使用预训练的卷积神经网络（猫狗图片分类）

本次所用数据来自ImageNet，使用预训练好的数据来预测一个新的数据集：猫狗图片分类。这里，使用VGG模型，这个模型内置在Keras中，直接导入就可以了。

from keras.applications import VGG16conv_base = VGG16(weights='imagenet',include_top=False,input_shape=(150, 150, 3))

说一下这三个参数：

weights：指定模型初始化权重检查点
include_top：指定模型最后是否包含密集连接分类器。默认情况下，这个密集连接分类器对应于ImageNet的1000个类别。因为我们打算使用自己的分类器（只有两个类别：cat和dog），所以不用包含。
input_shape：输入到网络中的图像张量（可选参数），如果不传入这个参数，那么网络可以处理任意形状的输入

看一下VGG16网络的详细构架：

conv_base.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
input_1 (InputLayer)         (None, 150, 150, 3)       0
_________________________________________________________________
block1_conv1 (Conv2D)        (None, 150, 150, 64)      1792
_________________________________________________________________
block1_conv2 (Conv2D)        (None, 150, 150, 64)      36928
_________________________________________________________________
block1_pool (MaxPooling2D)   (None, 75, 75, 64)        0
_________________________________________________________________
block2_conv1 (Conv2D)        (None, 75, 75, 128)       73856
_________________________________________________________________
block2_conv2 (Conv2D)        (None, 75, 75, 128)       147584
_________________________________________________________________
block2_pool (MaxPooling2D)   (None, 37, 37, 128)       0
_________________________________________________________________
block3_conv1 (Conv2D)        (None, 37, 37, 256)       295168
_________________________________________________________________
block3_conv2 (Conv2D)        (None, 37, 37, 256)       590080
_________________________________________________________________
block3_conv3 (Conv2D)        (None, 37, 37, 256)       590080
_________________________________________________________________
block3_pool (MaxPooling2D)   (None, 18, 18, 256)       0
_________________________________________________________________
block4_conv1 (Conv2D)        (None, 18, 18, 512)       1180160
_________________________________________________________________
block4_conv2 (Conv2D)        (None, 18, 18, 512)       2359808
_________________________________________________________________
block4_conv3 (Conv2D)        (None, 18, 18, 512)       2359808
_________________________________________________________________
block4_pool (MaxPooling2D)   (None, 9, 9, 512)         0
_________________________________________________________________
block5_conv1 (Conv2D)        (None, 9, 9, 512)         2359808
_________________________________________________________________
block5_conv2 (Conv2D)        (None, 9, 9, 512)         2359808
_________________________________________________________________
block5_conv3 (Conv2D)        (None, 9, 9, 512)         2359808
_________________________________________________________________
block5_pool (MaxPooling2D)   (None, 4, 4, 512)         0
=================================================================
Total params: 14,714,688
Trainable params: 14,714,688
Non-trainable params: 0

最后这个特征图形状为（4， 4， 512），我们在这个特征上面添加一个密集连接分类器。

不使用数据增强的快速特征提取（计算代价低）

首先，运行ImageDataGenerator实例，将图像及其标签提取为Numpy数组，调用conv_base模型的predict方法从这些图像的中提取特征。

import os
import numpy as np
from keras.preprocessing.image import ImageDataGeneratorbase_dir = '/Users/fchollet/Downloads/cats_and_dogs_small'train_dir = os.path.join(base_dir, 'train')
validation_dir = os.path.join(base_dir, 'validation')
test_dir = os.path.join(base_dir, 'test')datagen = ImageDataGenerator(rescale=1./255)
batch_size = 20def extract_features(directory, sample_count):features = np.zeros(shape=(sample_count, 4, 4, 512))labels = np.zeros(shape=(sample_count))generator = datagen.flow_from_directory(directory,target_size=(150, 150),batch_size=batch_size,class_mode='binary')i = 0for inputs_batch, labels_batch in generator:features_batch = conv_base.predict(inputs_batch)features[i * batch_size : (i + 1) * batch_size] = features_batchlabels[i * batch_size : (i + 1) * batch_size] = labels_batchi += 1if i * batch_size >= sample_count:break  # 这些生成器在循环中不断生成数据，所以你必须在读完所有图像之后终止循环return features, labelstrain_features, train_labels = extract_features(train_dir, 2000)
validation_features, validation_labels = extract_features(validation_dir, 1000)
test_features, test_labels = extract_features(test_dir, 1000)

目前，提取的特征形状为（samples， 4， 4， 512），我们要将其输入到密集连接分类器中去，所以必须首先对其形状展平为（samples ，8192）

train_features = np.reshape(train_features, (2000, 4 * 4 * 512))
validation_features = np.reshape(validation_features, (1000, 4 * 4 * 512))
test_features = np.reshape(test_features, (1000, 4 * 4 * 512))

