Tensorflow案例实战-11卷积神经网络模型架构(minst10分类任务)

网络模型

前期准备

import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
import input_data
minst=input_data.read_data_sets('data/',one_hot=True)
trainimg=minst.train.images
trainlabel=minst.train.labels
testimg=minst.test.images
testlabel=minst.test.labels

参数设置

n_input=784
n_output=10
weights={
    #filter:w,h 连接的输入的深度 输出
    'wc1':tf.Variable(tf.random_normal([3,3,1,64],stddev=0.1)),
    'wc2':tf.Variable(tf.random_normal([3,3,64,128],stddev=0.1)),
    #2*2变为1*1，即变为原来一半。28/2，两次pool,即为7.全连接层转换成向量所以是1024
    'wd1':tf.Variable(tf.random_normal([7*7*128,1024],stddev=0.1)),
    #前面是1024维向量，后面代表是10分类任务
    'wd2':tf.Variable(tf.random_normal([1024,n_output],stddev=0.1)),
}
biases={
    'bc1':tf.Variable(tf.random_normal([64],stddev=0.1)),
    'bc2':tf.Variable(tf.random_normal([128],stddev=0.1)),
    'bd1':tf.Variable(tf.random_normal([1024],stddev=0.1)),
    'bd2':tf.Variable(tf.random_normal([n_output],stddev=0.1)),
}

函数

def conv_basic(_input,_w,_b,_keepratio):
    #input预处理（转换成四维：batch_size,h,w,channel），-1：tensorflow自己推断
    _input_r=tf.reshape(_input,shape=[-1,28,28,1])
    
    #conv layer 1
    #strides:四个维度的stride
    _conv1=tf.nn.conv2d(_input_r,_w['wc1'],strides=[1,1,1,1],padding='SAME')
    #mean,_var=tf.nn.moments(_conv1,[0,1,2])
    #_conv1=tf.nn.batch_normalization(_conv1,_mean,_var,0,1,0.0001)
    _conv1=tf.nn.relu(tf.nn.bias_add(_conv1,_b['bc1']))
    _pool1=tf.nn.max_pool(_conv1,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME')
    #随机杀死节点，keepratio:保留比例
    _pool_dr1=tf.nn.dropout(_pool1,_keepratio)
    
    #conv layer2
    _conv2=tf.nn.conv2d(_pool_dr1,_w['wc2'],strides=[1,1,1,1],padding='SAME')
    #mean,_var=tf.nn.moments(_conv2,[0,1,2])
    #_conv2=tf.nn.batch_normalization(_conv2,_mean,_var,0,1,0.0001)
    _conv2=tf.nn.relu(tf.nn.bias_add(_conv2,_b['bc2']))
    _pool2=tf.nn.max_pool(_conv2,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME')
    _pool_dr2=tf.nn.dropout(_pool2,_keepratio)    
    
    #vectorize
    _densel=tf.reshape(_pool_dr2,[-1,_w['wd1'].get_shape().as_list()[0]])
    
    #fully connected layer1
    _fc1=tf.nn.relu(tf.add(tf.matmul(_densel,_w['wd1']),_b['bd1']))
    _fc_dr1=tf.nn.dropout(_fc1,_keepratio)
    
    #fully connected layer2
    _out=tf.add(tf.matmul(_fc_dr1,_w['wd2']),_b['bd2'])
    
    #return
    out={'input_r':_input_r,'_conv1':_conv1,'pool1':_pool1,'pool_dr1':_pool_dr1,
         'conv2':_conv2,'pool2':_pool2,'pool_dr2':_pool_dr2,'densel':_densel,
         'fcl':_fc1,'fc_dr1':_fc_dr1,'out':_out
        
    }
    return out   

TensorFlow中CNN的两种padding方式“SAME”和“VALID”：

步长为2，当向右滑动两步之后，VALID方式发现余下的窗口不到2×2 所以直接将第三列舍弃，而SAME方式并不会把多出的一列丢弃，但是只有一列了不够2×2怎么办？填充！

SAME会增加第四列以保证可以达到2×2，但为了不影响原始信息，一般以0来填充。这就不难理解不同的padding方式输出的形状会有所不同了。

x=tf.placeholder(tf.float32,[None,n_input])
y=tf.placeholder(tf.float32,[None,n_output])
keepratio=tf.placeholder(tf.float32)
#function
_pred=conv_basic(x,weights,biases,keepratio)['out']
cost=tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=_pred,labels=y))
optm=tf.train.AdamOptimizer(learning_rate=0.001).minimize(cost)
_corr=tf.equal(tf.argmax(_pred,1),tf.argmax(y,1))
accr=tf.reduce_mean(tf.cast(_corr,tf.float32))
init=tf.global_variables_initializer()

训练

training_epochs=15
batch_size=16
display_step=1
#session
sess=tf.Session()
sess.run(init)
#mini-batch leaning
for epoch in range(training_epochs):
    avg_cost=0
    #num_batch=int(minst.train.num_examples/batch_size)
    num_batch=10
    for i in range(num_batch):
        batch_xs,batch_ys=minst.train.next_batch(batch_size)
        #fit training using batch data
        sess.run(optm,feed_dict={x:batch_xs,y:batch_ys,keepratio:0.7})
        avg_cost+=sess.run(cost,feed_dict={x:batch_xs,y:batch_ys,keepratio:1.})
        avg_cost=avg_cost/num_batch
    #display
    if  epoch % display_step==0:
        print("Epoch:%03d/%03d cost: %.9f"%(epoch,training_epochs,avg_cost))
        feeds_train={x:batch_xs,y:batch_ys,keepratio:1.}
        train_acc=sess.run(accr,feed_dict=feeds_train)
        print("Train accrancy:%.3f"%(train_acc))
        feeds={x:minst.test.images,y:minst.test.labels,keepratio:1.}
        test_acc=sess.run(accr,feed_dict=feeds)
        print("Test accrancy:%.3f"%(test_acc))
print("optimization Done")