Tensorflow Slim

Posted on | In , |

[TOC]

1.变量的定义

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import tensorflow as tf

slim = tf.contrib.slim


#模型变量
weights = slim.model_variable('weights',
                              shape=[1, 1, 3 , 3],
                              initializer=tf.truncated_normal_initializer(stddev=0.1),
                              regularizer=slim.l2_regularizer(0.05),
                              device='/GPU:0')

# 局部变量
my_var = slim.variable('my_var',
                       shape=[10, 1],
                       initializer=tf.zeros_initializer())

#get_variables 返回所有的变量 
regular_variables_and_model_variables = slim.get_variables()

print()


with tf.Session() as sess:

    sess.run(tf.global_variables_initializer())
    print(sess.run(weights))
    print(sess.run(my_var))
    print(sess.run(regular_variables_and_model_variables))

#而模型变量会再save的时候保存下来。 诸如global_step之类的就是局部变量。
#slim中可以写明变量存放的设备，正则和初始化规则。还有获取变量的函数也需要注意一下，get_variables是返回所有的变量。

2.卷积的操作

2.1 传统的卷积

1
2
3
4
5
6
7
8
9
10
#传统的卷积
input = ...
with tf.name_scope('conv1_1') as scope:
  kernel = tf.Variable(tf.truncated_normal([3, 3, 64, 128], dtype=tf.float32,
                                           stddev=1e-1), name='weights')
  conv = tf.nn.conv2d(input, kernel, [1, 1, 1, 1], padding='SAME')
  biases = tf.Variable(tf.constant(0.0, shape=[128], dtype=tf.float32),
                       trainable=True, name='biases')
  bias = tf.nn.bias_add(conv, biases)
  conv1 = tf.nn.relu(bias, name=scope)

2.2 slim的卷积

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import tensorflow as tf
slim = tf.contrib.slim

#slim实现卷积
net = slim.conv2d(input, 128, [3, 3], scope='conv1_1')
'''
    底层代码
    @staticmethod
    def conv2d(features, weight):
        """Produces a convolutional layer that filters an image subregion
        :param features: The layer input.
        :param weight: The size of the layer filter.
        :return: Returns a convolutional layer.
        """
        return tf.nn.conv2d(features, weight, strides=[1, 1, 1, 1], padding='SAME')
'''

2.3 slim定义相同层

  • repeat操作 减少代码量
  • stack是处理卷积核或者输出不一样的情况
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function


import tensorflow as tf

slim = tf.contrib.slim

'''
repeat操作  减少代码量
stack是处理卷积核或者输出不一样的情况
'''

#1.
'''
假设定义三个相同的卷积层：
input = slim.conv2d(input, 256, [3, 3], scope='conv3_1')
input = slim.conv2d(input, 256, [3, 3], scope='conv3_2')
input = slim.conv2d(input, 256, [3, 3], scope='conv3_3')
input = slim.max_pool2d(input, [2, 2], scope='pool2')
'''
#在slim中的repeat操作可以减少代码量：
net = slim.repeat(input, 3, slim.conv2d, 256, [3, 3], scope='conv3')
net = slim.max_pool2d(input, [2, 2], scope='pool2')



#2.
'''
    假设定义三层FC：
        # Verbose way:
        x = slim.fully_connected(x, 32, scope='fc/fc_1')
        x = slim.fully_connected(x, 64, scope='fc/fc_2')
        x = slim.fully_connected(x, 128, scope='fc/fc_3')
'''
#使用stack操作：
slim.stack(x, slim.fully_connected, [32, 64, 128], scope='fc')


#3.
# 普通方法:
x = slim.conv2d(x, 32, [3, 3], scope='core/core_1')
x = slim.conv2d(x, 32, [1, 1], scope='core/core_2')
x = slim.conv2d(x, 64, [3, 3], scope='core/core_3')
x = slim.conv2d(x, 64, [1, 1], scope='core/core_4')

# 简便方法:
slim.stack(x, slim.conv2d, [(32, [3, 3]), (32, [1, 1]), (64, [3, 3]), (64, [1, 1])], scope='core')

2.4 slim中的argscope

​ 如果你的网络有大量相同的参数，如下：

  • arg_scope操作 用scope提取相同的特征
  • 嵌套的使用 和多层的定义
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import tensorflow as tf

slim = tf.contrib.slim

'''
      argscope 定义参数，不重复的定义
'''


net = slim.conv2d(inputs, 64, [11, 11], 4, padding='SAME',
                  weights_initializer=tf.truncated_normal_initializer(stddev=0.01),
                  weights_regularizer=slim.l2_regularizer(0.0005), scope='conv1')
net = slim.conv2d(net, 128, [11, 11], padding='VALID',
                  weights_initializer=tf.truncated_normal_initializer(stddev=0.01),
                  weights_regularizer=slim.l2_regularizer(0.0005), scope='conv2')
net = slim.conv2d(net, 256, [11, 11], padding='SAME',
                  weights_initializer=tf.truncated_normal_initializer(stddev=0.01),
                  weights_regularizer=slim.l2_regularizer(0.0005), scope='conv3')

