tensorflow1.12.1实现MNIST数据集训练和识别
1.mnist_train.py
# -*- coding: utf-8 -*- """ Created on Mon Dec 23 20:01:39 2019 tensorflow实现Lenet-5网络，mnist_train.py实现lenet-5训练过程 @author: zhaoy """ ##lenet-5训练过程 import os import tensorflow as tf from tensorflow.examples.tutorials.mnist import input_data import numpy as np import mnist_inference ##配置神经网络的参数 BATCH_SIZE=100 LEARNING_RATE_BASE=0.01 #基础学习率 LEARNING_RATE_DECAY=0.99 REGULARAZTION_RATE=0.0001 TRAINING_STEPS=30000 MOVING_AVERAGE_DECAY=0.99 ##模型保存的路径和文件名 MODEL_SAVE_PATH="./model/" MODEL_NAME="model.ckpt" ##定义训练过程 def train(mnist): #区别与全连接神经网络的输入是一个二维[None, mnist_inference.INPUT_NODE]， #卷积神经网络的输入x是一个四维数组 x=tf.placeholder(tf.float32,[ BATCH_SIZE, # 第一维表示一个batch中样例的个数 mnist_inference.IMAGE_SIZE, # 第二维和第三维表示图片的尺寸28 mnist_inference.IMAGE_SIZE, mnist_inference.NUM_CHANNELS], #图像通道数，黑白图像赋值1，彩色图像赋值3 name='x-input') y_ = tf.placeholder(tf.float32, [None, mnist_inference.OUTPUT_NODE] , name='y-input') regularizer=tf.contrib.layers.l2_regularizer(REGULARAZTION_RATE) y=mnist_inference.inference(x,True,regularizer) global_step=tf.Variable(0,trainable=False) #给定滑动平均衰减率和训练轮数的变量，初始化滑动平均类 variable_averages=tf.train.ExponentialMovingAverage(MOVING_AVERAGE_DECAY,global_step) #在所有代表神经网络参数的变量上使用滑动平均。 variables_averages_op=variable_averages.apply(tf.trainable_variables()) #计算交叉熵作为刻画预测值和真实值之间差距的损失函数 cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=y, labels=tf.argmax(y_,1)) #计算在当前batch中所有样例的交叉熵平均值 cross_entropy_mean=tf.reduce_mean(cross_entropy) #计算L2正则化损失函数 #regularizer=tf.contrib.layers.l2_regularizer(REGULARIZATION_RATE) #计算模型的正则化损失 #regularization=regularizer(weights1)+regularizer(weights2) #总损失等于交叉熵损失和正则化损失的和 loss=cross_entropy_mean+tf.add_n(tf.get_collection('losses')) #regularization #设置指数衰减的学习率 learning_rate = tf.train.exponential_decay(LEARNING_RATE_BASE, global_step, mnist.train.num_examples/BATCH_SIZE, LEARNING_RATE_DECAY, staircase=True) train_step=tf.train.GradientDescentOptimizer(learning_rate).minimize(loss,global_step=global_step) #with tf.control_dependencies([train_step,variables_averages_op]):train_op=tf.no_op(name='train') train_op=tf.group(train_step, variables_averages_op) #初始化tensorflow持久化类 saver=tf.train.Saver() ##初始化会话并开始训练过程 with tf.Session() as sess: tf.global_variables_initializer().run() print("****************开始训练************************") # validate_feed={x:mnist.validation.images,y_:mnist.validation.labels} #准备测试数据. #test_feed={x:mnist.test.images,y_:mnist.test.labels} #迭代地训练神经网络 for i in range(TRAINING_STEPS): xs,ys=mnist.train.next_batch(BATCH_SIZE) #区别于全连接神经网络，卷积神经网络的输入为四维数组 reshaped_xs = np.reshape(xs, (BATCH_SIZE, mnist_inference.IMAGE_SIZE, mnist_inference.IMAGE_SIZE, mnist_inference.NUM_CHANNELS)) train_op_renew,loss_value, step=sess.run([train_op,loss,global_step], feed_dict={x:reshaped_xs,y_:ys}) if i%1000==0: print("After %d training step(s),loss on training batch is %g."%(step,loss_value)) saver.save(sess,os.path.join(MODEL_SAVE_PATH,MODEL_NAME),global_step=global_step) def main(argv=None): mnist=input_data.read_data_sets("MNIST_data/",one_hot=True) train(mnist) if __name__=='__main__': tf.app.run()
2.mnist_inference.py
