神经网络学习小记录22——Xception模型的复现详解

学习前言

Xception是继Inception后提出的对Inception v3的另一种改进，学一学总是好的。

什么是Xception模型

Xception是谷歌公司继Inception后，提出的InceptionV3的一种改进模型，其改进的主要内容为采用depthwise separable convolution来替换原来Inception v3中的多尺寸卷积核特征响应操作。
在讲Xception模型之前，首先要讲一下什么是depthwise separable convolution。
对于一个卷积点而言：
假设有一个3×3大小的卷积层，其输入通道为16、输出通道为32。具体为，32个3×3大小的卷积核会遍历16个通道中的每个数据，最后可得到所需的32个输出通道，所需参数为16×32×3×3=4608个。

应用深度可分离卷积，用16个3×3大小的卷积核分别遍历16通道的数据，得到了16个特征图谱。在融合操作之前，接着用32个1×1大小的卷积核遍历这16个特征图谱，所需参数为16×3×3+16×32×1×1=656个。
可以看出来depthwise separable convolution可以减少模型的参数。

通俗地理解就是卷积核厚度只有一层，然后在输入张量上一层一层地滑动，所以一个卷积核就对应了一个输出通道，当卷积完成后，在利用1x1的卷积调整厚度。

不过Xception模型中的depthwise separable convolution和普通的不太一样，普通的depthwise separable convolution是先进行3x3操作再进行1x1操作（就好像mobileNet中的一样），而Xception模型中则是先进行1x1操作然后再进行3x3操作，中间为了保证数据不被破坏，没有添加relu层，而mobileNet添加了relu层。在建立模型的时候，可以使用Keras中的SeparableConv2D层建立相应的功能。

对于Xception模型而言，其一共可以分为3个flow，分别是Entry flow、Middle flow、Exit flow；分为14个block，其中Entry flow中有4个、Middle flow中有8个、Exit flow中有2个。具体结构如下：

其内部主要结构就是残差卷积网络搭配SeparableConv2D层实现一个个block，在Xception模型中，常见的两个block的结构如下。
这个主要在Entry flow和Exit flow中：

这个主要在Middle flow中：

Xception网络部分实现代码

#-------------------------------------------------------------#
#   Xception的网络部分
#-------------------------------------------------------------#
from keras.preprocessing import image

from keras.models import Model
from keras import layers
from keras.layers import Dense,Input,BatchNormalization,Activation,Conv2D,SeparableConv2D,MaxPooling2D
from keras.layers import GlobalAveragePooling2D,GlobalMaxPooling2D
from keras import backend as K
from keras.applications.imagenet_utils import decode_predictions


def Xception(input_shape = [299,299,3],classes=1000):


    img_input = Input(shape=input_shape)

    #--------------------------#
    # Entry flow
    #--------------------------#
    #--------------------#
    # block1
    #--------------------#
    # 299,299,3 -> 149,149,64
    x = Conv2D(32, (3, 3), strides=(2, 2), use_bias=False, name='block1_conv1')(img_input)
    x = BatchNormalization(name='block1_conv1_bn')(x)
    x = Activation('relu', name='block1_conv1_act')(x)
    x = Conv2D(64, (3, 3), use_bias=False, name='block1_conv2')(x)
    x = BatchNormalization(name='block1_conv2_bn')(x)
    x = Activation('relu', name='block1_conv2_act')(x)

    #--------------------#
    # block2
    #--------------------#
    # 149,149,64 -> 75,75,128
    residual = Conv2D(128, (1, 1), strides=(2, 2), padding='same', use_bias=False)(x)
    residual = BatchNormalization()(residual)

    x = SeparableConv2D(128, (3, 3), padding='same', use_bias=False, name='block2_sepconv1')(x)
    x = BatchNormalization(name='block2_sepconv1_bn')(x)
    x = Activation('relu', name='block2_sepconv2_act')(x)
    x = SeparableConv2D(128, (3, 3), padding='same', use_bias=False, name='block2_sepconv2')(x)
    x = BatchNormalization(name='block2_sepconv2_bn')(x)

    x = MaxPooling2D((3, 3), strides=(2, 2), padding='same', name='block2_pool')(x)
    x = layers.add([x, residual])

    #--------------------#
    # block3
    #--------------------#
    # 75,75,128 -> 38,38,256
    residual = Conv2D(256, (1, 1), strides=(2, 2),padding='same', use_bias=False)(x)
    residual = BatchNormalization()(residual)

    x = Activation('relu', name='block3_sepconv1_act')(x)
    x = SeparableConv2D(256, (3, 3), padding='same', use_bias=False, name='block3_sepconv1')(x)
    x = BatchNormalization(name='block3_sepconv1_bn')(x)
    x = Activation('relu', name='block3_sepconv2_act')(x)
    x = SeparableConv2D(256, (3, 3), padding='same', use_bias=False, name='block3_sepconv2')(x)
    x = BatchNormalization(name='block3_sepconv2_bn')(x)

    x = MaxPooling2D((3, 3), strides=(2, 2), padding='same', name='block3_pool')(x)
    x = layers.add([x, residual])