#arg_scope操作   用scope提取相同的特征
with slim.arg_scope([slim.conv2d], padding='SAME',
                      weights_initializer=tf.truncated_normal_initializer(stddev=0.01)
                      weights_regularizer=slim.l2_regularizer(0.0005)):
    net = slim.conv2d(inputs, 64, [11, 11], scope='conv1')
    net = slim.conv2d(net, 128, [11, 11], padding='VALID', scope='conv2')
    net = slim.conv2d(net, 256, [11, 11], scope='conv3')

#嵌套的使用  和多层的定义
with slim.arg_scope([slim.conv2d, slim.fully_connected],
                      activation_fn=tf.nn.relu,
                      weights_initializer=tf.truncated_normal_initializer(stddev=0.01),
                      weights_regularizer=slim.l2_regularizer(0.0005)):
  with slim.arg_scope([slim.conv2d], stride=1, padding='SAME'):
    net = slim.conv2d(inputs, 64, [11, 11], 4, padding='VALID', scope='conv1')
    net = slim.conv2d(net, 256, [5, 5],
                      weights_initializer=tf.truncated_normal_initializer(stddev=0.03),
                      scope='conv2')
    net = slim.fully_connected(net, 1000, activation_fn=None, scope='fc')



#定义VGG网络
def vgg16(inputs):
  with slim.arg_scope([slim.conv2d, slim.fully_connected],
                      activation_fn=tf.nn.relu,
                      weights_initializer=tf.truncated_normal_initializer(0.0, 0.01),
                      weights_regularizer=slim.l2_regularizer(0.0005)):
    net = slim.repeat(inputs, 2, slim.conv2d, 64, [3, 3], scope='conv1') # 定义两个conv  卷积核为3*2
    net = slim.max_pool2d(net, [2, 2], scope='pool1')  #定义池化层  2*2
    net = slim.repeat(net, 2, slim.conv2d, 128, [3, 3], scope='conv2')
    net = slim.max_pool2d(net, [2, 2], scope='pool2')
    net = slim.repeat(net, 3, slim.conv2d, 256, [3, 3], scope='conv3')
    net = slim.max_pool2d(net, [2, 2], scope='pool3')
    net = slim.repeat(net, 3, slim.conv2d, 512, [3, 3], scope='conv4')
    net = slim.max_pool2d(net, [2, 2], scope='pool4')
    net = slim.repeat(net, 3, slim.conv2d, 512, [3, 3], scope='conv5')
    net = slim.max_pool2d(net, [2, 2], scope='pool5')
    net = slim.fully_connected(net, 4096, scope='fc6')  #全连接网络
    net = slim.dropout(net, 0.5, scope='dropout6')
    net = slim.fully_connected(net, 4096, scope='fc7')
    net = slim.dropout(net, 0.5, scope='dropout7')  #dropout
    net = slim.fully_connected(net, 1000, activation_fn=None, scope='fc8')
  return net

3. slim封装网络

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
from tensorflow.contrib.slim.python.slim.nets import alexnet
from tensorflow.contrib.slim.python.slim.nets import inception
from tensorflow.contrib.slim.python.slim.nets import overfeat
from tensorflow.contrib.slim.python.slim.nets import resnet_utils
from tensorflow.contrib.slim.python.slim.nets import resnet_v1
from tensorflow.contrib.slim.python.slim.nets import resnet_v2
from tensorflow.contrib.slim.python.slim.nets import vgg
from tensorflow.python.util.all_util import make_all

import tensorflow as tf

vgg = tf.contrib.slim.nets.vgg

# Load the images and labels.
images, labels = ...

# Create the model.
predictions, _ = vgg.vgg_16(images)

# Define the loss functions and get the total loss.
loss = slim.losses.softmax_cross_entropy(predictions, labels)

4. loss

损失函数定义了我们想要最小化的数量。 对于分类问题，这通常是跨分类的真实分布和预测概率分布之间的交叉熵。 对于回归问题，这通常是预测值和真值之间的平方和差异。

某些模型（如多任务学习模型）需要同时使用多个损失函数。 换句话说，最终被最小化的损失函数是各种其他损失函数的总和。 例如，考虑预测图像中的场景类型以及每个像素的相机深度的模型。 这个模型的损失函数将是分类损失和深度预测损失的总和。

