Tensorflow手寫識別MNIST之CNN模型

 tensorflow很擅長於搭建神經網絡作圖像識別任務，咱們用MNIST手寫圖像集來作數據源，試驗經典的深度學習CNN模型。

（1）導入mnist數據集，這裏圖像image是[batch_size,784]大小，labels是[batch_size,10]大小

（2）設置X、Y佔位符

（3）搭建圖像卷積池化激活神經網絡

（4）訓練數據訓練模型，獲得loss和acc曲線，評估模型效果。

import warnings
warnings.filterwarnings('ignore')
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
import numpy as np
import os

# 讀取mnist數據，image是[batch_size,784]大小，labels是[batch_size,10]大小
mnist = input_data.read_data_sets('MNIST_data', one_hot=True)
batch_size = 100
X_holder = tf.placeholder(tf.float32,shape=[None, 784])
y_holder = tf.placeholder(tf.float32,shape=[None, 10])
images, labels = mnist.train.next_batch(batch_size)
print("image shape:%s, labels shape:%s"%(images.shape,labels.shape))

#權重在初始化時應該加入少許的噪聲來打破對稱性以及避免0梯度，避免神經元節點輸出恆爲0的問題（dead neurons）
def weight_variable(shape):
  initial = tf.truncated_normal(shape, stddev=0.1)
  return tf.Variable(initial)
 
 
def bias_variable(shape):
  initial = tf.constant(0.1, shape=shape)
  return tf.Variable(initial)
 
 
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')
 
#第一層卷積層，32個卷積核去分別關注32個特徵
W_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
x_image = tf.reshape(X_holder, [-1,28,28,1])#將單張圖片從784維向量從新還原爲28x28的矩陣圖片，-1表示取出全部的數據
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)
#第二層卷積層
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)
#全鏈接層
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)
#使用Dropout，訓練時爲0.5，測試時爲1，keep_prob表示保留不關閉的神經元的比例
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)
#把1024維的向量轉換成10維，對應10個類別
W_fc2 = weight_variable([1024, 10])
b_fc2 = bias_variable([10])
y_conv = tf.matmul(h_fc1_drop, W_fc2) + b_fc2
#交叉熵
loss = tf.reduce_mean(
    tf.nn.softmax_cross_entropy_with_logits(labels=y_holder, logits=y_conv))
#定義train_step
train = tf.train.AdamOptimizer(1e-4).minimize(loss)
#定義測試準確率
correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y_holder,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
 
session = tf.Session()
init = tf.global_variables_initializer()
session.run(init)

iterations = 1000
def  trainme():
	steps = np.zeros(iterations)
	LOSS = np.zeros_like(steps)
	for step in range(iterations):
		train_X, train_Y = mnist.train.next_batch(batch_size)
        _,loss_value,accuracy_value = session.run([train,loss,accuracy], feed_dict={X_holder:train_X, y_holder:train_Y,keep_prob:0.5})
		steps[step]=step
		LOSS[step]=accuracy_value
		if step % 25 == 0:
			print('step:%d accuracy:%.4f, loss:%s' %(step, accuracy_value,loss_value))
	#show plt
	import matplotlib.pyplot as plt
	fig = plt.figure()
	ax = fig.add_subplot(111)
	ax.plot(steps,LOSS,label='loss')
	ax.set_xlabel('step')
	ax.set_ylabel('loss')
	fig.suptitle('MSE')
	handles,labels = ax.get_legend_handles_labels()
	ax.legend(handles,labels=labels)
	plt.show()
        
		
def batchPredict(batch_size):
    test_X,test_Y = mnist.test.next_batch(batch_size)
    predict_labels = session.run(predict_y, feed_dict={X_holder:test_X, y_holder:test_Y})
    image_number = test_X.shape[0]
      
    for index in range(image_number):
        if index < image_number:
            image = test_X[index]
            actual = np.argmax(test_Y[index])
            predict = np.argmax(predict_labels[index])
            isTrue = actual==predict
            title = 'actual:%d ,predict:%d' %(actual,predict)
            if not isTrue:
                print(title)
                print(predict_labels[index])

trainme()