output_width,
output_height) output_height)
tf.summary.image('resized_image', tf.expand_dims(resized_image, 0)) if add_image_summaries:
tf.summary.image('resized_image', tf.expand_dims(resized_image, 0))
# Subtract off the mean and divide by the variance of the pixels. # Subtract off the mean and divide by the variance of the pixels.
return tf.image.per_image_standardization(resized_image) return tf.image.per_image_standardization(resized_image)
def preprocess_image(image, output_height, output_width, is_training=False): def preprocess_image(image, output_height, output_width, is_training=False,
add_image_summaries=True):
"""Preprocesses the given image. """Preprocesses the given image.
Args: Args:
...@@ -104,11 +113,16 @@ def preprocess_image(image, output_height, output_width, is_training=False): ...@@ -104,11 +113,16 @@ def preprocess_image(image, output_height, output_width, is_training=False):
output_width: The width of the image after preprocessing. output_width: The width of the image after preprocessing.
is_training: `True` if we're preprocessing the image for training and is_training: `True` if we're preprocessing the image for training and
`False` otherwise. `False` otherwise.
add_image_summaries: Enable image summaries.
Returns: Returns:
A preprocessed image. A preprocessed image.
""" """
if is_training: if is_training:
return preprocess_for_train(image, output_height, output_width) return preprocess_for_train(
image, output_height, output_width,
add_image_summaries=add_image_summaries)
else: else:
return preprocess_for_eval(image, output_height, output_width) return preprocess_for_eval(
image, output_height, output_width,
add_image_summaries=add_image_summaries)
research/slim/preprocessing/inception_preprocessing.py
View file @ 57014e4c
...@@ -155,7 +155,8 @@ def distorted_bounding_box_crop(image, ...@@ -155,7 +155,8 @@ def distorted_bounding_box_crop(image,
def preprocess_for_train(image, height, width, bbox, def preprocess_for_train(image, height, width, bbox,
fast_mode=True, fast_mode=True,
scope=None): scope=None,
add_image_summaries=True):
"""Distort one image for training a network. """Distort one image for training a network.
Distorting images provides a useful technique for augmenting the data Distorting images provides a useful technique for augmenting the data
...@@ -178,6 +179,7 @@ def preprocess_for_train(image, height, width, bbox, ...@@ -178,6 +179,7 @@ def preprocess_for_train(image, height, width, bbox,
fast_mode: Optional boolean, if True avoids slower transformations (i.e. fast_mode: Optional boolean, if True avoids slower transformations (i.e.
bi-cubic resizing, random_hue or random_contrast). bi-cubic resizing, random_hue or random_contrast).
scope: Optional scope for name_scope. scope: Optional scope for name_scope.
add_image_summaries: Enable image summaries.
Returns: Returns:
3-D float Tensor of distorted image used for training with range [-1, 1]. 3-D float Tensor of distorted image used for training with range [-1, 1].
""" """
...@@ -192,7 +194,8 @@ def preprocess_for_train(image, height, width, bbox, ...@@ -192,7 +194,8 @@ def preprocess_for_train(image, height, width, bbox,
# the coordinates are ordered [ymin, xmin, ymax, xmax]. # the coordinates are ordered [ymin, xmin, ymax, xmax].
image_with_box = tf.image.draw_bounding_boxes(tf.expand_dims(image, 0), image_with_box = tf.image.draw_bounding_boxes(tf.expand_dims(image, 0),
bbox) bbox)
tf.summary.image('image_with_bounding_boxes', image_with_box) if add_image_summaries:
tf.summary.image('image_with_bounding_boxes', image_with_box)
distorted_image, distorted_bbox = distorted_bounding_box_crop(image, bbox) distorted_image, distorted_bbox = distorted_bounding_box_crop(image, bbox)
# Restore the shape since the dynamic slice based upon the bbox_size loses # Restore the shape since the dynamic slice based upon the bbox_size loses
...@@ -200,8 +203,9 @@ def preprocess_for_train(image, height, width, bbox, ...@@ -200,8 +203,9 @@ def preprocess_for_train(image, height, width, bbox,
distorted_image.set_shape([None, None, 3]) distorted_image.set_shape([None, None, 3])
image_with_distorted_box = tf.image.draw_bounding_boxes( image_with_distorted_box = tf.image.draw_bounding_boxes(
tf.expand_dims(image, 0), distorted_bbox) tf.expand_dims(image, 0), distorted_bbox)
tf.summary.image('images_with_distorted_bounding_box', if add_image_summaries:
image_with_distorted_box) tf.summary.image('images_with_distorted_bounding_box',
image_with_distorted_box)
# This resizing operation may distort the images because the aspect # This resizing operation may distort the images because the aspect
# ratio is not respected. We select a resize method in a round robin # ratio is not respected. We select a resize method in a round robin
...@@ -215,8 +219,9 @@ def preprocess_for_train(image, height, width, bbox, ...@@ -215,8 +219,9 @@ def preprocess_for_train(image, height, width, bbox,
lambda x, method: tf.image.resize_images(x, [height, width], method), lambda x, method: tf.image.resize_images(x, [height, width], method),
num_cases=num_resize_cases) num_cases=num_resize_cases)
tf.summary.image('cropped_resized_image', if add_image_summaries:
tf.expand_dims(distorted_image, 0)) tf.summary.image('cropped_resized_image',
tf.expand_dims(distorted_image, 0))