TF-Slim提供了一个易于使用的机制，通过损失模块定义和跟踪损失功能。 考虑一下我们想要训练VGG网络的简单情况：

  • classification_loss = slim.losses.softmax_cross_entropy(scene_predictions, scene_labels)

  • sum_of_squares_loss = slim.losses.sum_of_squares(depth_predictions, depth_labels)

  • regularization_loss = slim.losses.get_regularization_losses()

slim.losses.get_total_loss(add_regularization_losses=False)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
import tensorflow as tf
vgg = tf.contrib.slim.nets.vgg

# Load the images and labels.
images, labels = ...

# Create the model.
predictions, _ = vgg.vgg_16(images)

# Define the loss functions and get the total loss.
loss = slim.losses.softmax_cross_entropy(predictions, labels)
在这个例子中，我们首先创建模型（使用TF-Slim的VGG实现），并添加标准分类损失。 现在，让我们假设有一个多任务模型，产生多个输出的情况：

# Load the images and labels.
images, scene_labels, depth_labels = ...

# Create the model.
scene_predictions, depth_predictions = CreateMultiTaskModel(images)

# Define the loss functions and get the total loss.
classification_loss = slim.losses.softmax_cross_entropy(scene_predictions, scene_labels)
sum_of_squares_loss = slim.losses.sum_of_squares(depth_predictions, depth_labels)

# The following two lines have the same effect:
total_loss = classification_loss + sum_of_squares_loss
total_loss = slim.losses.get_total_loss(add_regularization_losses=False)

在这个例子中，我们有两个损失，我们通过调用slim.losses.softmax_cross_entropy和slim.losses.sum_of_squares来添加。 我们可以通过将它们相加（total_loss）或调用slim.losses.get_total_loss（）来获得全部损失。 这是如何工作的？ 当您通过TF-Slim创建loss function时，TF-Slim将损失添加到损失函数中特殊的TensorFlow集合中。 这使您可以手动管理全部损失，或允许TF-Slim为您管理它们。


如果你想让TF-Slim管理你的损失，通过一个自定义的损失函数呢？ loss_ops.py也有一个功能，把这个损失添加到TF-Slims集合中。 例如：

# Load the images and labels.
images, scene_labels, depth_labels, pose_labels = ...

# Create the model.
scene_predictions, depth_predictions, pose_predictions = CreateMultiTaskModel(images)

# Define the loss functions and get the total loss.
classification_loss = slim.losses.softmax_cross_entropy(scene_predictions, scene_labels)
sum_of_squares_loss = slim.losses.sum_of_squares(depth_predictions, depth_labels)
pose_loss = MyCustomLossFunction(pose_predictions, pose_labels)
slim.losses.add_loss(pose_loss) # Letting TF-Slim know about the additional loss.

# The following two ways to compute the total loss are equivalent: 【regularization_loss】
regularization_loss = tf.add_n(slim.losses.get_regularization_losses())
total_loss1 = classification_loss + sum_of_squares_loss + pose_loss + regularization_loss

# (Regularization Loss is included in the total loss by default).
total_loss2 = slim.losses.get_total_loss()
在这个例子中，我们可以再次手动产生总损失函数，或者让TF-Slim知道额外的损失，让TF-Slim处理损失。

5. 保存读取模型

通过以下功能我们可以载入模型的部分变量：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
# Create some variables.
v1 = slim.variable(name="v1", ...)
v2 = slim.variable(name="nested/v2", ...)
...

# Get list of variables to restore (which contains only 'v2').
variables_to_restore = slim.get_variables_by_name("v2")

# Create the saver which will be used to restore the variables.
restorer = tf.train.Saver(variables_to_restore)

with tf.Session() as sess:
  # Restore variables from disk.
  restorer.restore(sess, "/tmp/model.ckpt")
  print("Model restored.")
除了这种部分变量加载的方法外，我们甚至还能加载到不同名字的变量中。

假设我们定义的网络变量是conv1/weights，而从VGG加载的变量名为vgg16/conv1/weights，正常load肯定会报错（找不到变量名），但是可以这样：

def name_in_checkpoint(var):
  return 'vgg16/' + var.op.name

variables_to_restore = slim.get_model_variables()
variables_to_restore = {name_in_checkpoint(var):var for var in variables_to_restore}
restorer = tf.train.Saver(variables_to_restore)

with tf.Session() as sess:
  # Restore variables from disk.
  restorer.restore(sess, "/tmp/model.ckpt")

通过这种方式我们可以加载不同变量名的变量！！

————————————————
版权声明：本文为CSDN博主「醉小义」的原创文章，遵循CC 4.0 BY-SA版权协议，转载请附上原文出处链接及本声明。
原文链接：https://blog.csdn.net/qq_30638831/article/details/81389533