def test():
with tf.Graph().as_default() as g:
x = tf.placeholder(tf.float32, [
BATCH_SIZE, cifar10_lenet5_forward.IMAGE_SIZE,
cifar10_lenet5_forward.IMAGE_SIZE,
cifar10_lenet5_forward.NUM_CHANNELS
])
y_ = tf.placeholder(tf.float32,
[None, cifar10_lenet5_forward.OUTPUT_NODE])
y = cifar10_lenet5_forward.forward(x, False, None)
ema = tf.train.ExponentialMovingAverage(
cifar10_lenet5_backward.MOVING_AVERAGE_DECAY)
ema_restore = ema.variables_to_restore()
saver = tf.train.Saver(ema_restore)
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
img_batch, label_batch = cifar10_lenet5_generateds.get_tfrecord(
TEST_NUM, isTrain=False)
while True:
with tf.Session() as sess:
ckpt = tf.train.get_checkpoint_state(
cifar10_lenet5_backward.MODEL_SAVE_PATH)
if ckpt and ckpt.model_checkpoint_path:
saver.restore(sess, ckpt.model_checkpoint_path)
global_step = ckpt.model_checkpoint_path.split(
'/')[-1].split('-')[-1]
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess,
coord=coord)
xs, ys = sess.run([img_batch, label_batch])
reshaped_xs = np.reshape(
xs, (BATCH_SIZE, cifar10_lenet5_forward.IMAGE_SIZE,
cifar10_lenet5_forward.IMAGE_SIZE,
cifar10_lenet5_forward.NUM_CHANNELS))
accuracy_score = sess.run(accuracy,
feed_dict={
x: reshaped_xs,
y_: ys
})
print("after %s training step(s), test accuracy = %g" %
(global_step, accuracy_score))
coord.request_stop()
coord.join(threads)
else:
print("No checkpoint file found")
return
time.sleep(INTERVAL_TIME)
def backward(): #执行反向传播，训练参数w
x = tf.placeholder(tf.float32, [None, 32,32,3])
y_ = tf.placeholder(tf.float32, [None]) #
y, end_points = nets.resnet_v2.resnet_v2_50(x, num_classes=10, is_training=True)
y = tf.reshape(y, shape=[-1, 10]) #预测值，相当于y
global_step = tf.Variable(0, trainable = False)
# 将label值进行onehot编码,10分类
one_hot_labels = tf.one_hot(indices=tf.cast(y_, tf.int32), depth=10)
# 定义损失函数和优化器
loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=y, labels=one_hot_labels))
learning_rate = tf.train.exponential_decay( # 设置指数衰减的学习率
LEARNING_RATE_BASE, # 基础学习率，随着迭代的进行，更新变量时使用的学习率在此基础上递减
global_step, # 当前迭代轮数
train_num_examples / BATCH_SIZE, # 过完所有训练数据所需迭代次数
LEARNING_RATE_DECAY, # 指数学习衰减率
staircase = True)
train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss,global_step = global_step)# 使用梯度下降优化损失函数，损失函数包含了交叉熵和正则化损失
ema = tf.train.ExponentialMovingAverage(MOVING_AVERAGE_DECAY, global_step) # 初始化滑动平均类
ema_op = ema.apply(tf.trainable_variables()) # 对所有表示神经网络参数的变量进行滑动平均
#正确率计算
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(one_hot_labels, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
with tf.control_dependencies([train_step, ema_op]): # 使用tf.control_dependencies机制一次完成多个操作。在此神经网络模型中，每过一遍数据既通过
train_op = tf.no_op(name = 'train') # 反向传播更新参数，又更新了每一个参数的滑动平均值
saver = tf.train.Saver()
img_batch, label_batch = cifar10_lenet5_generateds.get_tfrecord(BATCH_SIZE, isTrain=True)#img_batch = [[32,32,3],[32,32,3]], label_batch=[[0,0,0,1,0...], [0,1,0....]]