# -*- coding: utf-8 -*- """ Created on Mon Dec 23 19:44:19 2019 Tensorflow实现LeNet-5模型,mnist_inference.py实现神经卷积网络前向传播过程 @author: zhaoy """ ##实现神经卷积网络的前向传播过程 import tensorflow as tf INPUT_NODE=784 OUTPUT_NODE=10 IMAGE_SIZE=28 NUM_CHANNELS=1 NUM_LABELS=10 #第一层卷积层的尺寸和深度 CONV1_DEEP=32 CONV1_SIZE=5 #第二层卷积层的尺寸和深度 CONV2_DEEP=64 CONV2_SIZE=5 #全连接层的节点个数 FC_SISE=512 tf.reset_default_graph() #定义卷积神经网络的前向传播过程，这里添加了一个新的参数train用以区分训练过程和测试过程；train为布尔量： #采用dropout机制防止过拟合，在全连接层layer5-fc1中包含dropout机制， def inference(input_tensor,train,regularizer): ##声明第一层卷积层的变量并实现前向传播过程，这一层的输入为28*28*1的矩阵，由于采取全0填充，所以输出层为28*28*32的矩阵 with tf.variable_scope('layer1-conv1'): conv1_weights=tf.get_variable("weights",[CONV1_SIZE,CONV1_SIZE,NUM_CHANNELS,CONV1_DEEP], initializer=tf.truncated_normal_initializer(stddev=0.1)) conv1_biases=tf.get_variable("bias",[CONV1_DEEP],initializer=tf.constant_initializer(0.0)) #使用边长为5，深度为32的过滤器，过滤器移动的步长为1，且使用全0填充. conv1=tf.nn.conv2d(input_tensor,conv1_weights,strides=[1,1,1,1],padding='SAME') relu1=tf.nn.relu(tf.nn.bias_add(conv1,conv1_biases)) ##实现第二层（池化层的前向传播过程），这里选用最大池化层，池化层的过滤器的边长为2，所以输出为14*14*32 with tf.name_scope('layer2-poll'): pool1=tf.nn.max_pool(relu1,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME') ##声明第三层卷积层的变量并实现前向传播过程，这一层的输入为14*14*32的矩阵，输出为14*14*64 with tf.variable_scope('layer3-conv2'): conv2_weights=tf.get_variable("weights",[CONV2_SIZE,CONV2_SIZE,CONV1_DEEP,CONV2_DEEP], initializer=tf.truncated_normal_initializer(stddev=0.1)) conv2_biases=tf.get_variable("bias",[CONV2_DEEP],initializer=tf.constant_initializer(0.0)) #使用边长为5，深度为64的过滤器，过滤器移动的步长为1，且使用全0填充. conv2=tf.nn.conv2d(pool1,conv2_weights,strides=[1,1,1,1],padding='SAME') relu2=tf.nn.relu(tf.nn.bias_add(conv2,conv2_biases)) ##实现第四层池化层的前向传播过程,输入为14*14*64的矩阵，输出为7*7*64的矩阵 with tf.name_scope('layer4-pool2'): pool2=tf.nn.max_pool(relu2,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME') #第四层的输出为7*7*64，然而第五层全连接层需要的输入格式为向量，所以这里需要将7*7*64拉伸为一个向量。 #pool2.get_shape。因为每层网络的输入输出都是一batch，矩阵所以这里的维度也包含batch中数据的个数 pool_shape=pool2.get_shape().as_list() nodes=pool_shape[1]*pool_shape[2]*pool_shape[3] #通过tf.reshape函数将第四层的输出编程一个batch的向量 reshaped=tf.reshape(pool2,[pool_shape[0],nodes]) #声明第五层全连接层的变量并实现前向传播过程，这一层是一个拉直后的一组向量，向量长度为3136，输出为为一组长度为512的向量 #引入dropout机制，droupout在训练时会随机将部分节点输出为0，即使部分节点“死掉” with tf.variable_scope('layer5-fc1'): fc1_weights=tf.get_variable("weight",[nodes,FC_SISE], initializer=tf.truncated_normal_initializer(stddev=0.1)) if regularizer!=None: tf.add_to_collection('losses',regularizer(fc1_weights)) fc1_biases=tf.get_variable("bias",[FC_SISE],initializer=tf.constant_initializer(0.1)) fc1=tf.nn.relu(tf.matmul(reshaped,fc1_weights)+fc1_biases) #duopout系数为0.5 if train:fc1=tf.nn.dropout(fc1,0.5) ##声明第六层全连接层的变量并实现前向传播过程，这一层输入为长度为512的向量，输出为一组长度10的向量 with tf.variable_scope('layer6-fc2'): fc2_weights=tf.get_variable("weight",[FC_SISE,NUM_LABELS], initializer=tf.truncated_normal_initializer(stddev=0.1)) if regularizer!=None: tf.add_to_collection('losses',regularizer(fc2_weights)) fc2_biases=tf.get_variable("bias",[NUM_LABELS], initializer=tf.constant_initializer(0.1)) logit=tf.matmul(fc1,fc2_weights)+fc2_biases ##返回第六层的输出 return logit
3.mnist_test.py
# -*- coding: utf-8 -*- """ Created on Mon Dec 23 17:28:07 2019 tensorflow实现神经网络对mnist手写字的识别，mnist_test.py评估训练的神经网络在mnist测试集的正确率 @author: zhaoy """ import tensorflow as tf from tensorflow.examples.tutorials.mnist import input_data # 加载mnist_inference.py 和mnist_train.py中定义的常量和函数。 import mnist_inference import mnist_train def evaluate(mnist): with tf.Graph().as_default() as g: # 定义输入输出的格式。 x = tf.placeholder(tf.float32, [None, mnist_inference.INPUT_NODE], name='x-input') y_ = tf.placeholder(tf.float32, [None, mnist_inference.OUTPUT_NODE], name='y-input') test_feed = {x: mnist.test.images, y_: mnist.test.labels} # 直接通过调用封装好的函数来计算前向传播的结果。因为测试时不关注ze正则化损失的值 # 所以这里用于计算正则化损失的函数被设置为None。 