實例37：卷積函數的使用
# -*- coding: utf-8 -*-
import tensorflow as tf
# [batch, in_height, in_width, in_channels] [訓練時一個batch的圖片數量, 圖片高度, 圖片寬度, 圖像通道數]
input = tf.Variable(tf.constant(1.0,shape = [1, 5, 5, 1]))
input2 = tf.Variable(tf.constant(1.0,shape = [1, 5, 5, 2]))
input3 = tf.Variable(tf.constant(1.0,shape = [1, 4, 4, 1]))
# [filter_height, filter_width, in_channels, out_channels] [卷積核的高度，卷積核的寬度，圖像通道數，卷積核個數]
filter1 = tf.Variable(tf.constant([-1.0,0,0,-1],shape = [2, 2, 1, 1]))
filter2 = tf.Variable(tf.constant([-1.0,0,0,-1,-1.0,0,0,-1],shape = [2, 2, 1, 2]))
filter3 = tf.Variable(tf.constant([-1.0,0,0,-1,-1.0,0,0,-1,-1.0,0,0,-1],shape = [2, 2, 1, 3]))
filter4 = tf.Variable(tf.constant([-1.0,0,0,-1,
-1.0,0,0,-1,
-1.0,0,0,-1,
-1.0,0,0,-1],shape = [2, 2, 2, 2]))
filter5 = tf.Variable(tf.constant([-1.0,0,0,-1,-1.0,0,0,-1],shape = [2, 2, 2, 1]))
# padding的值爲‘VALID’，表示邊緣不填充, 當其爲‘SAME’時，表示填充到卷積核可以到達圖像邊緣
op1 = tf.nn.conv2d(input, filter1, strides=[1, 2, 2, 1], padding='SAME') #1個通道輸入，生成1個feature ma
op2 = tf.nn.conv2d(input, filter2, strides=[1, 2, 2, 1], padding='SAME') #1個通道輸入，生成2個feature map
op3 = tf.nn.conv2d(input, filter3, strides=[1, 2, 2, 1], padding='SAME') #1個通道輸入，生成3個feature map
op4 = tf.nn.conv2d(input2, filter4, strides=[1, 2, 2, 1], padding='SAME') # 2個通道輸入，生成2個feature
op5 = tf.nn.conv2d(input2, filter5, strides=[1, 2, 2, 1], padding='SAME') # 2個通道輸入，生成一個feature map
vop1 = tf.nn.conv2d(input, filter1, strides=[1, 2, 2, 1], padding='VALID') # 5*5 對於pading不同而不同
op6 = tf.nn.conv2d(input3, filter1, strides=[1, 2, 2, 1], padding='SAME')
vop6 = tf.nn.conv2d(input3, filter1, strides=[1, 2, 2, 1], padding='VALID') #4*4與pading無關
init = tf.global_variables_initializer()
with tf.Session() as sess:
sess.run(init)
print("op1:\n",sess.run([op1,filter1]))#1-1 後面補0
print("------------------")
print("op2:\n",sess.run([op2,filter2])) #1-2多卷積核 按列取
print("op3:\n",sess.run([op3,filter3])) #1-3
print("------------------")
print("op4:\n",sess.run([op4,filter4]))#2-2 通道疊加
print("op5:\n",sess.run([op5,filter5]))#2-1
print("------------------")
print("op1:\n",sess.run([op1,filter1]))#1-1
print("vop1:\n",sess.run([vop1,filter1]))
print("op6:\n",sess.run([op6,filter1]))
print("vop6:\n",sess.run([vop6,filter1]))
實例38：使用卷積提取圖片的輪廓
sobel算子：將彩色的圖片生成帶有邊緣化信息的圖片。
# -*- coding: utf-8 -*-
import matplotlib.pyplot as plt # plt 用於顯示圖片
import matplotlib.image as mpimg # mpimg 用於讀取圖片
import numpy as np
import tensorflow as tf
myimg = mpimg.imread('img.jpg') # 讀取和代碼處於同一目錄下的圖片
plt.imshow(myimg) # 顯示圖片
plt.axis('off') # 不顯示座標軸
plt.show()
print(myimg.shape)
full=np.reshape(myimg,[1,140, 121, 3])
inputfull = tf.Variable(tf.constant(1.0,shape = [1, 140, 121, 3]))
filter = tf.Variable(tf.constant([[-1.0,-1.0,-1.0], [0,0,0], [1.0,1.0,1.0],
[-2.0,-2.0,-2.0], [0,0,0], [2.0,2.0,2.0],
[-1.0,-1.0,-1.0], [0,0,0], [1.0,1.0,1.0]],shape = [3, 3, 3, 1]))
op = tf.nn.conv2d(inputfull, filter, strides=[1, 1, 1, 1], padding='SAME') #3個通道輸入，生成1個feature ma
