Tensorflow2.0 keras ResNet18 34 50 101 152系列 代码实现

模型介绍参看：博文

可以看出ResNet系列最多层数达到了152层，但是基本结构可以分为四个模块，即特征层分别为64，128，256，512的卷积层block；

每个卷积层block中如上图，由两组卷积层由两层卷积核大小为3x3组成，每一层采取了标准化（normalization）处理，激活函数是relu，第二个relu在残差处理后进行；
但是根据不同深度， 每个卷积层模块的block数量存在差异，可以分为两组（18，34）与（50，101，152）；
因此建立基本模块类建立ResNet是更优的选择。

迁移学习

康康标准答案，这里是ResNet152

import  tensorflow as tf
from    tensorflow import keras

base_model = keras.applications.ResNet152(weights='imagenet')
base_model.summary()

截取第5卷积层的第3个block的网络结果，

自建模型

提供两种不同的建模方法，keras搭建深度学习模型的若干方法：博文

ResNet50 101 152 models.Model建模方法

import  tensorflow as tf
from    tensorflow import keras
from    tensorflow.keras import layers, models, Sequential
from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, Dense, Flatten, Dropout, BatchNormalization, Activation, GlobalAveragePooling2D

# 继承Layer,建立resnet50 101 152卷积层模块
def conv_block(inputs, filter_num, stride=1, name=None):
    
    x = inputs
    x = Conv2D(filter_num[0], (1,1), strides=stride, padding='same', name=name+'_conv1')(x)
    x = BatchNormalization(axis=3, name=name+'_bn1')(x)
    x = Activation('relu', name=name+'_relu1')(x)

    x = Conv2D(filter_num[1], (3,3), strides=1, padding='same', name=name+'_conv2')(x)
    x = BatchNormalization(axis=3, name=name+'_bn2')(x)
    x = Activation('relu', name=name+'_relu2')(x)

    x = Conv2D(filter_num[2], (1,1), strides=1, padding='same', name=name+'_conv3')(x)
    x = BatchNormalization(axis=3, name=name+'_bn3')(x)
    
    # residual connection
    r = Conv2D(filter_num[2], (1,1), strides=stride, padding='same', name=name+'_residual')(inputs)
    x = layers.add([x, r])
    x = Activation('relu', name=name+'_relu3')(x)

    return x

def build_block (x, filter_num, blocks, stride=1, name=None):

    x = conv_block(x, filter_num, stride, name=name)

    for i in range(1, blocks):
        x = conv_block(x, filter_num, stride=1, name=name+'_block'+str(i))

    return x


# 创建resnet50 101 152
def ResNet(Netname, nb_classes):

    R