Skip to content
GitLab
Commit 2ca17f5f authored by Tiago de Freitas Pereira's avatar Tiago de Freitas Pereira
Browse files

Fixed the reuse in some operations of the inception-v1

parent 30dd85ea
Pipeline #29815 passed with stage
in 308 minutes and 2 seconds
Hide whitespace changes
Inline Side-by-side
bob/learn/tensorflow/network/InceptionResnetV1.py
View file @ 2ca17f5f
......@@ -32,24 +32,29 @@ def block35(net,
scale=1.0,
activation_fn=tf.nn.relu,
scope=None,
reuse=None,
trainable_variables=True):
"""Builds the 35x35 resnet block."""
with tf.variable_scope(scope, 'Block35', [net]):
with tf.variable_scope('Branch_0'):
tower_conv = slim.conv2d(
net, 32, 1, scope='Conv2d_1x1', trainable=trainable_variables)
net, 32, 1, scope='Conv2d_1x1',
reuse=reuse,
trainable=trainable_variables)
with tf.variable_scope('Branch_1'):
tower_conv1_0 = slim.conv2d(
net,
32,
1,
scope='Conv2d_0a_1x1',
reuse=reuse,
trainable=trainable_variables)
tower_conv1_1 = slim.conv2d(
tower_conv1_0,
32,
3,
scope='Conv2d_0b_3x3',
reuse=reuse,
trainable=trainable_variables)
with tf.variable_scope('Branch_2'):
tower_conv2_0 = slim.conv2d(
......@@ -57,18 +62,21 @@ def block35(net,
32,
1,
scope='Conv2d_0a_1x1',
reuse=reuse,
trainable=trainable_variables)
tower_conv2_1 = slim.conv2d(
tower_conv2_0,
32,
3,
scope='Conv2d_0b_3x3',
reuse=reuse,
trainable=trainable_variables)
tower_conv2_2 = slim.conv2d(
tower_conv2_1,
32,
3,
scope='Conv2d_0c_3x3',
reuse=reuse,
trainable=trainable_variables)
mixed = tf.concat([tower_conv, tower_conv1_1, tower_conv2_2], 3)
up = slim.conv2d(
......@@ -78,6 +86,7 @@ def block35(net,
normalizer_fn=None,
activation_fn=None,
scope='Conv2d_1x1',
reuse=reuse,
trainable=trainable_variables)
net += scale * up
if activation_fn:
......@@ -90,28 +99,32 @@ def block17(net,
scale=1.0,
activation_fn=tf.nn.relu,
scope=None,
reuse=None,
trainable_variables=True):
"""Builds the 17x17 resnet block."""
with tf.variable_scope(scope, 'Block17', [net]):
with tf.variable_scope('Branch_0'):
tower_conv = slim.conv2d(
net, 128, 1, scope='Conv2d_1x1', trainable=trainable_variables)
net, 128, 1, scope='Conv2d_1x1', trainable=trainable_variables, reuse=reuse)
with tf.variable_scope('Branch_1'):
tower_conv1_0 = slim.conv2d(
net,
128,
1,
scope='Conv2d_0a_1x1',
reuse=reuse,
trainable=trainable_variables)
tower_conv1_1 = slim.conv2d(
tower_conv1_0,
128, [1, 7],
scope='Conv2d_0b_1x7',
reuse=reuse,
trainable=trainable_variables)
tower_conv1_2 = slim.conv2d(
tower_conv1_1,
128, [7, 1],
scope='Conv2d_0c_7x1',
reuse=reuse,
trainable=trainable_variables)
mixed = tf.concat([tower_conv, tower_conv1_2], 3)
up = slim.conv2d(
......@@ -121,6 +134,7 @@ def block17(net,
normalizer_fn=None,
activation_fn=None,
scope='Conv2d_1x1',
reuse=reuse,
trainable=trainable_variables)
net += scale * up
if activation_fn:
......@@ -133,28 +147,32 @@ def block8(net,
scale=1.0,
activation_fn=tf.nn.relu,
scope=None,
reuse=None,
trainable_variables=True):
"""Builds the 8x8 resnet block."""
