卷积去噪自编码器的实现

import tensorflow as tf import matplotlib.pyplot as plt import numpy as np

查看tensorflow版本

1	tf.__version__

导入mnist数据集

1	(x_train,_),(x_test,_) = tf.keras.datasets.mnist.load_data()

查看数据维度

1	x_train.shape,x_test.shape

对数据最后一个维度进行扩展

1 2	x_train = np.expand_dims(x_train, -1) x_test = np.expand_dims(x_test, -1)

x_train.shape,x_test.shape
1
((60000, 28, 28, 1), (10000, 28, 28, 1))

数据归一化

1 2	x_train = tf.cast(x_train,tf.float32)/255 x_test = tf.cast(x_test,tf.float32)/255

噪声的添加

1
2
3

factor = 0.5 #增加噪声多少
x_train_noise = x_train + factor*np.random.normal(loc=0.0,scale=1.0,size=x_train.shape)
x_test_noise = x_test + factor*np.random.normal(loc=0.0,scale=1.0,size=x_test.shape)

数据范围规定

1 2	x_train_noise = np.clip(x_train_noise,0.,1.)#加噪后数值可能会不在0~1之间将范围规定到之中，大于1变成1小于0变成00 x_test_noise = np.clip(x_test_noise,0.,1.)

查看加噪后的图片

n=10
plt.figure(figsize=(10,2))
for i in range(1,n):
    ax = plt.subplot(1,n,i)
    plt.imshow(x_train_noise[i].reshape(28,28))

构建模型

input = tf.keras.layers.Input(shape=(28, 28, 1))

# Encoder
x = tf.keras.layers.Conv2D(16, (3, 3), activation='relu', padding='same')(input)
x = tf.keras.layers.MaxPool2D((2, 2), padding='same')(x)   # 14*14*16

x = tf.keras.layers.Conv2D(32, (3, 3), activation='relu', padding='same')(x)
x = tf.keras.layers.MaxPool2D((2, 2), padding='same')(x)   # 7*7*32


# Decoder
x = tf.keras.layers.Conv2DTranspose(16, (3, 3), strides=2,
                           activation='relu', padding='same')(x)   # 14*14*16
output = tf.keras.layers.Conv2DTranspose(1, (3, 3), strides=2,
                           activation='sigmoid', padding='same')(x)   # 28*28*1

model = tf.keras.Model(inputs=input, outputs=output)

查看模型数据流动

1	tf.keras.utils.plot_model(model,show_shapes=True)#可视化

模型配置

1	model.compile(optimizer='adam', loss='mse')

模型训练

model.fit(x_train_noisy, x_train,
                epochs=50,
                batch_size=256,
                validation_data=(x_test_noisy, x_test))

模型测试

pre_test = model.predict(x_test_noisy)
plt.figure(figsize=(20, 4))
for i in range(1, n):    
# 展示原始图像
    ax = plt.subplot(2, n, i)
    plt.imshow(x_test_noisy[i].reshape(28, 28))  
    # 展示自编码器重构后的图像
    ax = plt.subplot(2, n, i + n)
    plt.imshow(pre_test[i].reshape(28, 28))
plt.show()