Tensorflow之CNN实战
基本卷积操作
import tensorflow as tf
from tensorflow.keras import layers
# 4代表卷积核个数，也就是输出维度；5代表卷积核大小；strides=2则size减半
# padding=valid代表不padding，padding=same代表padding成size不变
layer = layers.Conv2D(4,kernel_size=5,strides=1,padding='valid')
# x : [b,32,32,3]
out = layer(x) # [b,28,28,4],此时权重w的size为[28,28,3,4],b:[4]
池化与采样
- 池化
x # [1,14,14,4]
# pooling不会改变channel，对每个channel进行同样的操作
pool = layers.MaxPool2D(2,strides=2)
out = pool(x) # [1,7,7,4]
- 上采样
x # [1,7,7,4]
layer = layers.Upsampling2D(size=3)
out = layer(x) # [1,21,21,4],每一个pixel变为3*3大小的块，这个块内的pixel值相等
- Relu函数
layers.ReLU()(x)
Batch Normalization
让每一层数据的分布都比较规范，易于学习，不易发生梯度离散或者梯度爆炸
-
batch normalization
[N,C,H*W] => [C] : 表示每一个channel上一个Batch的均值
-
layer normalization
[N,C,H*W] => [N] : 表示每一个实例的均值
-
instance normalization
[N,C,H*W] => [N,C] : 表示每一个实例上每一个channel的均值
-
group normalization : 表示每一个实例上一部分channel的均值
-
net = layers.BatchNormalization() 其参数如下：axis代表在哪个维度进行BN，center=True代表均值，scale=True代表方差，trainable=True
x = tf.random.normal([2,4,4,3],mean=1,stddev=0.5)
net = layers.BatchNormalization(axis=3)
out = net(x,training=True)
# net.variables : 可见逐渐均值为1，标准差为0.5
实战ResNet18
- ResNet实现
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers, Sequential
class BasicBlock(layers.Layer): # 2层
def __init__(self, filter_num, stride=1):
super(BasicBlock, self).__init__()
self.conv1 = layers.Conv2D(filter_num, (3, 3), strides=stride, padding='same') # 通过stride取值看是否下采样
self.bn1 = layers.BatchNormalization()
self.relu = layers.Activation('relu')
self.conv2 = layers.Conv2D(filter_num, (3, 3), strides=1, padding='same')
self.bn2 = layers.BatchNormalization()
if stride != 1: # 下采样
self.downsample = Sequential()
self.downsample.add(layers.Conv2D(filter_num, (1, 1), strides=stride))
self.downsample.add(layers.BatchNormalization())
else: # 未下采样
self.downsample = lambda x:x
def call(self, inputs, training=None):
# [b, h, w, c], pytorch是[b, c, h, w]
out = self.conv1(inputs)
out = self.bn1(out,training=training)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out,training=training)
identity = self.downsample(inputs)
output = layers.add([out, identity])
output = tf.nn.relu(output) # 不能用self.relu(),防止共享激活函数层
return output
class ResNet(keras.Model):
def __init__(self, layer_dims, num_classes=100): # [2, 2, 2, 2],每个resblock有几个basicblock
super(ResNet, self).__init__()
self.stem = Sequential([layers.Conv2D(64, (3, 3), strides=(1, 1)),
layers.BatchNormalization(),
layers.Activation('relu'),
layers.MaxPool2D(pool_size=(2, 2), strides=(1, 1), padding='same')
])
self.layer1 = self.build_resblock(64, layer_dims[0])
self.layer2 = self.build_resblock(128, layer_dims[1], stride=2)
self.layer3 = self.build_resblock(256, layer_dims[2], stride=2)
self.layer4 = self.build_resblock(512, layer_dims[3], stride=2)
# output: [b, 512, h, w],
self.avgpool = layers.GlobalAveragePooling2D()
self.fc = layers.Dense(num_classes)
def call(self, inputs, training=None):
x = self.stem(inputs,training=training)
x = self.layer1(x,training=training)
x = self.layer2(x,training=training)
x = self.layer3(x,training=training)
x = self.layer4(x,training=training)
x = self.avgpool(x)
# 输出为：[b, 100]
x = self.fc(x)
return x
def build_resblock(self, filter_num, blocks, stride=1):
# 参数：basicblock数量以及channel数量
res_blocks = Sequential()
# 只有这一个basicblock可以进行下采样
res_blocks.add(BasicBlock(filter_num, stride))
for _ in range(1, blocks):
res_blocks.add(BasicBlock(filter_num, stride=1))
return res_blocks
def resnet18():
return ResNet([2, 2, 2, 2]) # 有2*2*4+1+1=18层
def resnet34():
return ResNet([3, 4, 6, 3]) # 有3*2+4*2+6*2+3*2+1+1=34层
- 训练与测试
import os
import tensorflow as tf
from tensorflow.keras import layers, optimizers, datasets, Sequential
from resnet import resnet18
tf.random.set_seed(2345)
def preprocess(x, y):
# 数据范围预处理成：[-1~1]
x = tf.cast(x, dtype=tf.float32) / 255. - 0.5
y = tf.cast(y, dtype=tf.int32)
return x,y
(x,y), (x_test, y_test) = datasets.cifar100.load_data()
y = tf.squeeze(y, axis=1)
y_test = tf.squeeze(y_test, axis=1)
train_db = tf.data.Dataset.from_tensor_slices((x,y))
train_db = train_db.shuffle(1000).map(preprocess).batch(512)
test_db = tf.data.Dataset.from_tensor_slices((x_test,y_test))
test_db = test_db.map(preprocess).batch(512)
def main():
# [b, 32, 32, 3] => [b, 1, 1, 512]
model = resnet18()
model.build(input_shape=(None, 32, 32, 3))
model.summary()
optimizer = optimizers.Adam(lr=1e-3)
for epoch in range(500):
for step, (x,y) in enumerate(train_db):
with tf.GradientTape() as tape:
# [b, 32, 32, 3] => [b, 100]
logits = model(x,training=True)
# [b] => [b, 100]
y_onehot = tf.one_hot(y, depth=100)
# compute loss
loss = tf.losses.categorical_crossentropy(y_onehot, logits, from_logits=True)
loss = tf.reduce_mean(loss)
grads = tape.gradient(loss, model.trainable_variables)
optimizer.apply_gradients(zip(grads, model.trainable_variables))
if step % 50 == 0:
print(epoch, step, 'loss:', float(loss))
total_num = 0
total_correct = 0
for x,y in test_db:
logits = model(x,training=False)
prob = tf.nn.softmax(logits, axis=1)
pred = tf.argmax(prob, axis=1)
pred = tf.cast(pred, dtype=tf.int32)
correct = tf.cast(tf.equal(pred, y), dtype=tf.int32)
correct = tf.reduce_sum(correct)
total_num += x.shape[0]
total_correct += int(correct)
acc = total_correct / total_num
print(epoch, 'acc:', acc)
if __name__ == '__main__':
main()