下面定义一个密集连接分类器，并在刚刚保存好的数据和标签上训练分类器：

from keras import models
from keras import layers
from keras import optimizersmodel = models.Sequential()
model.add(layers.Dense(256, activation='relu', input_dim=4 * 4 * 512))
model.add(layers.Dropout(0.5))
model.add(layers.Dense(1, activation='sigmoid'))model.compile(optimizer=optimizers.RMSprop(lr=2e-5),loss='binary_crossentropy',metrics=['acc'])history = model.fit(train_features, train_labels,epochs=30,batch_size=20,validation_data=(validation_features, validation_labels))

训练速度非常快，因为只需要处理两个Dense层。下面看一下训练过程中的损失曲线和精度曲线：

import matplotlib.pyplot as pltacc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']epochs = range(len(acc))plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()plt.figure()plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()plt.show()

从图中可以看出，验证精度达到了约90%，比之前从一开始就训练小型模型效果要好很多，但是从图中也可以看出，虽然dropout比率比较大，但模型从一开始就出现了过拟合。这是因为本方法中没有使用数据增强，而数据增强对防止小型图片数据集过拟合非常重要。

使用数据增强的特征提取（计算代价高）

这种方法速度更慢，计算代价更高，但是可以在训练期间使用数据增强。这种方法是：扩展conv_base模型，然后在输入数据上端到端的运行模型。（这种方法计算代价很高，必须在GPU上运行）

承接我们之前定义的网络模型

from keras import models
from keras import layersmodel = models.Sequential()
model.add(conv_base)
model.add(layers.Flatten())
model.add(layers.Dense(256, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))
model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
vgg16 (Model)                (None, 4, 4, 512)         14714688
_________________________________________________________________
flatten_1 (Flatten)          (None, 8192)              0
_________________________________________________________________
dense_3 (Dense)              (None, 256)               2097408
_________________________________________________________________
dense_4 (Dense)              (None, 1)                 257
=================================================================
Total params: 16,812,353
Trainable params: 16,812,353
Non-trainable params: 0

我们可以看到，VGG16的卷积基一共有14714688个参数，其上添加的分类器一共有200万个参数，非常多。

在编译和训练模型之前，需要冻结卷积基。冻结一个或多个层是指在训练过程中保持其权重不变。如果不这么做，那么卷积基之前学到的表示将会在训练过程中被修改。因为其上添加的Dense是随机初始化的，所以非常打的权重更新会在网络中进行传播，对之前学到的表示造成很大破坏。

在Keras中，冻结网络的方法是将其trainable属性设置为False

print('This is the number of trainable weights ''before freezing the conv base:', len(model.trainable_weights))

This is the number of trainable weights before freezing the conv base: 30

conv_base.trainable = False
print('This is the number of trainable weights ''after freezing the conv base:', len(model.trainable_weights))

This is the number of trainable weights after freezing the conv base: 4

如此设置之后，只有添加的两个Dense层的权重才会被训练，总共有4个权重张量，每层2个（主权重矩阵和偏置向量），注意的是，如果想修改权重属性trainable，那么应该修改好属性之后再编译模型。

下面，我们可以训练模型了，并使用数据增强的办法：

from keras.preprocessing.image import ImageDataGeneratortrain_datagen = ImageDataGenerator(rescale=1./255,rotation_range=40,width_shift_range=0.2,height_shift_range=0.2,shear_range=0.2,zoom_range=0.2,horizontal_flip=True,fill_mode='nearest')# Note that the validation data should not be augmented!
test_datagen = ImageDataGenerator(rescale=1./255)train_generator = train_datagen.flow_from_directory(# This is the target directorytrain_dir,# All images will be resized to 150x150target_size=(150, 150),batch_size=20,# Since we use binary_crossentropy loss, we need binary labelsclass_mode='binary')validation_generator = test_datagen.flow_from_directory(validation_dir,target_size=(150, 150),batch_size=20,class_mode='binary')model.compile(loss='binary_crossentropy',optimizer=optimizers.RMSprop(lr=2e-5),metrics=['acc'])history = model.fit_generator(train_generator,steps_per_epoch=100,epochs=30,validation_data=validation_generator,validation_steps=50,verbose=2)

model.save('cats_and_dogs_small_3.h5')

我们再来看看验证精度：

acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']epochs = range(len(acc))plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()plt.figure()plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()plt.show()

验证精度到了将近96%，而且减少了过拟合。

微调模型

我们下面使用模型微调，进一步提高模型的性能。模型微调的步骤如下：

（1）在已经训练好的基网络（base network）上添加自定义网络
（2）冻结基网络
（3）训练所添加的部分
（4）解冻基网络的一些层
（5）联合训练解冻的这些层和添加的部分

