Tensorflow——卷积神经网络(CNN)应用于MNIST数据集手写字体识别+Tensorboard展示

一、卷积神经网络详解

参考：https://blog.csdn.net/gaoyu1253401563/article/details/83714865

 

二、Tensorflow实现手写数字识别的卷积神经网络

• 编译环境：jupyter notebook
• CNN设计：输入层-卷积层-池化层-卷积层-池化层-全连接层-全连接层-输出层

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
#载入数据
mnist = input_data.read_data_sets('MNIST_data', one_hot= True)
#设置批次的大小
batch_size = 100
#计算共有多少个批次
n_batch = mnist.train.num_examples // batch_size

#初始化权值
def weight_variable(shape):
    initial = tf.truncated_normal(shape,stddev=0.1)   #生成一个截断的正态分布
    return tf.Variable(initial)

#初始化偏置
def bias_variable(shape):
    initial = tf.constant(0.1,shape=shape)
    return tf.Variable(initial)

#卷积层
def conv2d(x,W):
    #tf.nn.conv2d()是tensorflow中实现卷积的函数
    #x:需要做卷积的输入图像，四维的[batch,in_height,in_width,in_channels]
    #W:filter参数，CNN中的卷积核,形状为[filter_height,filter_width,in_chnnesl,out_chnnels]
    #strides:卷积时在图像每一维上的步长，这是一个一维的向量，长度为4
    #padding:string类型的量，只能是“SAME”(卷积时会在外围补零)、"VALID",（不会在外围补零）这个值决定了不同的卷积方式
    #结果返回一个Tensor，即我们常说的feature map
    return tf.nn.conv2d(x,W,strides=[1,1,1,1],padding='SAME')

#池化层:最大池化
def max_pool_2x2(x):
    return tf.nn.max_pool(x,ksize=[1,2,2,1],strides=[1,2,2,1],padding="SAME")

#定义两个placeholder
x = tf.placeholder(tf.float32,[None,784])
y = tf.placeholder(tf.float32,[None,10])

#改变x的格式转成为4D的向量[batch,in_height,in_width,in_channels]
x_image = tf.reshape(x,[-1,28,28,1])


#初始化第一个卷积层的权值和偏置
#5*5的采样窗口，32个卷积核从1个平面抽取特征
W_conv1 = weight_variable([5,5,1,32])
#每一个卷积核对应医用偏置值，32是偏置值数量
b_conv1 = bias_variable([32])

#把x_image和权值向量进行卷积，再加上偏置值，然后用于ReLU激活函数
h_conv1 = tf.nn.relu(conv2d(x_image,W_conv1) + b_conv1)
#进行max—pooling操作
h_pool1 = max_pool_2x2(h_conv1)

#初始化第二个卷积层的权值和偏重
#5*5的采样窗口，64个卷积核从32个平面上抽取特征
W_conv2 = weight_variable([5,5,32,64])
#每一个卷积核对应一个偏置
b_conv2 = bias_variable([64])

#把h_pool1和权值向量进行卷积，再加上偏置值，然后用于ReLU激活函数
h_conv2 = tf.nn.relu(conv2d(h_pool1,W_conv2) + b_conv2)
#进行max—pooling层操作
h_pool2 = max_pool_2x2(h_conv2)

#28*28的图片第一次卷积后还是28*28，第一次池化后变成14*14
#第二次卷积后为14*14，第二次池化后为7*7
#通过以上操作得到64张7*7的平面

#初始化第一个全连接层的权值:上面有7*7*64个神经元，全连接层有1024个神经元，对应1024个偏置
W_fc1 = weight_variable([7*7*64,1024])
b_fc1 = bias_variable([1024])
#把池化层2的输出扁平化为1维
h_pool2_flat = tf.reshape(h_pool2,[-1,7*7*64])
#求第一个全连接层的输出
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat,W_fc1) + b_fc1)

#keep_prob用来表示神经元的输出概率
keep_prob = tf.placeholder(tf.float32)
#加入dropout操作，目的是使得一部分神经元工作，一部分神经元不工作
h_fc1_drop = tf.nn.dropout(h_fc1,keep_prob)

