# ResNet
## 残差快
```python
import tensorflow as tf
class Residual(tf.keras.Model):
def __init__(self, num_channels, use_1x1conv=False, strides=1):
super(Residual, self).__init__()
self.conv1 = tf.keras.layers.Conv2D(num_channels, kernel_size=3, strides=strides, padding='same')
self.conv2 = tf.keras.layers.Conv2D(num_channels, kernel_size=3, strides=strides, padding='same')
self.conv3=None
if use_1x1conv:
self.conv3=tf.keras.layers.Conv2D(num_channels, kernel_size=1, strides=strides)
self.bn1 = tf.keras.layers.BatchNormalization()
self.bn2 = tf.keras.layers.BatchNormalization()
def call(self, X):
Y = tf.keras.activations.relu(self.bn1(self.conv1(X)))
Y = self.bn2(self.conv2(Y))
if self.conv3 is not None:
X = self.conv3(x)
Y += X
return tf.keras.activations.relu(Y)
```
```python
blk = Residual(3)
X = tf.random.uniform((4, 6, 6, 3))
Y = blk(X)
Y.shape
```
## ResNet模型
ResNet 的前两层跟之前介绍的 GoogLeNet 中的一样： 在输出通道数为 64、步幅为 2 的 7×77×7 卷积层后，接步幅为 2 的 3×33×3 的最大汇聚层。 不同之处在于 ResNet 每个卷积层后增加了批量归一化层。
```python
b1 = tf.keras.models.Sequential([
tf.keras.layers.Conv2D(64, kernel_size=7, strides=2, padding='same'),
tf.keras.layers.BatchNormalization(),
tf.keras.layers.Activation('relu'),
tf.keras.layers.Maxpool2D(pool_size=3, strides=2, padding='same'),
])
```
GoogLeNet 在后面接了 4 个由Inception块组成的模块。 ResNet 则使用 4 个由残差块组成的模块，每个模块使用若干个同样输出通道数的残差块。 第一个模块的通道数同输入通道数一致。 由于之前已经使用了步幅为 2 的最大汇聚层，所以无须减小高和宽。 之后的每个模块在第一个残差块里将上一个模块的通道数翻倍，并将高和宽减半。
```python
class ResnetBlock(tf.keras.layers.Layer):
def __init__(self, num_channels, num_residuals, first_block=False, **kwargs):
super(ResnetBlock, self).__init__(**kwargs)
self.residual_layers = []
for i in range(num_residuals):
if i == 0 and not first_block:
self.residual_layers.append(Residual(num_channels, use_1x1conv=True, strides=2))
else:
self.residual_layers.append(Residual(num_channels))
def call(self, X):
for layer in self.residual_layers.layers:
X = layer(X)
return X
```
```python
b2 = ResnetBlock(64, 2, first_block=True)
b3 = ResnetBlock(128, 2)
b4 = ResnetBlock(256, 2)
b5 = ResnetBlock(512, 2)
```
```python
def net():
return tf.keras.Sequential([
tf.keras.layers.Conv2D(64, kernel_size=7, strides=2, padding='same'),
tf.keras.layers.BatchNormalization(),
tf.keras.layers.Activation('relu'),
tf.keras.layers.Maxpool2D(pool_size=3, strides=2, padding='same'),
ResnetBlock(64, 2, first_block=True),
ResnetBlock(128, 2),
ResnetBlock(256, 2),
ResnetBlock(512, 2),
tf.keras.layers.GlobalAvgPool2D(),
tf.keras.layers.Dense(units=10)
])
```