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
ckpt = tf.train.get_checkpoint_state("./model") # 从"./model"中加载训练好的模型
if ckpt and ckpt.model_checkpoint_path: # 若ckpt和保存的模型在指定路径中存在，则将保存的神经网络模型加载到当前会话中
saver.restore(sess, ckpt.model_checkpoint_path)
coord = tf.train.Coordinator() # 创建一个协调器，管理线程
threads = tf.train.start_queue_runners(sess=sess, coord=coord) # 启动QueueRunner,此时文件名队列已经进队
for i in range(STEPS):
b_image, b_label = sess.run([img_batch, label_batch])
_, step, accuracy_, loss_,= sess.run([train_op, global_step, accuracy, loss], feed_dict={x: b_image,
y_: b_label})
if i % 1000 == 0:
print("after %d training step(s),accuracy : %g, loss is %g." % (step,accuracy_, loss_))
saver.save(sess, os.path.join(MODEL_SAVE_PATH, MODEL_NAME),global_step = global_step) # 保存当前模型，globle_step参数可以使每个被保存模型的文件名末尾都加上训练的轮数
coord.request_stop()
coord.join(threads)
def test():
with tf.Graph().as_default() as g: #复现之前定义的计算图，并执行以下操作
x = tf.placeholder(
tf.float32,
[ #定义占位符x，以之代替输入图片
BATCH_SIZE, cifar10_lenet5_forward.IMAGE_SIZE,
cifar10_lenet5_forward.IMAGE_SIZE,
cifar10_lenet5_forward.NUM_CHANNELS
])
y_ = tf.placeholder(
tf.float32,
[None, cifar10_lenet5_forward.OUTPUT_NODE]) #定义占位符y_，用来接神经元计算结果
y = cifar10_lenet5_forward.forward(x, False, None) #y是神经元的计算结果，下一步喂给y_
ema = tf.train.ExponentialMovingAverage(
cifar10_lenet5_backward.MOVING_AVERAGE_DECAY
) # 实现滑动平均模型，参数MOVING_AVERAGE_DECAY用于控制模型更新的速度，训练过程中会对每一个变量维护一个影子变量
ema_restore = ema.variables_to_restore(
) # variable_to_restore()返回dict ({ema_variables : variables})，字典中保存变量的影子值和现值
saver = tf.train.Saver(
ema_restore) # 创建可还原滑动平均值的对象saver，测试时使用w的影子值，有更好的适配性
correct_prediction = tf.equal(
tf.argmax(y, 1), tf.argmax(y_, 1)
) # 比较预测值和标准输出得到correct_prediction，if tf.argmax(y, 1) equals to tf.argmax(y_, 1),correct_prediction will be set True
accuracy = tf.reduce_mean(
tf.cast(correct_prediction, tf.float32)
) # 将correct_prediction的值从boolean型转为tf.float32型，求均值，得出预测准确率
img_batch, label_batch = cifar10_lenet5_generateds.get_tfrecord(
TEST_NUM, isTrain=False) #2 一次批获取 TEST_NUM 张图片和标签
while True:
with tf.Session() as sess:
ckpt = tf.train.get_checkpoint_state(
cifar10_lenet5_backward.MODEL_SAVE_PATH) # 从指定路径中，加载训练好的模型
if ckpt and ckpt.model_checkpoint_path: # 若已有ckpt模型则执行以下恢复操作
saver.restore(sess,
ckpt.model_checkpoint_path) # 恢复会话到当前的神经网络
global_step = ckpt.model_checkpoint_path.split(
'/'
)[-1].split(
'-'
)[-1] # 从ckpt.model_checkpoint_path中，通过字符 "/" 和 "-"提取出最后一个整数（保存的轮数），恢复轮数，
coord = tf.train.Coordinator() #3开启线程协调器
threads = tf.train.start_queue_runners(sess=sess,
coord=coord) #4
xs, ys = sess.run([img_batch, label_batch
]) #5# 在 sess.run 中执行图片和标签的批获取
reshaped_xs = np.reshape(
xs,
( #导入部分，更改参数的形状
BATCH_SIZE, cifar10_lenet5_forward.IMAGE_SIZE,
cifar10_lenet5_forward.IMAGE_SIZE,
cifar10_lenet5_forward.NUM_CHANNELS))
accuracy_score = sess.run(
accuracy, # 计算准确率
feed_dict={
x: reshaped_xs,
y_: ys
})
print("after %s training step(s), test accuracy = %g" %
(global_step, accuracy_score))
coord.request_stop() #6
coord.join(threads) #7
else: #can not get checkpoint file ,print error infomation
print("No checkpoint file found")
return
time.sleep(
INTERVAL_TIME) # 设置等待时间，等backward生成新的checkpoint文件，再循环执行test函数
def backward(): #执行反向传播，训练参数w
x = tf.placeholder(
tf.float32,
[ #定义占位符x，以之代替输入图片
BATCH_SIZE, cifar10_lenet5_forward.IMAGE_HEIGHT,
cifar10_lenet5_forward.IMAGE_WIDTH,
cifar10_lenet5_forward.NUM_CHANNELS
])
y_ = tf.placeholder(tf.float32, [None, 10]) #定义占位符y_，作为传入标签
