TensorFlow-mnist

训练代码：

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

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

flags = tf.app.flags
FLAGS = flags.FLAGS
flags.DEFINE_string('data_dir', '/tmp/data/', 'Directory for storing data')

print(FLAGS.data_dir)
mnist = input_data.read_data_sets(FLAGS.data_dir, one_hot=True)

input=tf.placeholder(tf.float32,[None,784],name='input')
label=tf.placeholder(tf.float32,[None,10],name='label')
keep_prob=tf.placeholder(tf.float32,name='keep_prob')

image=tf.reshape(input,[-1,28,28,1])

conv1_W=tf.Variable(tf.truncated_normal([5,5,1,32],stddev=0.1))
conv1_b=tf.Variable(tf.constant(0.1,shape=[32]))
layer1=tf.nn.elu(tf.nn.conv2d(image,conv1_W,strides=[1,1,1,1],padding='SAME')+conv1_b)
layer2=tf.nn.max_pool(layer1,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME')

conv2_W=tf.Variable(tf.truncated_normal([5,5,32,64],stddev=0.1))
conv2_b=tf.Variable(tf.constant(0.1,shape=[64]))
layer3=tf.nn.elu(tf.nn.conv2d(layer2,conv2_W,strides=[1,1,1,1],padding='SAME')+conv2_b)
layer4=tf.nn.max_pool(layer3,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME')

layer5=tf.reshape(layer4,[-1,7*7*64])

fc1_W=tf.Variable(tf.truncated_normal([7*7*64,1024],stddev=0.1))
fc1_b=tf.Variable(tf.constant(0.1,shape=[1024]))
layer5=tf.reshape(layer4,[-1,7*7*64])
layer6=tf.nn.elu(tf.matmul(layer5,fc1_W)+fc1_b)

layer7=tf.nn.dropout(layer6,keep_prob)

fc2_W=tf.Variable(tf.truncated_normal([1024,10],stddev=0.1))
fc2_b=tf.Variable(tf.constant(0.1,shape=[10]))
output=tf.nn.softmax(tf.matmul(layer7,fc2_W)+fc2_b,name='output')

cross_entropy=tf.reduce_mean(-tf.reduce_sum(label*tf.log(output),reduction_indices=[1]))

train_step=tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_predition=tf.equal(tf.argmax(output,1),tf.arg_max(label,1))
accuracy=tf.reduce_mean(tf.cast(correct_predition,tf.float32),name='accuracy')

sess = tf.InteractiveSession()
sess.run(tf.global_variables_initializer())
batch = mnist.train.next_batch(50)
for i in range(20000):
    batch = mnist.train.next_batch(50)
    train_step.run(feed_dict={input: batch[0], label: batch[1], keep_prob: 0.5})
    if i%100==0:
        train_accuracy = accuracy.eval(feed_dict={input:batch[0], label:batch[1], keep_prob: 1.0})
        print("%d:training accuracy %g"%(i,train_accuracy))

saver = tf.train.Saver()
save_path = saver.save(sess,"E:/dnn/model")

测试代码：

from __future__ import division
import numpy as np
import tensorflow as tf
from PIL import Image

img = Image.open('E:/dnn/test.bmp').convert('L')
if img.size[0] != 28 or img.size[1] != 28:
    img = img.resize((28, 28))
arr = []
for i in range(28):
    for j in range(28):
        pixel = 1.0 - float(img.getpixel((j, i)))/255.0
        arr.append(pixel)
image = np.array(arr).reshape((1, 28, 28, 1))

saver = tf.train.import_meta_graph('E:/dnn/model.meta')
graph = tf.get_default_graph()
input=graph.get_tensor_by_name('input:0')
label=graph.get_tensor_by_name('label:0')
output=graph.get_tensor_by_name('output:0')
keep_prob=graph.get_tensor_by_name('keep_prob:0')
accuracy=graph.get_tensor_by_name('accuracy:0')

with tf.Session() as sess:
    saver.restore(sess, tf.train.latest_checkpoint('E:/dnn'))
    test = sess.run(output, feed_dict={input: image.reshape(-1,784), label: np.full(10,1e-10).reshape(-1,10), keep_prob: 1.0})
    print(test)
    ans=0
    for i in range(10):
        if (test[0][i]>test[0][ans]):
            ans=i
    print(ans)

测试结果：