#初始化第二个全连接层
W_fc2 = weight_variable([1024,10])
b_fc2 = bias_variable([10])

#计算输出层的输出，使用softmax函数计算概率得到输出
prediction = tf.nn.softmax(tf.matmul(h_fc1_drop,W_fc2) + b_fc2)


#使用交叉熵代价函数计算loss
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels = y, logits = prediction))
#使用AdamOptimizer优化器进行优化，使得loss最小
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
#将结果存放在一个布尔列表中,argmax()函数返回一维张量中最大值所在的位置:tf.equal(A,B)函数,若A=B,则返回True，否则返回False
correct_prediction = tf.equal(tf.argmax(prediction,1),tf.argmax(y,1))
#求准确率：tf.cast(x,dtype,name=None),类型转换函数，将x转化为dtype类型
accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))


#定义会话
with tf.Session() as sess:
    #初始化变量
    sess.run(tf.global_variables_initializer())
    #迭代21个周期
    for epoch in range(21):
        for batch in range(n_batch):
            batch_xs,batch_ys = mnist.train.next_batch(batch_size)
            sess.run(train_step,feed_dict = {x:batch_xs, y:batch_ys, keep_prob:0.7})  #每次训练有70%神经元在工作
        acc = sess.run(accuracy,feed_dict = {x:mnist.test.images,y:mnist.test.labels,keep_prob:1.0})
        print("Iter " + str(epoch) + ", Testing Accuracy = " + str(acc))

测试准确度结果：

Extracting MNIST_data/train-images-idx3-ubyte.gz
Extracting MNIST_data/train-labels-idx1-ubyte.gz
Extracting MNIST_data/t10k-images-idx3-ubyte.gz
Extracting MNIST_data/t10k-labels-idx1-ubyte.gz
Iter 0, Testing Accuracy = 0.951
Iter 1, Testing Accuracy = 0.97
Iter 2, Testing Accuracy = 0.9743
Iter 3, Testing Accuracy = 0.981
Iter 4, Testing Accuracy = 0.9838
Iter 5, Testing Accuracy = 0.9861
Iter 6, Testing Accuracy = 0.9866
Iter 7, Testing Accuracy = 0.9873
Iter 8, Testing Accuracy = 0.9859
Iter 9, Testing Accuracy = 0.9868
Iter 10, Testing Accuracy = 0.9883
Iter 11, Testing Accuracy = 0.9888
Iter 12, Testing Accuracy = 0.9886
Iter 13, Testing Accuracy = 0.9898
Iter 14, Testing Accuracy = 0.9896
Iter 15, Testing Accuracy = 0.9901
Iter 16, Testing Accuracy = 0.9909
Iter 17, Testing Accuracy = 0.9897
Iter 18, Testing Accuracy = 0.9908
Iter 19, Testing Accuracy = 0.9913
Iter 20, Testing Accuracy = 0.9906

 

 

三、Tensorflow实现手写数字识别的卷积神经网络——tensorboard展示

• 使用tensorboard代码

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
#载入数据
mnist = input_data.read_data_sets('MNIST_data', one_hot= True)
#设置批次的大小
batch_size = 100
#计算共有多少个批次
n_batch = mnist.train.num_examples // batch_size


#生成tensorboard的参数概要
def variable_summaries(var):
    with tf.name_scope('summaries'):
        mean = tf.reduce_mean(var)
        tf.summary.scalar('mean',mean)
        with tf.name_scope('stddev'):
            stddev = tf.sqrt(tf.reduce_mean(tf.square(var - mean)))
        tf.summary.scalar('stddev',stddev)  #标准差
        tf.summary.scalar('max',tf.reduce_max(var))   #最大值
        tf.summary.scalar('min',tf.reduce_min(var))   #最小值
        tf.summary.histogram('histogram',var)  #直方图



#初始化权值
def weight_variable(shape,name):
    #生成一个截断的正态分布
    initial = tf.truncated_normal(shape,stddev=0.1)
    return tf.Variable(initial,name = name)

#初始化偏置
def bias_variable(shape,name):
    initial = tf.constant(0.1,shape=shape)
    return tf.Variable(initial,name = name)

