浅入浅出TensorFlow 4 - 训练CIFAR数据

一.  CIFAR数据集

       CIFAR数据集是一个经典的数据集，提供两个版本的分类样本，CIFAR-10和CIFAR-100。

       CIFAR-10 提供10类标注数据，每类6000张（32*32），其中5000张用于训练，1000张用于测试。

    获取数据集的方法：

   git clone https://github.com/tensorflow/models.git
   cd models/tutorials/image/cifar10

       

       可以看一下我们从github上down下来的数据，外面不看了，直接进 tutorials/image，教程专用，看来是基础的不能再基础了。

       里面提供了几个典型的数据集的 下载、训练等接口，方便直接在python里调用。

       进入cifar10，能够看到：

        

      其中文件 cifar10.py 和 cifar10_input.py 就是接下来我们要 import 的。

二.  代码实现

       撸一段 Python 代码，可以View里面的注释讲解：

#coding=utf-8
import cifar10,cifar10_input
import tensorflow as tf
import numpy as np
import time

# define max_iter_step  batch_size
max_iter_step = 1000
batch_size = 128

# define variable_with_weight_loss
# 和之前定义的weight有所不同， 
# 这里定义附带loss的weight，通过权重惩罚避免部分权重系数过大，导致overfitting 
def variable_with_weight_loss(shape,stddev,w1):
	var = tf.Variable(tf.truncated_normal(shape,stddev=stddev))
	if w1 is not None:
		weight_loss = tf.multiply(tf.nn.l2_loss(var),w1,name='weight_loss')
		tf.add_to_collection('losses',weight_loss)
	return var

# 下载数据集 － 调用cifar10函数下载并解压
cifar10.maybe_download_and_extract()

cifar_dir = '/tmp/cifar10_data/cifar-10-batches-bin'

# 采用 data augmentation进行数据处理
# 生成训练数据，训练数据通过cifar10_input的distort变化 
images_train, labels_train = cifar10_input.distorted_inputs(data_dir=cifar_dir,batch_size=batch_size)
# 测试数据（eval_data 测试数据）
images_test,labels_test = cifar10_input.inputs(eval_data=True,data_dir=cifar_dir,batch_size=batch_size)

# 创建输入数据，采用 placeholder
x_input = tf.placeholder(tf.float32,[batch_size,24,24,3])
y_input = tf.placeholder(tf.int32,[batch_size])

# 创建第一个卷积层 input:3(channel) kernel:64 size:5*5
weight1 = variable_with_weight_loss(shape=[5,5,3,64],stddev=5e-2,w1=0.0)
bias1 = tf.Variable(tf.constant(0.0,shape=[64]))
conv1 = tf.nn.conv2d(x_input,weight1,[1,1,1,1],padding='SAME')
relu1 = tf.nn.relu(tf.nn.bias_add(conv1,bias1))
pool1 = tf.nn.max_pool(conv1,ksize=[1,3,3,1],strides=[1,2,2,1],padding='SAME')
norm1 = tf.nn.lrn(pool1,4,bias=1.0,alpha=0.001/9.0,beta=0.75)

# 创建第二个卷积层 input:64 kernel:64 size:5*5
weight2 = variable_with_weight_loss(shape=[5,5,64,64],stddev=5e-2,w1=0.0)
bias2 = tf.Variable(tf.constant(0,1,shape=[64]))
conv2 = tf.nn.conv2d(norm1,weight2,[1,1,1,1],padding='SAME')
relu2 = tf.nn.relu(tf.nn.bias_add(conv2,bias2))
norm2 = tf.nn.lrn(relu2,4,bias=1.0,alpha=0.001/9.0,beta=0.75)
pool2 = tf.nn.max_pool(norm2,ksize=[1,3,3,1],strides=[1,2,2,1],padding='SAME')

# 创建第三个层－全连接层  output:384
reshape = tf.reshape(pool2,[batch_size,-1])
dim = reshape.get_shape()[1].value
weight3 = variable_with_weight_loss(shape=[dim,384],stddev=0.04,w1=0.004)
bias3 = tf.Variable(tf.constant(0.1,shape=[384]))
local3 = tf.nn.relu(tf.matmul(reshape,weight3)+bias3)

# 创建第四个层－全连接层  output:192
weight4 = variable_with_weight_loss(shape=[384,192],stddev=0.04,w1=0.004)
bias4 = tf.Variable(tf.constant(0.1,shape=[192]))

# 最后一层  output:10
weight5 = variable_with_weight_loss(shape=[192,10],stddev=1/192.0,w1=0.0)
bias5 = tf.Variable(tf.constant(0.0,shape=[10]))
results = tf.add(tf.matmul(local4,weight5),bias5)

# 定义loss
def loss(results,labels):
	labels = tf.cast(labels,tf.int64)
	cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=results,labels=labels,name='cross_entropy_per_example')
	cross_entropy_mean = tf.reduce_mean(cross_entropy,name='cross_entropy')
	tf.add_to_collection('losses',cross_entropy_mean)
	return tf.add_n(tf.get_collection('losses'),name='total_loss')

# 计算loss
loss = loss(results,y_input)

train_op = tf.train.AdamOptimizer(1e-3).minimize(loss)  # Adam
top_k_op = tf.nn.in_top_k(results,y_input,1)  # top1 准确率

sess = tf.InteractiveSession()         # 创建session
tf.global_variable_initializer().run() # 初始化全部模型

tf.train.start_queue_runners()  # 启动多线程加速

# 开始训练
for step in range(max_steps):
	start_time = time.time()
	image_batch,label_batch = sess.run([images_train,labels_train])
	_,loss_value = sess.run([train_op,loss],
		feed_dict={x_input:image_batch, y_input:label_batch})
	duration = time.time() - start_time
	if step % 10 == 0:
		examples_per_sec = batch_size/duration
		sec_per_batch = float(duration)

		format_str = ('step %d,loss=%.2f (%.1f examples/sec; %.3f sec/batch')
		print(format_str % (step,loss_value,examples_per_sec,sec_per_batch))

# 评测模型在测试集上的准确度
num_examples = 10000
import math
num_iter = int(math.ceil(num_examples/batch_size))
true_count = 0
total_sample_count = num_iter * batch_size
step = 0
while step < num_iter:
	image_batch,label_batch = sess.run([images_test,labels_test])
	predictions = sess.run([top_k_op],feed_dict={x_input:image_batch,y_input:label_batch})
	true_count += np.sum(predictions)
	step += 1

# 打印结果
precision = true_count / total_sample_count
print('precision @ 1 = %.3f' % precision)

      注意，这里与前面不一样的地方在于引入了权值惩罚，另外，top_k的用法也是第一次，将代码另存为 .py文件，copy到models/tutorials/image/cifar10目录下调用，观察下载数据及训练过程，然后再Review代码，相信会有新的收获！