keras自动编码器实现系列之卷积自动编码器操作

图片的自动编码很容易就想到用卷积神经网络做为编码-解码器。在实际的操作中，

也经常使用卷积自动编码器去解决图像编码问题，而且非常有效。

下面通过**keras**完成简单的卷积自动编码。 编码器有堆叠的卷积层和池化层(max pooling用于空间降采样)组成。 对应的解码器由卷积层和上采样层组成。

@requires_authorization
# -*- coding:utf-8 -*-

from keras.layers import Input, Dense, Conv2D, MaxPooling2D, UpSampling2D
from keras.models import Model
from keras import backend as K
import os

## 网络结构 ##
input_img = Input(shape=(28,28,1)) # Tensorflow后端， 注意要用channel_last
# 编码器部分
x = Conv2D(16, (3,3), activation='relu', padding='same')(input_img)
x = MaxPooling2D((2,2), padding='same')(x)
x = Conv2D(8,(3,3), activation='relu', padding='same')(x)
x = MaxPooling2D((2,2), padding='same')(x)
x = Conv2D(8, (3,3), activation='relu', padding='same')(x)
encoded = MaxPooling2D((2,2), padding='same')(x)

# 解码器部分
x = Conv2D(8, (3,3), activation='relu', padding='same')(encoded)
x = UpSampling2D((2, 2))(x)
x = Conv2D(8, (3,3), activation='relu', padding='same')(x)
x = UpSampling2D((2, 2))(x)
x = Conv2D(16, (3, 3), activation='relu'， padding='same')(x)
x = UpSampling2D((2, 2))(x)
decoded = Conv2D(1, (3, 3), activation='sigmoid', padding='same')(x)

autoencoder = Model(input_img, decoded)
autoencoder.compile(optimizer='adam', loss='binary_crossentropy')

# 得到编码层的输出
encoder_model = Model(inputs=autoencoder.input, outputs=autoencoder.get_layer('encoder_out').output)

## 导入数据， 使用常用的手写识别数据集
def load_mnist(dataset_name):
'''
load the data
'''
 data_dir = os.path.join("./data", dataset_name)
 f = np.load(os.path.join(data_dir, 'mnist.npz'))
 train_data = f['train'].T
 trX = train_data.reshape((-1, 28, 28, 1)).astype(np.float32)
 trY = f['train_labels'][-1].astype(np.float32)
 test_data = f['test'].T
 teX = test_data.reshape((-1, 28, 28, 1)).astype(np.float32)
 teY = f['test_labels'][-1].astype(np.float32)

# one-hot
 # y_vec = np.zeros((len(y), 10), dtype=np.float32)
 # for i, label in enumerate(y):
 #   y_vec[i, y[i]] = 1
 # keras.utils里带的有one-hot的函数， 就直接用那个了
 return trX / 255., trY, teX/255., teY

# 开始导入数据
x_train, _ , x_test, _= load_mnist('mnist')

# 可视化训练结果， 我们打开终端， 使用tensorboard
# tensorboard --logdir=/tmp/autoencoder # 注意这里是打开一个终端， 在终端里运行

# 训练模型， 并且在callbacks中使用tensorBoard实例， 写入训练日志 http://0.0.0.0:6006
from keras.callbacks import TensorBoard
autoencoder.fit(x_train, x_train,
       epochs=50,
       batch_size=128,
       shuffle=True,
       validation_data=(x_test, x_test),
       callbacks=[TensorBoard(log_dir='/tmp/autoencoder')])

# 重建图片
import matplotlib.pyplot as plt
decoded_imgs = autoencoder.predict(x_test)
encoded_imgs = encoder_model.predict(x_test)
n = 10
plt.figure(figsize=(20, 4))
for i in range(n):
 k = i + 1
 # 画原始图片
 ax = plt.subplot(2, n, k)
 plt.imshow(x_test[k].reshape(28, 28))
 plt.gray()
 ax.get_xaxis().set_visible(False)
 # 画重建图片
 ax = plt.subplot(2, n, k + n)
 plt.imshow(decoded_imgs[i].reshape(28, 28))
 plt.gray()
 ax.get_xaxis().set_visible(False)
 ax.get_yaxis().set_visible(False)
plt.show()

