tensorflow layers

发表于 | 更新于 | 分类于 | 评论数：

tf.layers

这个模块定义在tf.contrib.layers中。主要是构建神经网络，正则化和summaries等op。它包括1个模块，19个类，以及一系列函数。

模块

experimental module

tf.layers.experimental的公开的API

类

class Conv2D

二维卷积类。

API

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
__init__(
    filters, # 卷积核的数量
    kernel_size, # 卷积核的大小
    strides=(1, 1),
    padding='valid',
    data_format='channels_last', # string, "channels_last", "channels_first"
    dilation_rate=(1, 1), #
    activation=None, # 激活函数
    use_bias=True,
    kernel_initializer=None, # 卷积核的构造器
    bias_initializer=tf.zeros_initializer(), # bias的构造器
    kernel_regularizer=None, #  卷积核的正则化
    bias_regularizer=None,
    activity_regularizer=None,
    kernel_constraint=None,
    bias_constraint=None,
    trainable=True, # 如果为True的话，将变量添加到TRANABLE_VARIABELS collection中
    name=None,
    **kwargs
)

示例

其他

所有类

  • class AveragePooling1D
  • class AveragePooling2D
  • class AveragePooling3D
  • class BatchNormalization
  • class Conv1D
  • class Conv2D
  • class Conv2DTranspose
  • class Conv3D
  • class Conv3DTranspose
  • class Dense
  • class Dropout
  • class Flatten
  • class InputSpec
  • class Layer
  • class MaxPooling1D
  • class MaxPooling2D
  • class MaxPooling3D
  • class SeparableConv1D
  • class SeparableConv2D

函数

conv2d

API

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
tf.layers.conv2d(
    inputs, # 输入
    filters, #  一个整数,输出的维度，就是有几个卷积核
    kernel_size,
    strides=(1, 1),
    padding='valid',
    data_format='channels_last',
    dilation_rate=(1, 1),
    activation=None,
    use_bias=True,
    kernel_initializer=None,
    bias_initializer=tf.zeros_initializer(),
    kernel_regularizer=None,
    bias_regularizer=None,
    activity_regularizer=None,
    kernel_constraint=None,
    bias_constraint=None,
    trainable=True,
    name=None,
    reuse=None
)

示例

其他

所有函数

需要注意的是，下列所有函数在以后版本都将被弃用。

  • average_pooling1d(…)
  • average_pooling2d(…)
  • average_pooling3d(…)
  • batch_normalization(…)
  • conv1d(…)
  • conv2d(…)
  • conv2d_transpose(…)
  • conv3d(…)
  • conv3d_transpose(…)
  • dense(…)
  • dropout(…)
  • flatten(…)
  • max_pooling1d(…)
  • max_pooling2d(…)
  • max_pooling3d(…)
  • separable_conv1d(…)
  • separable_conv2d(…)

tf.layers.conv2d vs tf.layers.Conv2d

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
tf.layers.Conv2d.__init__(
    filters,
    kernel_size,
    strides=(1, 1),
    padding='valid',
    data_format='channels_last',
    dilation_rate=(1, 1),
    activation=None,
    use_bias=True,
    kernel_initializer=None,
    bias_initializer=tf.zeros_initializer(),
    kernel_regularizer=None,
    bias_regularizer=None,
    activity_regularizer=None,
    kernel_constraint=None,
    bias_constraint=None,
    trainable=True,
    name=None,
    **kwargs
)
tf.layers.conv2d(
    inputs,
    filters,
    kernel_size,
    strides=(1, 1),
    padding='valid',
    data_format='channels_last',
    dilation_rate=(1, 1),
    activation=None,
    use_bias=True,
    kernel_initializer=None,
    bias_initializer=tf.zeros_initializer(),
    kernel_regularizer=None,
    bias_regularizer=None,
    activity_regularizer=None,
    kernel_constraint=None,
    bias_constraint=None,
    trainable=True,
    name=None,
    reuse=None
)

conv2d是函数；Conv2d是类。
conv2d运行的时候需要传入卷积核参数，输入；Conv2d在构造的时候需要实例化卷积核参数，实例化后，可以使用不用的输入得到不同的输出。
调用conv2d就相当于调用Conv2d对象的apply(inputs)函数。

参考文献

1.https://www.tensorflow.org/api_docs/python/tf/layers
4.https://www.tensorflow.org/api_docs/python/tf/layers/Conv2D
5.https://www.tensorflow.org/api_docs/python/tf/layers/conv2d
6.https://stackoverflow.com/questions/52011509/what-is-difference-between-tf-layers-conv2d-and-tf-layers-conv2d/52035621