#True表示训练阶段，在进行forward时，if语句成立，进行dropout
y = cifar10_lenet5_forward.forward(x, True, REGULARIZER) #y是神经元的计算结果
y = tf.reshape(y, [-1, 10])
# print "y:",y
global_step = tf.Variable(0,
trainable=False) #定义变量global_step，并把它的属性设置为不可训练
# y = tf.reshape(y,[-1,18,10]) #将神经网络的计算结果y，转换为batch*18行*10列的tensor
# for i in range(18):
ce = tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=
y, #用交叉熵ce（cross entropy），使用tf.nn.sparse_softmax_cross_entropy_with_logits函数计算交叉熵，
labels=tf.argmax(y_,
1)) #其第一个参数是神经网络不包括softmax层的前向传播结果，第二个参数是训练数据的正确答案
# tf.argmax(vector, 1)：返回的是vector中的最大值的索引号
cem = tf.reduce_mean(ce) #计算在当前batch中所有样例的交叉熵平均值
loss = cem + tf.add_n(
tf.get_collection('losses')
) # 总损失等于交叉熵损失 + 正则化损失的和,losses保存有正则化的计算结果（forward中getweight（）对参数进行了正则化计算）
learning_rate = tf.train.exponential_decay( # 设置指数衰减的学习率
LEARNING_RATE_BASE, # 基础学习率，随着迭代的进行，更新变量时使用的学习率在此基础上递减
global_step, # 当前迭代轮数
train_num_examples / BATCH_SIZE, # 过完所有训练数据所需迭代次数
LEARNING_RATE_DECAY, # 指数学习衰减率
staircase=True)
train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(
loss, global_step=global_step) # 使用梯度下降优化损失函数，损失函数包含了交叉熵和正则化损失
ema = tf.train.ExponentialMovingAverage(MOVING_AVERAGE_DECAY,
global_step) # 初始化滑动平均类
ema_op = ema.apply(tf.trainable_variables()) # 对所有表示神经网络参数的变量进行滑动平均
#bind operation train_step & ema_op together to realize two operations at time
with tf.control_dependencies(
[train_step,
ema_op]): # 使用tf.control_dependencies机制一次完成多个操作。在此神经网络模型中，每过一遍数据既通过
train_op = tf.no_op(name='train') # 反向传播更新参数，又更新了每一个参数的滑动平均值
saver = tf.train.Saver() # 声明tf.train.Saver类用于保存模型
img_batch, lable_batch = cifar10_lenet5_generateds.get_tfrecord(
BATCH_SIZE, isTrain=True) #3一次批获取 batch_size 张图片和标签
with tf.Session() as sess:
sess.run(tf.global_variables_initializer()) # 初始化所有变量
# 恢复模块
ckpt = tf.train.get_checkpoint_state("./model") # 从"./model"中加载训练好的模型
if ckpt and ckpt.model_checkpoint_path: # 若ckpt和保存的模型在指定路径中存在，则将保存的神经网络模型加载到当前会话中
print ckpt.model_checkpoint_path
saver.restore(sess, ckpt.model_checkpoint_path)
coord = tf.train.Coordinator() #4开启线程协调器
threads = tf.train.start_queue_runners(sess=sess, coord=coord) #5
for i in range(STEPS): # 迭代地训练神经网络
xs, ys = sess.run([img_batch,
lable_batch]) #6将一个batch的训练数数据和对应标签分别赋给xs，ys
reshaped_xs = np.reshape(
xs,
( #导入部分，更改参数的形状
BATCH_SIZE, cifar10_lenet5_forward.IMAGE_HEIGHT,
cifar10_lenet5_forward.IMAGE_WIDTH,
cifar10_lenet5_forward.NUM_CHANNELS))
reshaped_ys = np.reshape(ys, (-1, 10))
_, loss_value, step = sess.run(
[
train_op, loss, global_step
], # 计算损失函数结果，计算节点train_op, loss,global_step并返回结果至 _, loss_value, step ，
feed_dict={
x: reshaped_xs,
y_: reshaped_ys
}
) #'_' means an anonymous variable which will not in use any more
if i % 1000 == 0: # 每1000轮打印损失函数信息，并保存当前的模型
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
) # 保存当前模型，globle_step参数可以使每个被保存模型的文件名末尾都加上训练的轮数
# 文件的名字是MODEL_SAVE_PATH + MODEL_NAME + global_step
coord.request_stop() #7关闭线程协调器
coord.join(threads) #8
#coding:utf-8
import tensorflow as tf
import tensorflow.contrib.slim.nets as nets
import cifar10_lenet5_generateds
BATCH_SIZE = 100
save_dir = r"./train_image_63.model"
batch_size_ = 100
lr = tf.Variable(0.0001, dtype=tf.float32)
x = tf.placeholder(tf.float32, [None, 32, 32, 3])
y_ = tf.placeholder(tf.float32, [None]) #
img_batch, label_batch = cifar10_lenet5_generateds.get_tfrecord(
BATCH_SIZE, isTrain=True
) #img_batch = [[32,32,3],[32,32,3]], label_batch=[[0,0,0,1,0...], [0,1,0....]]
