# CNN手写数字识别（改进了保存模型以及预测自己的图片）

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
from PIL import Image
import numpy as np
from PIL import *
import os

# 每个批次的大小
batch_size = 100
# 计算一共有多少个批次
n_batch = mnist.train.num_examples // batch_size

# 初始化权值
def weight_variable(shape):
# 生成一个截断的正态分布
initial = tf.truncated_normal(shape, stddev=0.1)
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):
# x input tensor of shape [batch, in_height, in_width, in_channels]
# W filter / kernel tensor of shape [filter_height, filter_width, in_channels, out_channels]
# in_channels代表输入通道数,out_channels代表输出方向数
# strides[0] = strides[1] = 0
# strides[1]代表x方向步长, strides[2]代表y方向步长
# 'SAME'在外面补零
return tf.nn.conv2d(x, W, strides=[1,1,1,1], padding = 'SAME')

# 池化层
def max_pool_2x2(x):
# ksize [1,x,y,1]

def pre_pic(pic_Name):
img = Image.open(pic_Name)
# 第二个参数表示抗锯齿
reIm = img.resize((28, 28), Image.ANTIALIAS)
# 变为灰度图并转化为矩阵
im_arr = np.array(reIm.convert('L'))
# 注意输入的数字为黑底白字，需要进行转化
# 对图像做二值化处理
threshold = 50
for i in range(28):
for j in range(28):
im_arr[i][j] = 255 - im_arr[i][j]
if im_arr[i][j] < threshold:
im_arr[i][j] = 0
else:
im_arr[i][j] = 255

nm_arr = im_arr.reshape([1, 28 * 28])
return nm_arr

files = ['0.png', '1.png', '2.png', '3.png', '4.png', '5.png', '6.png', '7.png', '8.png', '9.png']
ndarrayImgs = np.zeros((len(files), 784))  # x行784列
index = 0
for file in files:
ndarrayImg =  pre_pic(file)
ndarrayImgs[index] = ndarrayImg
index = index + 1

ndarrayLabels = np.array([[1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
[0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],
[0., 0., 1., 0., 0., 0., 0., 0., 0., 0.],
[0., 0., 0., 1., 0., 0., 0., 0., 0., 0.],
[0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],
[0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],
[0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],
[0., 0., 0., 0., 0., 0., 0., 0., 1., 0.],
[0., 0., 0., 0., 0., 0., 0., 0., 0., 1.]])

# 定义两个占位符
x = tf.placeholder(tf.float32,[None,28*28])
y = tf.placeholder(tf.float32,[None,10])

# 改变x的值为4D向量[batch_size,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
W_conv1 = weight_variable([5,5,1,32]) # 5*5采样窗口，32个卷积核从1个平面提取数据
b_conv1 = bias_variable([32]) # 每个卷积核，一个偏置

# 把x_image和权值向量进行卷积，再加上偏置，然后应用于激活函数
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
# 进行池化操作
h_pool1 = max_pool_2x2(h_conv1)

# 初始化第二个卷积层的全值和偏置
W_conv2 = weight_variable([5,5,32,64])   # 5*5采样窗口，64个卷积核从32个平面抽取特征
b_conv2 = weight_variable([64]) # 每个卷积核，一个偏置

# 把h_pool1和权值向量进行卷积，再加上偏置，然后应用于激活函数
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)
# 进行池化操作
h_pool2 = max_pool_2x2(h_conv2)

# 28*28的图片，第一次卷积还是28*28，第一池化变为14*14
# 第二次卷积还是14*14，第二次池化变为7*7
# 通过上面操作，得到64*7*7的平面

# 初始化第一个全连接层的权值
W_fcl = weight_variable([64*7*7,1024]) # 上一张有7*7*64的输入,1024个神经元
b_fcl = bias_variable([1024])

# 将池化后的图片扁平化为一维
h_pool2_flact = tf.reshape(h_pool2,[-1, 7*7*64])
# 求第一个全连接层的输出
h_fcl = tf.nn.relu(tf.matmul(h_pool2_flact, W_fcl) + b_fcl)

# keep_prob 用来表示神经元的输出概率
keep_prob = tf.placeholder(tf.float32)
h_fcl_drop =tf.nn.dropout(h_fcl, keep_prob)

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

# 计算输出
prediction = tf.nn.softmax(tf.matmul(h_fcl_drop,W_fc2) + b_fc2)

# 交叉熵代价函数
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y, logits=prediction))

# 使用优化器进行优化
correct_prediction = tf.equal(tf.argmax(prediction, 1), tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
# 保存模型
saver = tf.train.Saver()
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
if not os.path.exists('checkpoint'):
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})
if batch % 100 == 0:
print(str(batch) + "/" + str(n_batch))
acc = sess.run(accuracy, feed_dict={x: mnist.test.images, y: mnist.test.labels, keep_prob: 1.0})
print("Iter " + str(epoch) + " Test Accuracy " + str(acc))
# 保存训练好的模型
saver.save(sess, './mnist.ckpt')
else:
# 读取训练好的模型
saver.restore(sess, './mnist.ckpt')
accu = accuracy.eval(feed_dict={x: ndarrayImgs, y: ndarrayLabels, keep_prob: 1.0})
print(accu)
output = sess.run(prediction, feed_dict={x: ndarrayImgs, keep_prob: 1.0})
print('预测值：', output.argmax(axis=1))  # axis：0表示按列，1表示按行
print('实际值：', ndarrayLabels.argmax(axis=1))