TensorFlow 之 keras.layers.Conv2D( ) 主要参数讲解
keras.layers.Conv2D( ) 函数参数
def __init__(self, filters,
kernel_size,
strides=(1, 1),
padding='valid',
data_format=None,
dilation_rate=(1, 1),
activation=None,
use_bias=True,
kernel_initializer='glorot_uniform',
bias_initializer='zeros',
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
**kwargs):
参数：
filters 卷积核个数的变化，filters 影响的是最后输入结果的的第三个维度的变化，例如，输入的维度是 (600, 600, 3), filters 的个数是 64，转变后的维度是 (600, 600, 64)
>>> from keras.layers import (Input, Reshape)
>>> input = Input(shape=(600, 600, 3))
>>> x = Conv2D(64, (1, 1), strides=(1, 1), name='conv1')(input)
>>> x
<tf.Tensor 'conv1_1/BiasAdd:0' shape=(?, 600, 600, 64) dtype=float32>
kernel_size 参数 表示卷积核的大小，可以直接写一个数，影响的是输出结果前两个数据的维度，例如，（600, 600, 3）=> (599, 599, 64)
>>> from keras.layers import (Input, Conv2D)
>>> input = Input(shape=(600, 600, 3))
>>> Conv2D(64, (2, 2), strides=(1, 1), name='conv1')(input)
<tf.Tensor 'conv1/BiasAdd:0' shape=(?, 599, 599, 64) dtype=float32>
直接写 2 也是可以的
>>> from keras.layers import (Input, Conv2D)
>>> input = Input(shape=(600, 600, 3))
>>> Conv2D(64, 2, strides=(1, 1), name='conv1')(input)
<tf.Tensor 'conv1_2/BiasAdd:0' shape=(?, 599, 599, 64) dtype=float32>
strides 步长 同样会影响输出的前两个维度，例如，(600, 600, 3）=> (300, 300, 64)，值得注意的是，括号里的数据可以不一致，分别控制横坐标和纵坐标，这里步长的计算公式为：
>>> from keras.layers import (Input, Conv2D)
>>> input = Input(shape=(600, 600, 3))
>>> Conv2D(64, 1, strides=(2, 2), name='conv1')(input)
<tf.Tensor 'conv1_4/BiasAdd:0' shape=(?, 300, 300, 64) dtype=float32>
padding 是否对周围进行填充，“same” 即使通过kernel_size 缩小了维度，但是四周会填充 0，保持原先的维度；“valid”表示存储不为0的有效信息。多个对比效果如下：
>>> Conv2D(64, 1, strides=(2, 2), padding="same", name='conv1')(input)
<tf.Tensor 'conv1_6/BiasAdd:0' shape=(?, 300, 300, 64) dtype=float32>
>>> Conv2D(64, 3, strides=(2, 2), padding="same", name='conv1')(input)
<tf.Tensor 'conv1_7/BiasAdd:0' shape=(?, 300, 300, 64) dtype=float32>
>>> Conv2D(64, 3, strides=(1, 1), padding="same", name='conv1')(input)
<tf.Tensor 'conv1_8/BiasAdd:0' shape=(?, 600, 600, 64) dtype=float32>
>>> Conv2D(64, 3, strides=(1, 1), padding="valid", name='conv1')(input)
<tf.Tensor 'conv1_9/BiasAdd:0' shape=(?, 598, 598, 64) dtype=float32>
通过这种最简单的方式，可以观察 ResNet50 的组成结构
Conv Block 的架构：
def conv_block(input_tensor, kernel_size, filters, stage, block, strides):
filters1, filters2, filters3 = filters # filters1 64, filters3 256 将数值传入到filters。。。中
conv_name_base = 'res' + str(stage) + block + '_branch'
bn_name_base = 'bn' + str(stage) + block + '_branch'
x = Conv2D(filters1, (1, 1), strides=strides, name=conv_name_base + '2a')(input_tensor)
x = BatchNormalization(name=bn_name_base + '2a')(x)
x = Activation('relu')(x)
x = Conv2D(filters2, kernel_size, padding='same', name=conv_name_base + '2b')(x)
x = BatchNormalization(name=bn_name_base + '2b')(x)
x = Activation('relu')(x)
x = Conv2D(filters3, (1, 1), name=conv_name_base + '2c')(x)
x = BatchNormalization(name=bn_name_base + '2c')(x)
shortcut = Conv2D(filters3, (1, 1), strides=strides, name=conv_name_base + '1')(input_tensor)
shortcut = BatchNormalization(name=bn_name_base + '1')(shortcut)
x = layers.add([x, shortcut])
x = Activation("relu")(x)
return x
Identity Block 的架构：
def identity_block(input_tensor, kernel_size, filters, stage, block):
filters1, filters2, filters3 = filters
conv_name_base = 'res' + str(stage) + block + '_branch'
bn_name_base = 'bn' + str(stage) + block + '_branch'
x = Conv2D(filters1, (1, 1), name=conv_name_base + '2a')(input_tensor)
x = BatchNormalization(name=bn_name_base + '2a')(x)
x = Activation('relu')(x)
x = Conv2D(filters2, kernel_size, padding='same', name=conv_name_base + '2b')(input_tensor)
x = BatchNormalization(name=bn_name_base + '2b')(x)
x = Activation('relu')(x)
x = Conv2D(filters3, (1, 1), name=conv_name_base + '2c')(input_tensor)
x = BatchNormalization(name=bn_name_base + '2c')(x)
x = layers.add([x, input_tensor])
x = Activation('relu')(x)
return x
最后是整体架构：
def ResNet50(inputs):
#-----------------------------------#
# 假设输入进来的图片是600,600,3
#-----------------------------------#
img_input = inputs
# 600,600,3 -> 300,300,64
x = ZeroPadding2D((3, 3))(img_input)
x = Conv2D(64, (7, 7), strides=(2, 2), name='conv1')(x)
x = BatchNormalization(name='bn_conv1')(x)
x = Activation('relu')(x)
# 300,300,64 -> 150,150,64
x = MaxPooling2D((3, 3), strides=(2, 2), padding="same")(x)
# 150,150,64 -> 150,150,256
x = conv_block(x, 3, [64, 64, 256], stage=2, block='a', strides=(1, 1))
x = identity_block(x, 3, [64, 64, 256], stage=2, block='b')
x = identity_block(x, 3, [64, 64, 256], stage=2, block='c')
# 150,150,256 -> 75,75,512
x = conv_block(x, 3, [128, 128, 512], stage=3, block='a')
x = identity_block(x, 3, [128, 128, 512], stage=3, block='b')
x = identity_block(x, 3, [128, 128, 512], stage=3, block='c')
x = identity_block(x, 3, [128, 128, 512], stage=3, block='d')
# 75,75,512 -> 38,38,1024
x = conv_block(x, 3, [256, 256, 1024], stage=4, block='a')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='b')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='c')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='d')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='e')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='f')
# 最终获得一个38,38,1024的共享特征层
return x
附上理论链接 Resnet-50网络结构详解 //www.cnblogs.com/qianchaomoon/p/12315906.html