在做特征提取的时候已经完成了前三个步骤。我们继续第四个步骤，先解冻conv_base，然后冻结其中的部分层。

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
input_1 (InputLayer)         (None, 150, 150, 3)       0
_________________________________________________________________
block1_conv1 (Conv2D)        (None, 150, 150, 64)      1792
_________________________________________________________________
block1_conv2 (Conv2D)        (None, 150, 150, 64)      36928
_________________________________________________________________
block1_pool (MaxPooling2D)   (None, 75, 75, 64)        0
_________________________________________________________________
block2_conv1 (Conv2D)        (None, 75, 75, 128)       73856
_________________________________________________________________
block2_conv2 (Conv2D)        (None, 75, 75, 128)       147584
_________________________________________________________________
block2_pool (MaxPooling2D)   (None, 37, 37, 128)       0
_________________________________________________________________
block3_conv1 (Conv2D)        (None, 37, 37, 256)       295168
_________________________________________________________________
block3_conv2 (Conv2D)        (None, 37, 37, 256)       590080
_________________________________________________________________
block3_conv3 (Conv2D)        (None, 37, 37, 256)       590080
_________________________________________________________________
block3_pool (MaxPooling2D)   (None, 18, 18, 256)       0
_________________________________________________________________
block4_conv1 (Conv2D)        (None, 18, 18, 512)       1180160
_________________________________________________________________
block4_conv2 (Conv2D)        (None, 18, 18, 512)       2359808
_________________________________________________________________
block4_conv3 (Conv2D)        (None, 18, 18, 512)       2359808
_________________________________________________________________
block4_pool (MaxPooling2D)   (None, 9, 9, 512)         0
_________________________________________________________________
block5_conv1 (Conv2D)        (None, 9, 9, 512)         2359808
_________________________________________________________________
block5_conv2 (Conv2D)        (None, 9, 9, 512)         2359808
_________________________________________________________________
block5_conv3 (Conv2D)        (None, 9, 9, 512)         2359808
_________________________________________________________________
block5_pool (MaxPooling2D)   (None, 4, 4, 512)         0
=================================================================
Total params: 14,714,688
Trainable params: 14,714,688
Non-trainable params: 0

再回顾一下这些层，我们将微调最后三个卷积层，也就是说，知道block4_pool的所有层都应该被冻结，后面三层来进行训练。

要知道，训练的参数越多，过拟合的风险越大。卷积基有1500万个参数，所以你在小型数据集上训练这么多参数是有风险的。因此，这种情况下最好的策略是仅微调卷积基最后三两层。

conv_base.trainable = Trueset_trainable = False
for layer in conv_base.layers:if layer.name == 'block5_conv1':set_trainable = Trueif set_trainable:layer.trainable = Trueelse:layer.trainable = False

现在可以微调网络了，我们将使用学习率非常小的RMSProp优化器来实现。之所以让学习率很小，是因为对于微调网络的三层表示，我们希望其变化范围不要太大，太大的权重可能会破坏这些表示。

model.compile(loss='binary_crossentropy',optimizer=optimizers.RMSprop(lr=1e-5),metrics=['acc'])history = model.fit_generator(train_generator,steps_per_epoch=100,epochs=100,validation_data=validation_generator,validation_steps=50)
model.save('cats_and_dogs_small_4.h5')

下面，绘制曲线看看效果：

acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']epochs = range(len(acc))plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()plt.figure()plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()plt.show()

这些曲线看起来包含噪音。为了让图像更具有可读性，可以让每个损失精度替换为指数移动平均，从而让曲线变得更加平滑，下面用一个简单实用函数来实现：

def smooth_curve(points, factor=0.8):smoothed_points = []for point in points:if smoothed_points:previous = smoothed_points[-1]smoothed_points.append(previous * factor + point * (1 - factor))else:smoothed_points.append(point)return smoothed_pointsplt.plot(epochs,smooth_curve(acc), 'bo', label='Smoothed training acc')
plt.plot(epochs,smooth_curve(val_acc), 'b', label='Smoothed validation acc')
plt.title('Training and validation accuracy')
plt.legend()plt.figure()plt.plot(epochs,smooth_curve(loss), 'bo', label='Smoothed training loss')
plt.plot(epochs,smooth_curve(val_loss), 'b', label='Smoothed validation loss')
plt.title('Training and validation loss')
plt.legend()plt.show()

通过指数移动平均，验证曲线变得更清楚了。可以看到，精度提高了1%，约从96%提高到了97%。

下面，在测试集上评估一下这个模型

test_generator = test_datagen.flow_from_directory(test_dir,target_size=(150, 150),batch_size=20,class_mode='binary')test_loss, test_acc = model.evaluate_generator(test_generator, steps=50)
print('test acc:', test_acc)

Found 1000 images belonging to 2 classes.
test acc: 0.967999992371

得到了差不多97%的测试精度，在关于这个数据集的原始Kaggle竞赛中，这个结果是最佳结果之一。
值得注意的是，我们只是用了一小部分训练数据（约10%）就得到了这个结果。训练20000个样本和训练2000个样本还是有很大差别的。

更多精彩内容，欢迎关注我的微信公众号：数据瞎分析