o=tf.cast( ((op-tf.reduce_min(op))/(tf.reduce_max(op)-tf.reduce_min(op)) ) *255 ,tf.uint8)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer() )
t,f=sess.run([o,filter],feed_dict={ inputfull:full})
#print(f)
t=np.reshape(t,[140, 121])
plt.imshow(t,cmap='Greys_r') # 顯示圖片
plt.axis('off') # 不顯示座標軸
plt.show()
實例39：池化函數的使用
- tf.nn.max_pool(input, ksize, strides, padding, name = None)
- tf.nn.avg_pool(input, ksize, strides, padding, name = None)
# -*- coding: utf-8 -*-
import tensorflow as tf
img=tf.constant([
[[0.0,4.0],[0.0,4.0],[0.0,4.0],[0.0,4.0]],
[[1.0,5.0],[1.0,5.0],[1.0,5.0],[1.0,5.0]],
[[2.0,6.0],[2.0,6.0],[2.0,6.0],[2.0,6.0]],
[[3.0,7.0],[3.0,7.0], [3.0,7.0],[3.0,7.0]]
])
img=tf.reshape(img,[1,4,4,2])
pooling=tf.nn.max_pool(img,[1,2,2,1],[1,2,2,1],padding='VALID')
pooling1=tf.nn.max_pool(img,[1,2,2,1],[1,1,1,1],padding='VALID')
pooling2=tf.nn.avg_pool(img,[1,4,4,1],[1,1,1,1],padding='SAME')
pooling3=tf.nn.avg_pool(img,[1,4,4,1],[1,4,4,1],padding='SAME')
nt_hpool2_flat = tf.reshape(tf.transpose(img), [-1, 16])
pooling4=tf.reduce_mean(nt_hpool2_flat,1) #1對行求均值（1表示軸是列） 0 對列求均值
with tf.Session() as sess:
print("image:")
image=sess.run(img)
print (image)
result=sess.run(pooling)
print ("reslut:\n",result)
result=sess.run(pooling1)
print ("reslut1:\n",result)
result=sess.run(pooling2)
print ("reslut2:\n",result)
result=sess.run(pooling3)
print ("reslut3:\n",result)
flat,result=sess.run([nt_hpool2_flat,pooling4])
print ("reslut4:\n",result)
print("flat:\n",flat)
實例40：導入並顯示CIFAR數據集
# -*- coding: utf-8 -*-
#放在cifar目錄下
import cifar10
cifar10.maybe_download_and_extract(data_dir="H:/tensorflow_projects/chap8/")
# -*- coding: utf-8 -*-
#放在cifar目錄下
import cifar10_input
import tensorflow as tf
import pylab
import numpy as np
#取數據
batch_size = 12
data_dir = 'H:/tensorflow_projects/chap8/cifar-10-batches-bin'
images_test, labels_test = cifar10_input.inputs(eval_data = True, data_dir = data_dir, batch_size = batch_size)
#sess = tf.InteractiveSession()
#tf.global_variables_initializer().run()
#tf.train.start_queue_runners()
#image_batch, label_batch = sess.run([images_test, labels_test])
#print("__\n",image_batch[0])
#
#print("__\n",label_batch[0])
#pylab.imshow(image_batch[0])
#pylab.show()
#
sess = tf.Session()
tf.global_variables_initializer().run(session=sess)
tf.train.start_queue_runners(sess=sess)
image_batch, label_batch = sess.run([images_test, labels_test])
print("__\n",image_batch[0])
print("__\n",label_batch[0])
pylab.imshow( (image_batch[0]-np.min(image_batch[0])) / (np.max(image_batch[0])-np.min(image_batch[0]) ) )
pylab.show()
#with tf.Session() as sess:
# tf.global_variables_initializer().run()
# tf.train.start_queue_runners()
# image_batch, label_batch = sess.run([images_test, labels_test])
# print("__\n",image_batch[0])
#
# print("__\n",label_batch[0])
# pylab.imshow(image_batch[0])
# pylab.show()
實例41：顯示CIFAR數據集的原始圖片
# -*- coding: utf-8 -*-
import numpy as np
