在今年的3月7号，谷歌在 Tensorflow Developer Summit 2019 大会上发布 TensorFlow 2.0 Alpha
版，随后又发布了Beta版本。

Resnet18结构

Tensorflow搭建Resnet18

导入第三方库

import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers,Sequential

搭建BasicBlock

class BasicBlock(layers.Layer):
    def __init__(self,filter_num,stride=1):
        super(BasicBlock, self).__init__()
        self.conv1=layers.Conv2D(filter_num,(3,3),strides=stride,padding='same')
        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:
            self.downsample=Sequential()
            self.downsample.add(layers.Conv2D(filter_num,(1,1),strides=stride))
        else:
            self.downsample=lambda x:x
    def call(self,input,training=None):
        out=self.conv1(input)
        out=self.bn1(out)
        out=self.relu(out)

        out=self.conv2(out)
        out=self.bn2(out)

        identity=self.downsample(input)
        output=layers.add([out,identity])
        output=tf.nn.relu(output)
        return output

搭建ResNet

class ResNet(keras.Model):
    def __init__(self,layer_dims,num_classes=10):
        super(ResNet, self).__init__()
        # 预处理层
        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')
        ])
        # resblock
        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)

        # there are [b,512,h,w]
        # 自适应
        self.avgpool=layers.GlobalAveragePooling2D()
        self.fc=layers.Dense(num_classes)



    def call(self,input,training=None):
        x=self.stem(input)
        x=self.layer1(x)
        x=self.layer2(x)
        x=self.layer3(x)
        x=self.layer4(x)
        # [b,c]
        x=self.avgpool(x)
        x=self.fc(x)
        return x

    def build_resblock(self,filter_num,blocks,stride=1):
        res_blocks= Sequential()
        # may down sample
        res_blocks.add(BasicBlock(filter_num,stride))
        # just down sample one time
        for pre in range(1,blocks):
            res_blocks.add(BasicBlock(filter_num,stride=1))
        return res_blocks
def resnet18():
    return  ResNet([2,2,2,2])

训练数据

为了数据获取方便，这里使用的是CIFAR10的数据，可以在代码中直接使用keras.datasets.cifar10.load_data()方法获取，非常的方便
训练代码如下：

import os
import tensorflow as tf
from Resnet import resnet18
from tensorflow.keras import datasets,layers,optimizers,Sequential,metrics

os.environ["TF_CPP_MIN_LOG_LEVEL"]='2'
tf.random.set_seed(2345)


def preprocess(x,y):
    x=2*tf.cast(x,dtype=tf.float32)/255.-1
    y=tf.cast(y,dtype=tf.int32)
    return x,y
(x_train,y_train),(x_test,y_test)=datasets.cifar10.load_data()
y_train=tf.squeeze(y_train,axis=1)
y_test=tf.squeeze(y_test,axis=1)
# print(x_train.shape,y_train.shape,x_test.shape,y_test.shape)
train_data=tf.data.Dataset.from_tensor_slices((x_train,y_train))
train_data=train_data.shuffle(1000).map(preprocess).batch(64)

test_data=tf.data.Dataset.from_tensor_slices((x_test,y_test))
test_data=test_data.map(preprocess).batch(64)

sample=next(iter(train_data))
print('sample:',sample[0].shape,sample[1].shape,
      tf.reduce_min(sample[0]),tf.reduce_max(sample[0]))

def main():
    model=resnet18()
    model.build(input_shape=(None,32,32,3))
    model.summary()
    optimizer=optimizers.Adam(lr=1e-3)
    for epoch in range(50):
        for step,(x,y) in enumerate(train_data):
            with tf.GradientTape() as tape:
                logits=model(x)
                y_onehot=tf.one_hot(y,depth=10)
                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%100==0:
                print(epoch,step,'loss',float(loss))
        total_num=0
        total_correct=0
        for x,y in test_data:
            logits=model(x)
            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()

训练数据

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
sequential (Sequential)      multiple                  2048      
_________________________________________________________________
sequential_1 (Sequential)    multiple                  148736    
_________________________________________________________________
sequential_2 (Sequential)    multiple                  526976    
_________________________________________________________________
sequential_4 (Sequential)    multiple                  2102528   
_________________________________________________________________
sequential_6 (Sequential)    multiple                  8399360   
_________________________________________________________________
global_average_pooling2d (Gl multiple                  0         
_________________________________________________________________
dense (Dense)                multiple                  5130      
=================================================================
Total params: 11,184,778
Trainable params: 11,176,970
Non-trainable params: 7,808
_________________________________________________________________
0 0 loss 2.2936558723449707
0 100 loss 1.855604887008667
0 200 loss 1.9335857629776
0 300 loss 1.508711576461792
0 400 loss 1.5679863691329956
0 500 loss 1.5649926662445068
0 600 loss 1.147849202156067
0 700 loss 1.3818628787994385
0 acc: 0.5424
1 0 loss 1.3022596836090088
1 100 loss 1.4624202251434326
1 200 loss 1.3188159465789795
1 300 loss 1.1521495580673218
1 400 loss 0.9550357460975647
1 500 loss 1.2304189205169678
1 600 loss 0.7009983062744141
1 700 loss 0.8488335609436035
1 acc: 0.644
2 0 loss 0.9625152945518494
2 100 loss 1.174363374710083
2 200 loss 1.1750390529632568
2 300 loss 0.7221378087997437
2 400 loss 0.7162064909934998
2 500 loss 0.926654040813446
2 600 loss 0.6159981489181519
2 700 loss 0.6437114477157593
2 acc: 0.6905
3 0 loss 0.7495195865631104
3 100 loss 0.9840961694717407
3 200 loss 0.9429250955581665
3 300 loss 0.5575872659683228
3 400 loss 0.5735365152359009
3 500 loss 0.7843905687332153
3 600 loss 0.6125107407569885
3 700 loss 0.6241222620010376
3 acc: 0.6933
4 0 loss 0.7694090604782104
4 100 loss 0.5488263368606567
4 200 loss 0.9142876863479614
4 300 loss 0.4908181428909302
4 400 loss 0.5889899730682373
4 500 loss 0.7341771125793457
4 600 loss 0.4880038797855377
4 700 loss 0.5088012218475342
4 acc: 0.7241
5 0 loss 0.5378311276435852
5 100 loss 0.5630106925964355
5 200 loss 0.8578733205795288
5 300 loss 0.3617972433567047
5 400 loss 0.29359108209609985
5 500 loss 0.5915042757987976
5 600 loss 0.3684327006340027
5 700 loss 0.40654802322387695
5 acc: 0.7005
6 0 loss 0.5005596280097961
6 100 loss 0.40528279542922974
6 200 loss 0.4127967953681946
6 300 loss 0.4062516987323761
6 400 loss 0.40751856565475464
6 500 loss 0.45849910378456116
6 600 loss 0.4571283459663391
6 700 loss 0.32558882236480713
6 acc: 0.7119

可以看到使用ResNet18网络结构，参数量是非常大的，有 11,184,778，所以训练起来的话，很耗时间，这里笔者没有训练完，有兴趣的同学，可以训练一下