卷积神经网络CNN学习记录-实战篇（Minst手写数据集识别）

from tensorflow.examples.tutorials.mnist import input_data
import tensorflow as tf

minst = input_data.read_data_sets("data/",one_hot="True")
sess = tf.InteractiveSession()

def weight_var(shape):
    initial = tf.truncated_normal(shape,stddev=0.1)
    return tf.Variable(initial)

def bias_var(shape):
    initial =tf.constant(0.1, shpae = shape)
    return tf.Variable(initial)

def conv2d(x,W):
    return tf.nn.conv2d(x,W,strides=[1,1,1,1],padding = "SAME")

def max_pooling(x):
    return tf.nn.max_pool(x,ksize=[1,2,2,1],strides = [1,2,2,1],padding = "SAME")

x = tf.placeholder(tf.float32,[None,784])
y_ = tf.placeholder(tf.float32,[None,10])
x_imag = tf.reshape(x,[-1,28,28,1])

#第一个卷积块
W_conv1 = weight_var([5,5,1,32])
b_conv1 = bias_var((32))
h_conv1 = tf.nn.relu(conv2d(x_imag,W_conv1)+b_conv1)
h_pool1 = max_pooling(h_conv1)

#第二个卷积块
W_conv2 = weight_var([5,5,1,64])
b_conv2 = bias_var([64])
h_conv2 = tf.nn.relu(conv2d(h_pool1,W_conv2)+b_conv2)
h_pool2 = max_pooling(h_conv2)

#全连接层
W_fc1 = weight_var([7*7*64,1024])
b_fc1 = bias_var([1024])
h_flat = tf.reshape(h_pool2,[-1,7*7*64])
h_fc1 = tf.nn.relu(tf.matmul(h_flat,W_fc1)+b_fc1)

keep_prob = tf.placeholder(tf.float32)
h_drop = tf.nn.dropout(h_fc1,keep_prob)

W_fc2 = weight_var([1024,10])
b_fc2 = bias_var([10])
y_conv = tf.nn.softmax(tf.matmul(h_drop,W_fc2)+b_fc2)

cross_ent = tf.reduce_mean(tf.reduce_sum(y_*tf.log(y_conv),reduction_indices=[1]))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_ent)

correct_pre = tf.equal(tf.arg_max(y_conv,1),tf.arg_max(y_,1))
acc = tf.reduce_mean(tf.cast(correct_pre),tf.float32)

tf.global_variables_initializer().run()
for i in range(20000):
    batch = minst.train.next_batch(50)
    if i%100 == 0:
        train_acc = acc.eval(feed_dict={x:batch[0],y_:batch[1],keep_prob:1.0})
        print("step %d,train acc %g"%(i,train_acc))
    train_step.run(feed_dict={x:batch[0],y_:batch[1],keep_prob:0.5})

print("test acc %g"%acc.eval(feed_dict={x:minst.test.images,y_:minst.test.labels,keep_prob:1.0}))
           

经测试，识别准确率达99%以上