import cifar10
import cifar10_input
import tensorflow as tf
import numpy as np
import time

# 定義batch_size, 訓(xùn)練輪數(shù)max_steps, 以及下載CIFAR-10數(shù)據(jù)的默認(rèn)路徑
max_steps = 3000
batch_size = 128
data_dir = 'E:\\tmp\cifar10_data\cifar-10-batches-bin'

# 定義初始化weight的函數(shù)，定義的同時，對weight加一個L2 loss，放在集'losses'中
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

# 使用cifar10類下載數(shù)據(jù)集，并解壓、展開到其默認(rèn)位置
#cifar10.maybe_download_and_extract()

# 在使用cifar10_input類中的distorted_inputs函數(shù)產(chǎn)生訓(xùn)練需要使用的數(shù)據(jù)。需要注意的是，返回的是已經(jīng)封裝好的tensor，
# 且對數(shù)據(jù)進(jìn)行了Data Augmentation（水平翻轉(zhuǎn)、隨機(jī)剪切、設(shè)置隨機(jī)亮度和對比度、對數(shù)據(jù)進(jìn)行標(biāo)準(zhǔn)化）
images_train, labels_train = cifar10_input.distorted_inputs(data_dir=data_dir, batch_size=batch_size)

# 再使用cifar10_input.inputs函數(shù)生成測試數(shù)據(jù)，這里不需要進(jìn)行太多處理
images_test, labels_test = cifar10_input.inputs(eval_data=True,
                        data_dir=data_dir,
                        batch_size=batch_size)

# 創(chuàng)建數(shù)據(jù)的placeholder
image_holder = tf.placeholder(tf.float32, [batch_size, 24, 24, 3])
label_holder = tf.placeholder(tf.int32, [batch_size])

# 創(chuàng)建第一個卷積層
weight1 = variable_with_weight_loss(shape=[5, 5, 3, 64], stddev=5e-2,
                  w1=0.0)
kernel1 = tf.nn.conv2d(image_holder, weight1, strides=[1, 1, 1, 1], padding='SAME')
bias1 = tf.Variable(tf.constant(0.0, shape=[64]))
conv1 = tf.nn.relu(tf.nn.bias_add(kernel1, bias1))
pool1 = tf.nn.max_pool(conv1, ksize=[1, 3, 3, 1], strides=[1, 2, 2, 1],
            padding='SAME')
# LRN層對ReLU會比較有用，但不適合Sigmoid這種有固定邊界并且能抑制過大值的激活函數(shù)
norm1 = tf.nn.lrn(pool1, 4, bias=1.0, alpha=0.001 / 9.0, beta=0.75)

# 創(chuàng)建第二個卷積層
weight2 = variable_with_weight_loss(shape=[5, 5, 64, 64], stddev=5e-2,
                  w1=0.0)
kernel2 = tf.nn.conv2d(norm1, weight2, strides=[1, 1, 1, 1], padding='SAME')
bias2 = tf.Variable(tf.constant(0.1, shape=[64]))
conv2 = tf.nn.relu(tf.nn.bias_add(kernel2, bias2))
norm2 = tf.nn.lrn(conv2, 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')

# 使用一個全連接層
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)

# 再使用一個全連接層，隱含節(jié)點數(shù)下降了一半，只有192個，其他的超參數(shù)保持不變
weight4 = variable_with_weight_loss(shape=[384, 192], stddev=0.04, w1=0.004)
bias4 = tf.Variable(tf.constant(0.1, shape=[192]))
local4 = tf.nn.relu(tf.matmul(local3, weight4) + bias4)

# 最后一層，將softmax放在了計算loss部分
weight5 = variable_with_weight_loss(shape=[192, 10], stddev=1 / 192.0, w1=0.0)
bias5 = tf.Variable(tf.constant(0.0, shape=[10]))
logits = tf.add(tf.matmul(local4, weight5), bias5)

# 定義loss
def loss(logits, labels):
  labels = tf.cast(labels, tf.int64)
  cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=logits, 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')

# 獲取最終的loss
loss = loss(logits, label_holder)

# 優(yōu)化器
train_op = tf.train.AdamOptimizer(1e-3).minimize(loss)

# 使用tf.nn.in_top_k函數(shù)求輸出結(jié)果中top k的準(zhǔn)確率，默認(rèn)使用top 1，也就是輸出分?jǐn)?shù)最高的那一類的準(zhǔn)確率
top_k_op = tf.nn.in_top_k(logits, label_holder, 1)

# 使用tf.InteractiveSession創(chuàng)建默認(rèn)的session，接著初始化全部模型參數(shù)
sess = tf.InteractiveSession()
tf.global_variables_initializer().run()

# 啟動圖片數(shù)據(jù)增強(qiáng)線程
tf.train.start_queue_runners()

# 正式開始訓(xùn)練
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={image_holder: image_batch, label_holder: label_batch})
  duration = time.time() - start_time
  if step % 10 == 0:
    example_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, example_per_sec, sec_per_batch))

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={image_holder: image_batch, label_holder: label_holder})
  true_count += np.sum(predictions)
  step += 1

precision = true_count / total_sample_count
print('precision @ 1 = %.3f'%precision)