资源简介

有训练代码和测试代码和我已经训练好的模型,还有几张我的测试图片 详情见我的博客:https://blog.csdn.net/qq_38269418/article/details/78991649

资源截图

代码片段和文件信息

from tensorflow.examples.tutorials.mnist import input_data

import tensorflow as tf

mnist = input_data.read_data_sets(‘F:/DEEPLEARN/Anaconda/Lib/site-packages/tensorflow/examples/tutorials/mnist/MNIST_data‘ one_hot=True)

x = tf.placeholder(tf.float32 [None 784])

y_ = tf.placeholder(tf.float32 [None 10])

def weight_variable(shape):
    initial = tf.truncated_normal(shapestddev = 0.1)
    return tf.Variable(initial)

def bias_variable(shape):
    initial = tf.constant(0.1shape = shape)
    return tf.Variable(initial)

def conv2d(xW):
    return tf.nn.conv2d(x W strides = [1111] padding = ‘SAME‘)

def max_pool_2x2(x):
    return tf.nn.max_pool(x ksize=[1221] strides=[1221] padding=‘SAME‘)

W_conv1 = weight_variable([5 5 1 32])
b_conv1 = bias_variable([32])

x_image = tf.reshape(x[-128281])

h_conv1 = tf.nn.relu(conv2d(x_imageW_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)

W_conv2 = weight_variable([5 5 32 64])
b_conv2 = bias_variable([64])

h_conv2 = tf.nn.relu(conv2d(h_pool1 W_conv2) + b_conv2)
h_pool2 = max_pool_2x2(h_conv2)

W_fc1 = weight_variable([7 * 7 * 64 1024])
b_fc1 = bias_variable([1024])

h_pool2_flat = tf.reshape(h_pool2 [-1 7*7*64])
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat W_fc1) + b_fc1)

keep_prob = tf.placeholder(“float“)
h_fc1_drop = tf.nn.dropout(h_fc1 keep_prob)

W_fc2 = weight_variable([1024 10])
b_fc2 = bias_variable([10])

y_conv=tf.nn.softmax(tf.matmul(h_fc1_drop W_fc2) + b_fc2)

cross_entropy = -tf.reduce_sum(y_*tf.log(y_conv))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv1) tf.argmax(y_1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction “float“))

saver = tf.train.Saver() 

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    for i in range(20000):
        batch = mnist.train.next_batch(50)
        if i % 100 == 0:
            train_accuracy = accuracy.eval(feed_dict={
                x: batch[0] y_: batch[1] keep_prob: 1.0})
            print(‘step %d training accuracy %g‘ % (i train_accuracy))
        train_step.run(feed_dict={x: batch[0] y_: batch[1] keep_prob: 0.5})
    saver.save(sess ‘C:/Users/mercheve/Desktop/SAVE/model.ckpt‘)

    print(‘test accuracy %g‘ % accuracy.eval(feed_dict={
        x: mnist.test.images y_: mnist.test.labels keep_prob: 1.0}))


 属性            大小     日期    时间   名称
----------- ---------  ---------- -----  ----

     文件       3602  2018-01-06 16:16  5.png

     文件       3761  2018-01-06 16:21  6.png

     文件       2460  2018-10-22 14:47  mnistdeep.py

     文件       3808  2018-01-06 20:22  test.png

     文件       2950  2018-10-25 14:28  test.py

     文件        195  2018-01-06 11:54  SAVE\checkpoint

     文件   39295616  2018-01-06 11:54  SAVE\model.ckpt.data-00000-of-00001

     文件        914  2018-01-06 11:54  SAVE\model.ckpt.index

     文件      72562  2018-01-06 11:54  SAVE\model.ckpt.meta

     文件       3525  2018-01-06 16:59  4.png

     目录          0  2018-07-11 14:28  SAVE

----------- ---------  ---------- -----  ----

             39389393                    11


评论

共有 条评论