with tf.variable_scope(scope, 'Block8', [net]):
with tf.variable_scope('Branch_0'):
tower_conv = slim.conv2d(
net, 192, 1, scope='Conv2d_1x1', trainable=trainable_variables)
net, 192, 1, scope='Conv2d_1x1', trainable=trainable_variables,reuse=reuse)
with tf.variable_scope('Branch_1'):
tower_conv1_0 = slim.conv2d(
net,
192,
1,
scope='Conv2d_0a_1x1',
reuse=reuse,
trainable=trainable_variables)
tower_conv1_1 = slim.conv2d(
tower_conv1_0,
192, [1, 3],
scope='Conv2d_0b_1x3',
reuse=reuse,
trainable=trainable_variables)
tower_conv1_2 = slim.conv2d(
tower_conv1_1,
192, [3, 1],
scope='Conv2d_0c_3x1',
reuse=reuse,
trainable=trainable_variables)
mixed = tf.concat([tower_conv, tower_conv1_2], 3)
up = slim.conv2d(
......@@ -164,6 +182,7 @@ def block8(net,
normalizer_fn=None,
activation_fn=None,
scope='Conv2d_1x1',
reuse=reuse,
trainable=trainable_variables)
net += scale * up
if activation_fn:
......@@ -171,7 +190,7 @@ def block8(net,
return net
def reduction_a(net, k, l, m, n, trainable_variables=True):
def reduction_a(net, k, l, m, n, trainable_variables=True, reuse=None):
with tf.variable_scope('Branch_0'):
tower_conv = slim.conv2d(
net,
......@@ -180,15 +199,17 @@ def reduction_a(net, k, l, m, n, trainable_variables=True):
stride=2,
padding='VALID',
scope='Conv2d_1a_3x3',
reuse=reuse,
trainable=trainable_variables)
with tf.variable_scope('Branch_1'):
tower_conv1_0 = slim.conv2d(
net, k, 1, scope='Conv2d_0a_1x1', trainable=trainable_variables)
net, k, 1, scope='Conv2d_0a_1x1', trainable=trainable_variables, reuse=reuse)
tower_conv1_1 = slim.conv2d(
tower_conv1_0,
l,
3,
scope='Conv2d_0b_3x3',
reuse=reuse,
trainable=trainable_variables)
tower_conv1_2 = slim.conv2d(
tower_conv1_1,
......@@ -196,6 +217,7 @@ def reduction_a(net, k, l, m, n, trainable_variables=True):
3,
stride=2,
padding='VALID',
reuse=reuse,
scope='Conv2d_1a_3x3',
trainable=trainable_variables)
with tf.variable_scope('Branch_2'):
......@@ -205,21 +227,22 @@ def reduction_a(net, k, l, m, n, trainable_variables=True):
return net
def reduction_b(net, trainable_variables=True):
def reduction_b(net, trainable_variables=True, reuse=None):
with tf.variable_scope('Branch_0'):
tower_conv = slim.conv2d(
net, 256, 1, scope='Conv2d_0a_1x1', trainable=trainable_variables)
net, 256, 1, scope='Conv2d_0a_1x1', trainable=trainable_variables, reuse=reuse)
tower_conv_1 = slim.conv2d(
tower_conv,
384,
3,
stride=2,
padding='VALID',
reuse=reuse,
scope='Conv2d_1a_3x3',
trainable=trainable_variables)
with tf.variable_scope('Branch_1'):
tower_conv1 = slim.conv2d(
net, 256, 1, scope='Conv2d_0a_1x1', trainable=trainable_variables)
net, 256, 1, scope='Conv2d_0a_1x1', trainable=trainable_variables, reuse=reuse)
tower_conv1_1 = slim.conv2d(
tower_conv1,
256,
......@@ -227,15 +250,17 @@ def reduction_b(net, trainable_variables=True):
stride=2,
padding='VALID',
scope='Conv2d_1a_3x3',
reuse=reuse,
trainable=trainable_variables)
with tf.variable_scope('Branch_2'):
tower_conv2 = slim.conv2d(
net, 256, 1, scope='Conv2d_0a_1x1', trainable=trainable_variables)
net, 256, 1, scope='Conv2d_0a_1x1', trainable=trainable_variables, reuse=reuse)
tower_conv2_1 = slim.conv2d(
tower_conv2,
256,
3,
3,
scope='Conv2d_0b_3x3',
reuse=reuse,
trainable=trainable_variables)
tower_conv2_2 = slim.conv2d(
tower_conv2_1,
......@@ -244,6 +269,7 @@ def reduction_b(net, trainable_variables=True):
stride=2,
padding='VALID',
scope='Conv2d_1a_3x3',
reuse=reuse,
trainable=trainable_variables)
with tf.variable_scope('Branch_3'):
tower_pool = slim.max_pool2d(
......@@ -400,6 +426,7 @@ def inception_resnet_v1(inputs,
3,
stride=2,
padding='VALID',
reuse=reuse,
scope=name,
trainable=trainable)
end_points[name] = net
......@@ -419,6 +446,7 @@ def inception_resnet_v1(inputs,
3,
padding='VALID',
scope=name,
reuse=reuse,
trainable=trainable)
end_points[name] = net
......@@ -433,7 +461,7 @@ def inception_resnet_v1(inputs,
name = "Conv2d_2b_3x3"
trainable = is_trainable(name, trainable_variables, mode=mode)
net = slim.conv2d(
net, 64, 3, scope=name, trainable=trainable)
net, 64, 3, scope=name, trainable=trainable, reuse=reuse)
end_points[name] = net
# 73 x 73 x 64
......@@ -457,6 +485,7 @@ def inception_resnet_v1(inputs,
1,
padding='VALID',
scope=name,
reuse=reuse,
trainable=trainable)
end_points[name] = net
......@@ -476,6 +505,7 @@ def inception_resnet_v1(inputs,
3,
padding='VALID',
scope=name,
reuse=reuse,
trainable=trainable)
end_points[name] = net
......@@ -496,6 +526,7 @@ def inception_resnet_v1(inputs,
stride=2,
padding='VALID',
scope=name,
reuse=reuse,
trainable=trainable)
end_points[name] = net
......@@ -514,6 +545,7 @@ def inception_resnet_v1(inputs,
5,
block35,
scale=0.17,
reuse=reuse,
trainable_variables=trainable)
end_points[name] = net
......@@ -568,7 +600,7 @@ def inception_resnet_v1(inputs,
with tf.variable_scope(name):
net = reduction_b(
net, trainable_variables=trainable)
net, trainable_variables=trainable, reuse=reuse)
end_points[name] = net
# 5 x Inception-Resnet-C
......
Supports Markdown
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