y = mnist_inference.inference(x, None) # 使用前向传播的结果计算正确率。如果需要对未知的样例进行分类，那么使用 # tf.argmax(y,1)就可以得到输入样例的预测类别了。 correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1)) accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) # 同训练模型一样，定义一个滑动平均类 variable_averages = tf.train.ExponentialMovingAverage( mnist_train.MOVING_AVERAGE_DECAY ) #在使用滑动平均进行模型训练时，模型除了保存网络参数以外，还会保存相应的滑动平均参数， #此时加载模型参数需要联通滑动参数一起加载模型时，需要用到.variables_to_restore variables_to_restore = variable_averages.variables_to_restore() saver = tf.train.Saver(variables_to_restore) with tf.Session() as sess: # tf.train.get_checkpoint_state函数会通过checkpoint文件自动 # 找到目录中最新模型的文件名。 ckpt = tf.train.get_checkpoint_state(mnist_train.MODEL_SAVE_PATH) if ckpt and ckpt.model_checkpoint_path: # 加载模型。 saver.restore(sess, ckpt.model_checkpoint_path) # 通过文件名得到模型保存时迭代的轮数。 #split函数：拆分字符串。通过指定分隔符对字符串进行切片，并返回分割后的字符串列表（list）。 #split函数返回值为：分割后的字符串列表。 #list[n]：即表示选取第n个分片,n为-1即为末尾倒数第一个分片（分片即为在返回值列表中元素） global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1] accuracy_score = sess.run(accuracy, feed_dict=test_feed) print("After %s training step(s), test " "accuracy = %g" % (global_step, accuracy_score)) else: print("No checkpoint file found") return def main(argv=None): mnist = input_data.read_data_sets("C:/Users/zhaoy/Desktop/Tensorflow/sample/data/MNIST", one_hot=True) #one_hot表示0-1编码 evaluate(mnist) if __name__ == "__main__": tf.app.run()
4.predict.py
# -*- coding: utf-8 -*- """ Created on Fri Dec 27 17:35:39 2019 @author: zhaoy """ import tensorflow as tf # 加载mnist_inference.py 和mnist_train.py中定义的常量和函数。 import mnist_inference import mnist_train import cv2 def imageprepare(file_name): im = cv2.imread(file_name,0) print(file_name) pixels = [] h, w = im.shape #normalize pixels to 0 and 1. 0 is pure white, 1 is pure black. for i in range(h): for j in range(w): #如果是白底黑字则为：pixels.append((255-im[i,j])*1.0/255.0) pixels.append(im[i, j]*1.0/255.0) #print(pixels) return pixels # ============================================================================= # def imageprepare(file_name): # image = tf.gfile.FastGFile(file_name, 'rb').read() # print(file_name) # pixels = [] # image_data = tf.image.decode_jpeg(image) # image_data = tf.image.convert_image_dtype(image_data, dtype=tf.float32) # # pixels.append(image_data) # return pixels # ============================================================================= x = tf.placeholder(tf.float32, [None, mnist_inference.INPUT_NODE], name='x-input') y_ = tf.placeholder(tf.float32, [None, mnist_inference.OUTPUT_NODE], name='y-input') y = mnist_inference.inference(x, None) variable_averages = tf.train.ExponentialMovingAverage(mnist_train.MOVING_AVERAGE_DECAY) variables_to_restore = variable_averages.variables_to_restore() saver = tf.train.Saver(variables_to_restore) prediction = tf.argmax(y, 1) probability = tf.nn.softmax(y) with tf.Session() as sess: result = imageprepare('4.jpg') tf.global_variables_initializer().run() ckpt = tf.train.get_checkpoint_state(mnist_train.MODEL_SAVE_PATH) if ckpt and ckpt.model_checkpoint_path: # 加载模型。 saver.restore(sess, ckpt.model_checkpoint_path) pre, prob = sess.run([prediction, probability], feed_dict={x:[result]}) #pre = prediction.eval(feed_dict={x:[result]}, session=sess) #global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1] #accuracy_score = sess.run(accuracy, feed_dict=test_feed) print( "recognize result = %d," "the probability is %g " % (pre[0],prob[0][pre])) else: print("No checkpoint file found")