#with tf.Session() as sess:#???
# print sess.run(img_batch), sess.run(label_batch)#???
# 将label值进行onehot编码,10分类
one_hot_labels = tf.one_hot(indices=tf.cast(y_, tf.int32), depth=10)
#one_hot_labels = tf.argmax(y_,1)
pred, end_points = nets.resnet_v2.resnet_v2_50(x,
num_classes=10,
is_training=True)
pred = tf.reshape(pred, shape=[-1, 10]) #预测值，相当于y
# 定义损失函数和优化器
loss = tf.reduce_mean(
tf.nn.sigmoid_cross_entropy_with_logits(logits=pred,
labels=one_hot_labels))
optimizer = tf.train.AdamOptimizer(learning_rate=lr).minimize(loss)
# 准确度
def test():
with tf.Graph().as_default() as g: #复现之前定义的计算图，并执行以下操作
x = tf.placeholder(tf.float32, [None, 32, 32, 3])
y_ = tf.placeholder(tf.float32, [None]) #
y, end_points = nets.resnet_v2.resnet_v2_50(x,
num_classes=10,
is_training=True)
y = tf.reshape(y, shape=[-1, 10]) #预测值，相当于y
# ema = tf.train.ExponentialMovingAverage(cifar10_lenet5_backward.MOVING_AVERAGE_DECAY)# 实现滑动平均模型，参数MOVING_AVERAGE_DECAY用于控制模型更新的速度，训练过程中会对每一个变量维护一个影子变量
# ema_restore = ema.variables_to_restore() # variable_to_restore()返回dict ({ema_variables : variables})，字典中保存变量的影子值和现值
# saver = tf.train.Saver(ema_restore) # 创建可还原滑动平均值的对象saver，测试时使用w的影子值，有更好的适配性
saver = tf.train.Saver() # 创建可还原滑动平均值的对象saver，测试时使用w的影子值，有更好的适配性
one_hot_labels = tf.one_hot(indices=tf.cast(y_, tf.int32), depth=10)
correct_prediction = tf.equal(tf.argmax(y, 1),
tf.argmax(one_hot_labels, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
img_batch, label_batch = cifar10_lenet5_generateds.get_tfrecord(
TEST_NUM, isTrain=False) #2 一次批获取 TEST_NUM 张图片和标签
print "before true"
while True:
with tf.Session() as sess:
ckpt = tf.train.get_checkpoint_state(
cifar10_lenet5_backward.MODEL_SAVE_PATH) # 从指定路径中，加载训练好的模型
if ckpt and ckpt.model_checkpoint_path: # 若已有ckpt模型则执行以下恢复操作
saver.restore(sess,
ckpt.model_checkpoint_path) # 恢复会话到当前的神经网络
global_step = ckpt.model_checkpoint_path.split(
'/'
)[-1].split(
'-'
)[-1] # 从ckpt.model_checkpoint_path中，通过字符 "/" 和 "-"提取出最后一个整数（保存的轮数），恢复轮数，
coord = tf.train.Coordinator() #3开启线程协调器
threads = tf.train.start_queue_runners(sess=sess,
coord=coord) #4
xs, ys = sess.run([img_batch, label_batch
]) #5# 在 sess.run 中执行图片和标签的批获取
accuracy_score = sess.run(
accuracy, # 计算准确率
feed_dict={
x: xs,
y_: ys
})
print("after %s training step(s), test accuracy = %g" %
(global_step, accuracy_score))
coord.request_stop() #6
coord.join(threads) #7
else: #can not get checkpoint file ,print error infomation
print("No checkpoint file found")
return
time.sleep(
INTERVAL_TIME) # 设置等待时间，等backward生成新的checkpoint文件，再循环执行test函数