from scipy.misc import imsave
filename = '../cifar-10-batches-bin/test_batch.bin'
bytestream = open(filename, "rb")
buf = bytestream.read(10000 * (1 + 32 * 32 * 3))
bytestream.close()
data = np.frombuffer(buf, dtype=np.uint8)
data = data.reshape(10000, 1 + 32*32*3)
labels_images = np.hsplit(data, [1])
labels = labels_images[0].reshape(10000)
images = labels_images[1].reshape(10000, 32, 32, 3)
img = np.reshape(images[0], (3, 32, 32)) #導出第一幅圖
img = img.transpose(1, 2, 0)
import pylab
print(labels[0])
pylab.imshow(img)
pylab.show()
# -*- coding: utf-8 -*-
import tensorflow as tf
#創建長度爲100的隊列
queue = tf.FIFOQueue(100,"float")
c = tf.Variable(0.0) #計數器
#加1操作
op = tf.assign_add(c,tf.constant(1.0))
#操作:將計數器的結果加入隊列
enqueue_op = queue.enqueue(c)
#創建一個隊列管理器QueueRunner，用這兩個操作向q中添加元素。目前我們只使用一個線程:
qr = tf.train.QueueRunner(queue,enqueue_ops=[op,enqueue_op])
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
coord = tf.train.Coordinator()
## 啓動入隊線程, Coordinator是線程的參數
enqueue_threads = qr.create_threads(sess, coord = coord,start=True) # 啓動入隊線程
# 主線程
for i in range(0, 10):
print ("-------------------------")
print(sess.run(queue.dequeue()))
coord.request_stop() #通知其他線程關閉 其他所有線程關閉之後，這一函數才能返回
#join操作經常用在線程當中,其作用是等待某線程結束
#coord.join(enqueue_threads)
實例42：協調器的用法演示
# -*- coding: utf-8 -*-
#放在cifar目錄下
import cifar10_input
import tensorflow as tf
import pylab
#取數據
batch_size = 12
data_dir = '../cifar-10-batches-bin'
images_test, labels_test = cifar10_input.inputs(eval_data = True, data_dir = data_dir, batch_size = batch_size)
with tf.Session() as sess:
tf.global_variables_initializer().run()
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess, coord)
image_batch, label_batch = sess.run([images_test, labels_test])
print("__\n",image_batch[0])
print("__\n",label_batch[0])
pylab.imshow(image_batch[0])
pylab.show()
coord.request_stop()
實例43：爲session中的隊列加上協調器
import cifar10_input
import tensorflow as tf
import pylab
import numpy as np
#取數據
batch_size = 12
data_dir = 'H:/tensorflow_projects/chap8/cifar-10-batches-bin'
images_test, labels_test = cifar10_input.inputs(eval_data = True, data_dir = data_dir, batch_size = batch_size)
with tf.Session() as sess:
tf.global_variables_initializer().run()
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess, coord)
image_batch, label_batch = sess.run([images_test, labels_test])
print("__\n",image_batch[0])
print("__\n",label_batch[0])
pylab.imshow(image_batch[0])
pylab.show()
coord.request_stop()
實例44：建立一個帶有全局平均池化層的卷積神經網絡
# -*- coding: utf-8 -*-
import cifar10_input
import tensorflow as tf
import numpy as np
# 1. 引入數據集
batch_size = 128
data_dir = '../cifar-10-batches-bin'
print("begin")
images_train, labels_train = cifar10_input.inputs(
eval_data=False, data_dir=data_dir, batch_size=batch_size)
images_test, labels_test = cifar10_input.inputs(
eval_data=True, data_dir=data_dir, batch_size=batch_size)
print("begin data")
# 2. 定義網絡結構
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')