# Randomly flip the image horizontally. # Randomly flip the image horizontally.
distorted_image = tf.image.random_flip_left_right(distorted_image) distorted_image = tf.image.random_flip_left_right(distorted_image)
...@@ -227,8 +232,9 @@ def preprocess_for_train(image, height, width, bbox, ...@@ -227,8 +232,9 @@ def preprocess_for_train(image, height, width, bbox,
lambda x, ordering: distort_color(x, ordering, fast_mode), lambda x, ordering: distort_color(x, ordering, fast_mode),
num_cases=4) num_cases=4)
tf.summary.image('final_distorted_image', if add_image_summaries:
tf.expand_dims(distorted_image, 0)) tf.summary.image('final_distorted_image',
tf.expand_dims(distorted_image, 0))
distorted_image = tf.subtract(distorted_image, 0.5) distorted_image = tf.subtract(distorted_image, 0.5)
distorted_image = tf.multiply(distorted_image, 2.0) distorted_image = tf.multiply(distorted_image, 2.0)
return distorted_image return distorted_image
...@@ -278,7 +284,8 @@ def preprocess_for_eval(image, height, width, ...@@ -278,7 +284,8 @@ def preprocess_for_eval(image, height, width,
def preprocess_image(image, height, width, def preprocess_image(image, height, width,
is_training=False, is_training=False,
bbox=None, bbox=None,
fast_mode=True): fast_mode=True,
add_image_summaries=True):
"""Pre-process one image for training or evaluation. """Pre-process one image for training or evaluation.
Args: Args:
...@@ -295,6 +302,7 @@ def preprocess_image(image, height, width, ...@@ -295,6 +302,7 @@ def preprocess_image(image, height, width,
where each coordinate is [0, 1) and the coordinates are arranged as where each coordinate is [0, 1) and the coordinates are arranged as
[ymin, xmin, ymax, xmax]. [ymin, xmin, ymax, xmax].
fast_mode: Optional boolean, if True avoids slower transformations. fast_mode: Optional boolean, if True avoids slower transformations.
add_image_summaries: Enable image summaries.
Returns: Returns:
3-D float Tensor containing an appropriately scaled image 3-D float Tensor containing an appropriately scaled image
...@@ -303,6 +311,7 @@ def preprocess_image(image, height, width, ...@@ -303,6 +311,7 @@ def preprocess_image(image, height, width,
ValueError: if user does not provide bounding box ValueError: if user does not provide bounding box
""" """
if is_training: if is_training:
return preprocess_for_train(image, height, width, bbox, fast_mode) return preprocess_for_train(image, height, width, bbox, fast_mode,
add_image_summaries=add_image_summaries)
else: else:
return preprocess_for_eval(image, height, width) return preprocess_for_eval(image, height, width)
research/slim/preprocessing/preprocessing_factory.py
View file @ 57014e4c
...@@ -54,6 +54,8 @@ def get_preprocessing(name, is_training=False): ...@@ -54,6 +54,8 @@ def get_preprocessing(name, is_training=False):
'inception_resnet_v2': inception_preprocessing, 'inception_resnet_v2': inception_preprocessing,
'lenet': lenet_preprocessing, 'lenet': lenet_preprocessing,
'mobilenet_v1': inception_preprocessing, 'mobilenet_v1': inception_preprocessing,
'nasnet_mobile': inception_preprocessing,
'nasnet_large': inception_preprocessing,
'resnet_v1_50': vgg_preprocessing, 'resnet_v1_50': vgg_preprocessing,
'resnet_v1_101': vgg_preprocessing, 'resnet_v1_101': vgg_preprocessing,
'resnet_v1_152': vgg_preprocessing, 'resnet_v1_152': vgg_preprocessing,
......
research/slim/scripts/export_mobilenet.sh
View file @ 57014e4c
#!/bin/bash #!/bin/bash
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# This script prepares the various different versions of MobileNet models for # This script prepares the various different versions of MobileNet models for
# use in a mobile application. If you don't specify your own trained checkpoint # use in a mobile application. If you don't specify your own trained checkpoint
# file, it will download pretrained checkpoints for ImageNet. You'll also need # file, it will download pretrained checkpoints for ImageNet. You'll also need
......
research/slim/scripts/finetune_inception_resnet_v2_on_flowers.sh
View file @ 57014e4c
#!/bin/bash #!/bin/bash
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# #
# This script performs the following operations: # This script performs the following operations:
# 1. Downloads the Flowers dataset # 1. Downloads the Flowers dataset
......
research/slim/scripts/finetune_inception_v1_on_flowers.sh
View file @ 57014e4c
#!/bin/bash #!/bin/bash
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# #
# This script performs the following operations: # This script performs the following operations:
# 1. Downloads the Flowers dataset # 1. Downloads the Flowers dataset
......
research/slim/scripts/finetune_inception_v3_on_flowers.sh
View file @ 57014e4c
#!/bin/bash #!/bin/bash
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# #
# This script performs the following operations: # This script performs the following operations:
# 1. Downloads the Flowers dataset # 1. Downloads the Flowers dataset
......
research/slim/scripts/finetune_resnet_v1_50_on_flowers.sh
View file @ 57014e4c
#!/bin/bash #!/bin/bash
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# #
# This script performs the following operations: # This script performs the following operations:
# 1. Downloads the Flowers dataset # 1. Downloads the Flowers dataset
......
research/slim/scripts/train_cifarnet_on_cifar10.sh
View file @ 57014e4c
#!/bin/bash #!/bin/bash
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# #
# This script performs the following operations: # This script performs the following operations:
# 1. Downloads the Cifar10 dataset # 1. Downloads the Cifar10 dataset
......
research/slim/scripts/train_lenet_on_mnist.sh
View file @ 57014e4c
#!/bin/bash #!/bin/bash
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# #
# This script performs the following operations: # This script performs the following operations:
# 1. Downloads the MNIST dataset # 1. Downloads the MNIST dataset
......
research/slim/setup.py
View file @ 57014e4c
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Setup script for slim.""" """Setup script for slim."""
from setuptools import find_packages from setuptools import find_packages
......
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