    #--------------------#
    # block4
    #--------------------#
    # 38,38,256 -> 19,19,728
    residual = Conv2D(728, (1, 1), strides=(2, 2),padding='same', use_bias=False)(x)
    residual = BatchNormalization()(residual)

    x = Activation('relu', name='block4_sepconv1_act')(x)
    x = SeparableConv2D(728, (3, 3), padding='same', use_bias=False, name='block4_sepconv1')(x)
    x = BatchNormalization(name='block4_sepconv1_bn')(x)
    x = Activation('relu', name='block4_sepconv2_act')(x)
    x = SeparableConv2D(728, (3, 3), padding='same', use_bias=False, name='block4_sepconv2')(x)
    x = BatchNormalization(name='block4_sepconv2_bn')(x)

    x = MaxPooling2D((3, 3), strides=(2, 2), padding='same', name='block4_pool')(x)
    x = layers.add([x, residual])

    #--------------------------#
    # Middle flow
    #--------------------------#
    #--------------------#
    # block5--block12
    #--------------------#
    # 19,19,728 -> 19,19,728
    for i in range(8):
        residual = x
        prefix = 'block' + str(i + 5)

        x = Activation('relu', name=prefix + '_sepconv1_act')(x)
        x = SeparableConv2D(728, (3, 3), padding='same', use_bias=False, name=prefix + '_sepconv1')(x)
        x = BatchNormalization(name=prefix + '_sepconv1_bn')(x)
        x = Activation('relu', name=prefix + '_sepconv2_act')(x)
        x = SeparableConv2D(728, (3, 3), padding='same', use_bias=False, name=prefix + '_sepconv2')(x)
        x = BatchNormalization(name=prefix + '_sepconv2_bn')(x)
        x = Activation('relu', name=prefix + '_sepconv3_act')(x)
        x = SeparableConv2D(728, (3, 3), padding='same', use_bias=False, name=prefix + '_sepconv3')(x)
        x = BatchNormalization(name=prefix + '_sepconv3_bn')(x)

        x = layers.add([x, residual])

    #--------------------------#
    # Exit flow
    #--------------------------#
    #--------------------#
    # block13
    #--------------------#
    # 19,19,728 -> 10,10,1024
    residual = Conv2D(1024, (1, 1), strides=(2, 2),
                      padding='same', use_bias=False)(x)
    residual = BatchNormalization()(residual)

    x = Activation('relu', name='block13_sepconv1_act')(x)
    x = SeparableConv2D(728, (3, 3), padding='same', use_bias=False, name='block13_sepconv1')(x)
    x = BatchNormalization(name='block13_sepconv1_bn')(x)
    x = Activation('relu', name='block13_sepconv2_act')(x)
    x = SeparableConv2D(1024, (3, 3), padding='same', use_bias=False, name='block13_sepconv2')(x)
    x = BatchNormalization(name='block13_sepconv2_bn')(x)

    x = MaxPooling2D((3, 3), strides=(2, 2), padding='same', name='block13_pool')(x)
    x = layers.add([x, residual])

    #--------------------#
    # block14
    #--------------------#
    # 10,10,1024 -> 10,10,2048
    x = SeparableConv2D(1536, (3, 3), padding='same', use_bias=False, name='block14_sepconv1')(x)
    x = BatchNormalization(name='block14_sepconv1_bn')(x)
    x = Activation('relu', name='block14_sepconv1_act')(x)

    x = SeparableConv2D(2048, (3, 3), padding='same', use_bias=False, name='block14_sepconv2')(x)
    x = BatchNormalization(name='block14_sepconv2_bn')(x)
    x = Activation('relu', name='block14_sepconv2_act')(x)

    x = GlobalAveragePooling2D(name='avg_pool')(x)
    x = Dense(classes, activation='softmax', name='predictions')(x)

    inputs = img_input

    model = Model(inputs, x, name='xception')

    model.load_weights("xception_weights_tf_dim_ordering_tf_kernels.h5")

    return model

图片预测

建立网络后，可以用以下的代码进行预测。

def preprocess_input(x):
    x /= 255.
    x -= 0.5
    x *= 2.
    return x


if __name__ == '__main__':
    model = Xception()

    img_path = 'elephant.jpg'
    img = image.load_img(img_path, target_size=(299, 299))
    x = image.img_to_array(img)
    x = np.expand_dims(x, axis=0)
    x = preprocess_input(x)
    print('Input image shape:', x.shape)

    preds = model.predict(x)
    print(np.argmax(preds))
    print('Predicted:', decode_predictions(preds))

预测所需的已经训练好的Xception模型可以在https://github.com/fchollet/deep-learning-models/releases下载。非常方便。
预测结果为：

Predicted: [[('n02504458', 'African_elephant', 0.47570863), ('n01871265', 'tusker', 0.3173351), ('n02504013', 'Indian_elephant', 0.030323735), ('n02963159', 'cardigan', 0.0007877756), ('n02410509', 'bison', 0.00075616257)]]