#卷积层
def conv2d(x,W):
    #tf.nn.conv2d()是tensorflow中实现卷积的函数
    #x:需要做卷积的输入图像，四维的[batch,in_height,in_width,in_channels]
    #W:filter参数，CNN中的卷积核,形状为[filter_height,filter_width,in_chnnesl,out_chnnels]
    #strides:卷积时在图像每一维上的步长，这是一个一维的向量，长度为4
    #padding:string类型的量，只能是“SAME”(卷积时会在外围补零)或"VALID",（不会在外围补零）这个值决定了不同的卷积方式
    #结果返回一个Tensor，即我们常说的feature map
    return tf.nn.conv2d(x,W,strides=[1,1,1,1],padding='SAME')

#池化层:最大池化
def max_pool_2x2(x):
    return tf.nn.max_pool(x,ksize=[1,2,2,1],strides=[1,2,2,1],padding="SAME")



#命名空间
with tf.name_scope('input'):
    #定义两个placeholder
    x = tf.placeholder(tf.float32,[None,784],name = 'x-input')
    y = tf.placeholder(tf.float32,[None,10],name = 'y-input')

with tf.name_scope('x_image'):
    #改变x的格式转成为4D的向量[batch,in_height,in_width,in_channels]
    x_image = tf.reshape(x,[-1,28,28,1], name = 'x_image')

with tf.name_scope('Conv1'):
    with tf.name_scope('W_conv1'):
        #初始化第一个卷积层的权值和偏置
        #5*5的采样窗口，32个卷积核从1个平面抽取特征
        W_conv1 = weight_variable([5,5,1,32],name='W_conv1')
    with tf.name_scope('b_conv1'):
        #每一个卷积核对应医用偏置值，32是偏置值数量
        b_conv1 = bias_variable([32],name = 'b_conv1')

    #把x_image和权值向量进行卷积，再加上偏置值，然后用于ReLU激活函数
    with tf.name_scope('conv2d_1'):
        conv2d_1 = conv2d(x_image,W_conv1) + b_conv1
    with tf.name_scope('relu'):  
        h_conv1 = tf.nn.relu(conv2d_1)
    with tf.name_scope('h_pool1'):
        #进行max—pooling操作
        h_pool1 = max_pool_2x2(h_conv1)

with tf.name_scope('Conv2'):
    #初始化第二个卷积层的权值和偏重
    #5*5的采样窗口，64个卷积核从32个平面上抽取特征
    with tf.name_scope('W_conv2'):
        W_conv2 = weight_variable([5,5,32,64],name='W_conv2')
    with tf.name_scope('b_conv2'):
        #每一个卷积核对应一个偏置
        b_conv2 = bias_variable([64],name = 'b_conv2')

    #把h_pool1和权值向量进行卷积，再加上偏置值，然后用于ReLU激活函数
    with tf.name_scope('conv2d_2'):
        conv2d_2 = conv2d(h_pool1,W_conv2) + b_conv2
    with tf.name_scope('relu'):
        h_conv2 = tf.nn.relu(conv2d_2)
    with tf.name_scope('h_pool2'):
        #进行max—pooling层操作
        h_pool2 = max_pool_2x2(h_conv2)

#28*28的图片第一次卷积后还是28*28，第一次池化后变成14*14
#第二次卷积后为14*14，第二次池化后为7*7
#通过以上操作得到64张7*7的平面

#初始化第一个全连接层的权值:上面有7*7*64个神经元，全连接层有1024个神经元，对应1024个偏置
with tf.name_scope('fc1'):
    with tf.name_scope('W_fc1'):
        W_fc1 = weight_variable([7*7*64,1024],name = 'W_fc1')
    with tf.name_scope('b_fc1'):
        b_fc1 = bias_variable([1024],name = 'b_fc1')
    #把池化层2的输出扁平化为1维
    with tf.name_scope('h_pool2_flat'):
        h_pool2_flat = tf.reshape(h_pool2,[-1,7*7*64],name = 'h_pool2_flat')
    #求第一个全连接层的输出
    with tf.name_scope('wx_plus_b1'):
        wx_plus_b1 = tf.matmul(h_pool2_flat,W_fc1) +b_fc1
    with tf.name_scope('relu'):
        h_fc1 = tf.nn.relu(wx_plus_b1)
    