# 编码得到的特征
n = 10
plt.figure(figsize=(20, 8))
for i in range(n):
 k = i + 1
 ax = plt.subplot(1, n, k)
 plt.imshow(encoded[k].reshape(4, 4 * 8).T)
 plt.gray()
 ax.get_xaxis().set_visible(False)
 ax.get_yaxis().set_visible(False)
plt.show()

补充知识：keras搬砖系列-单层卷积自编码器

考试成绩出来了，竟然有一门出奇的差，只是有点意外。

觉得应该不错的，竟然考差了，它估计写了个随机数吧。

头文件

from keras.layers import Input,Dense
from keras.models import Model
from keras.datasets import mnist
import numpy as np
import matplotlib.pyplot as plt

导入数据

(X_train,_),(X_test,_) = mnist.load_data()

X_train = X_train.astype('float32')/255.
X_test = X_test.astype('float32')/255.
X_train = X_train.reshape((len(X_train),-1))
X_test = X_test.reshape((len(X_test),-1))

这里的X_train和X_test的维度分别为(60000L,784L),(10000L,784L)

这里进行了归一化，将所有的数值除上255.

设定编码的维数与输入数据的维数

encoding_dim = 32

input_img = Input(shape=(784,))

构建模型

encoded = Dense(encoding_dim,activation='relu')(input_img)
decoded = Dense(784,activation='relu')(encoded)

autoencoder = Model(inputs = input_img,outputs=decoded)
encoder = Model(inputs=input_img,outputs=encoded)

encoded_input = Input(shape=(encoding_dim,))
decoder_layer = autoencoder.layers[-1]
deconder = Model(inputs=encoded_input,outputs = decoder_layer(encoded_input))

模型编译

autoencoder.compile(optimizer='adadelta',loss='binary_crossentropy')

模型训练

autoencoder.fit(X_train,X_train,epochs=50,batch_size=256,shuffle=True,validation_data=(X_test,X_test))

预测

encoded_imgs = encoder.predict(X_test)

decoded_imgs = deconder.predict(encoded_imgs)

数据可视化

n = 10
for i in range(n):
ax = plt.subplot(2,n,i+1)
plt.imshow(X_test[i].reshape(28,28))
plt.gray()
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)

ax = plt.subplot(2,n,i+1+n)
plt.imshow(decoded_imgs[i].reshape(28,28))
plt.gray()
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)
plt.show()

完成代码

from keras.layers import Input,Dense
from keras.models import Model
from keras.datasets import mnist
import numpy as np
import matplotlib.pyplot as plt

(X_train,_),(X_test,_) = mnist.load_data()

X_train = X_train.astype('float32')/255.
X_test = X_test.astype('float32')/255.
X_train = X_train.reshape((len(X_train),-1))
X_test = X_test.reshape((len(X_test),-1))

encoding_dim = 32
input_img = Input(shape=(784,))

encoded = Dense(encoding_dim,activation='relu')(input_img)
decoded = Dense(784,activation='relu')(encoded)

autoencoder = Model(inputs = input_img,outputs=decoded)
encoder = Model(inputs=input_img,outputs=encoded)

encoded_input = Input(shape=(encoding_dim,))
decoder_layer = autoencoder.layers[-1]
deconder = Model(inputs=encoded_input,outputs = decoder_layer(encoded_input))

autoencoder.compile(optimizer='adadelta',loss='binary_crossentropy')
autoencoder.fit(X_train,X_train,epochs=50,batch_size=256,shuffle=True,validation_data=(X_test,X_test))

encoded_imgs = encoder.predict(X_test)
decoded_imgs = deconder.predict(encoded_imgs)

##via
n = 10
for i in range(n):
ax = plt.subplot(2,n,i+1)
plt.imshow(X_test[i].reshape(28,28))
plt.gray()
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)

ax = plt.subplot(2,n,i+1+n)
plt.imshow(decoded_imgs[i].reshape(28,28))
plt.gray()
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)
plt.show()

来源：https://blog.csdn.net/xuyang508/article/details/74276181