基于python实现的手写数字识别

【实例截图】

│  8.png
│  checkpoint
│  demo01_ce.py
│  demo02_tfdemo.py
│  demo03_mnist.py
│  demo04_mnist_test.py
│  mnist_model.ckpt.data-00000-of-00001
│  mnist_model.ckpt.index
│  mnist_model.ckpt.meta
│  神经网络分类模型_初始.png
│  神经网络分类模型_结果.png
│
└─MNIST_data
        t10k-images-idx3-ubyte.gz
        t10k-images.idx3-ubyte
        t10k-labels-idx1-ubyte.gz
        t10k-labels.idx1-ubyte
        train-images-idx3-ubyte.gz
        train-images.idx3-ubyte
        train-labels-idx1-ubyte.gz
        train-labels.idx1-ubyte

【核心代码】

import tensorflow as tf
import cv2 as cv
import numpy as np

#生成权重
def weight_variable(shape):
  initial = tf.random_normal(shape, stddev=0.1)
  return tf.Variable(initial)

#生成b
def bias_variable(shape):
  initial = tf.constant(0.1, shape=shape)
  return tf.Variable(initial)

#卷积层
def conv2d(x, W):
  return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')

def max_pool_2x2(x):
  return tf.nn.max_pool(x, ksize=[1, 2, 2, 1],
                        strides=[1, 2, 2, 1], padding='SAME')

x = tf.placeholder("float", shape=[None, 784])
y_ = tf.placeholder("float", shape=[None, 10])

x_image = tf.reshape(x, [-1,28,28,1])

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

h_conv1 = tf.nn.relu(conv2d(x_image, W_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)

h_pool2_flat = tf.reshape(h_pool2, [-1, 7*7*64])

W_fc1 = weight_variable([7 * 7 * 64, 1024])
b_fc1 = bias_variable([1024])
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1)   b_fc1)

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

y_conv=tf.nn.softmax(tf.matmul(h_fc1, W_fc2)   b_fc2)

saver = tf.train.Saver()

with tf.Session() as sess:
  # 加载模型参数
  saver.restore(sess, 'mnist_model.ckpt')
  print('Variables restored.')
  # 读入测试图片
  original = cv.imread('8.png')
  # 图片转灰度
  gray = cv.cvtColor(original, cv.COLOR_BGR2GRAY)
  # 把灰度小于等于10的=0，大于10的=1（纯 黑色像素 值为 0， 纯 白色像素值为 1）
  gray[gray<=10] = 0
  gray[gray>10] = 1
  # 修改灰度图片的像素改为28x28， 并且转为二维数组
  gray = cv.resize(gray, (28, 28)).reshape(1, 28*28)
  # 喂数据（把gray代入方程获取测试结果）
  pred_y = sess.run(y_conv, feed_dict={x:gray})
  # 取预测结果的最大值的索引就是真实预测结果
  print(pred_y.argmax())

  # from tensorflow.examples.tutorials.mnist import input_data
  # import tensorflow as tf
  # mnist = input_data.read_data_sets('MNIST_data', one_hot=True)
  # batch = mnist.test.next_batch(5000)
  # test_x = batch[0]
  # test_y = batch[1]
  # pred_y = sess.run(y_conv, feed_dict={x:test_x})
  # for pred, label in zip(
  #   np.argmax(pred_y, axis=1),
  #   np.argmax(test_y, axis=1)):
  #   print(pred, '<-', label)
  #
  # print((np.argmax(pred_y, axis=1) == np.argmax(test_y, axis=1)).sum() / 50)