    
    #keep_prob用来表示神经元的输出概率
    with tf.name_scope('keep_prob'):
        keep_prob = tf.placeholder(tf.float32,name = 'keep_prob')
    #加入dropout操作，目的是使得一部分神经元工作，一部分神经元不工作
    with tf.name_scope('h_fc1_drop'):
        h_fc1_drop = tf.nn.dropout(h_fc1,keep_prob,name = 'h_fc1_drop')

#初始化第二个全连接层   
with tf.name_scope('fc2'):
    with tf.name_scope('W_fc2'):
        W_fc2 = weight_variable([1024,10], name = 'W_fc2')
    with tf.name_scope('b_fc2'):
        b_fc2 = bias_variable([10],name = 'b_fc2')
    with tf.name_scope('wx_plus_b2'):
        wx_plus_b2 = tf.matmul(h_fc1_drop,W_fc2) + b_fc2
    with tf.name_scope('softmax'):
    #计算输出层的输出，使用softmax函数计算概率得到输出
        prediction = tf.nn.softmax(wx_plus_b2)
    
    
#使用交叉熵代价函数计算loss
with tf.name_scope('cross_entropy'):
    cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels = y, logits = prediction),name = 'cross_entropy')
    tf.summary.scalar('cross_entropy',cross_entropy)
    

#使用AdamOptimizer优化器进行优化，使得loss最小
with tf.name_scope('train'):
    train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)


#将结果存放在一个布尔列表中,argmax()函数返回一维张量中最大值所在的位置:tf.equal(A,B)函数,若A=B,则返回True，否则返回False
with tf.name_scope('accuracy'):
    with tf.name_scope('correct_prediction'):
        correct_prediction = tf.equal(tf.argmax(prediction,1),tf.argmax(y,1))
#求准确率：tf.cast(x,dtype,name=None),类型转换函数，将x转化为dtype类型
    with tf.name_scope('accuracy'):
        accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
        tf.summary.scalar('accuracy',accuracy)

#合并所有的summary
merged = tf.summary.merge_all()

#定义会话
with tf.Session() as sess:
    #初始化变量
    sess.run(tf.global_variables_initializer())
    train_writer = tf.summary.FileWriter('tensorboard/logs/train',sess.graph)
    test_writer = tf.summary.FileWriter('tensorboard/logs/test',sess.graph)

    for i in range(1001):
        #训练模型
        batch_xs,batch_ys = mnist.train.next_batch(batch_size)
        sess.run(train_step,feed_dict = {x:batch_xs, y:batch_ys, keep_prob:0.5})  #每次训练有50%神经元在工作
        #记录训练集计算的参数
        summary = sess.run(merged,feed_dict = {x:batch_xs,y:batch_ys,keep_prob:1.0})
        train_writer.add_summary(summary,i)
        #记录测试机计算的参数
        batch_xs,batch_ys = mnist.test.next_batch(batch_size)
        summary = sess.run(merged,feed_dict = {x:batch_xs,y:batch_ys,keep_prob:1.0})
        test_writer.add_summary(summary,i)
        
        if i%100 == 0:
            test_acc = sess.run(accuracy,feed_dict = {x:mnist.test.images,y:mnist.test.labels,keep_prob:1.0})
            train_acc = sess.run(accuracy,feed_dict = {x:mnist.train.images[:10000],y:mnist.train.labels[:10000],keep_prob:1.0})
            print("Iter " + str(i) + ", Testing Accuracy = " + str(test_acc) + ", Training Accuracy = " + str(train_acc))

 

• 如何生成tensorboard网址
  • 在生成的目录tensorboard/logs下运行终端：

tensorboard --logdir=/home/xx（自己的）/tensorflow学习代码实例/tsorboard/logs

• 打开红线网址

如下就是tensorboard

各部件的表示含义：

（1）网络结构的展示：

形式一：

 

形式二：

    拿其中一个展开 

（2）参数的展示 

• accuracy

 

•  cross_entropy