def avg_pool_6x6(x):
return tf.nn.avg_pool(x, ksize=[1, 6, 6, 1],
strides=[1, 6, 6, 1], padding='SAME')
# tf Graph Input
# cifar data image of shape 24*24*3
x = tf.placeholder(tf.float32, [None, 24, 24, 3])
y = tf.placeholder(tf.float32, [None, 10]) # 0-9 數字=> 10 classes
W_conv1 = weight_variable([5, 5, 3, 64])
b_conv1 = bias_variable([64])
x_image = tf.reshape(x, [-1, 24, 24, 3])
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)
W_conv2 = weight_variable([5, 5, 64, 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)
W_conv3 = weight_variable([5, 5, 64, 10])
b_conv3 = bias_variable([10])
h_conv3 = tf.nn.relu(conv2d(h_pool2, W_conv3) + b_conv3)
nt_hpool3 = avg_pool_6x6(h_conv3) # 10
nt_hpool3_flat = tf.reshape(nt_hpool3, [-1, 10])
y_conv = tf.nn.softmax(nt_hpool3_flat)
cross_entropy = -tf.reduce_sum(y*tf.log(y_conv))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
# 3. 運行session進行訓練
sess = tf.Session()
sess.run(tf.global_variables_initializer())
tf.train.start_queue_runners(sess=sess)
for i in range(15000):
# 20000
image_batch, label_batch = sess.run([images_train, labels_train])
label_b = np.eye(10, dtype=float)[label_batch] # one hot
train_step.run(feed_dict={x: image_batch, y: label_b}, session=sess)
if i % 200 == 0:
train_accuracy = accuracy.eval(
feed_dict={x: image_batch, y: label_b}, session=sess)
print("step %d, training accuracy %g" % (i, train_accuracy))
# 4. 評估結果
image_batch, label_batch = sess.run([images_test, labels_test])
label_b = np.eye(10, dtype=float)[label_batch] # one hot
print("finished！ test accuracy %g" % accuracy.eval(feed_dict={
x: image_batch, y: label_b}, session=sess))
begin begin data step 0, training accuracy 0.164062 step 200, training accuracy 0.320312 step 400, training accuracy 0.414062 step 600, training accuracy 0.4375 step 800, training accuracy 0.429688 step 1000, training accuracy 0.507812 step 1200, training accuracy 0.367188 step 1400, training accuracy 0.523438 step 1600, training accuracy 0.507812 step 1800, training accuracy 0.445312 step 2000, training accuracy 0.5625 step 2200, training accuracy 0.5 step 2400, training accuracy 0.507812 step 2600, training accuracy 0.460938 step 2800, training accuracy 0.585938 step 3000, training accuracy 0.617188 step 3200, training accuracy 0.554688 step 3400, training accuracy 0.546875 step 3600, training accuracy 0.554688 step 3800, training accuracy 0.523438 step 4000, training accuracy 0.554688 step 4200, training accuracy 0.59375 step 4400, training accuracy 0.546875 step 4600, training accuracy 0.601562 step 4800, training accuracy 0.625 step 5000, training accuracy 0.554688 step 5200, training accuracy 0.632812 step 5400, training accuracy 0.617188 step 5600, training accuracy 0.5625 step 5800, training accuracy 0.59375 step 6000, training accuracy 0.515625 step 6200, training accuracy 0.6875 step 6400, training accuracy 0.578125 step 6600, training accuracy 0.578125 step 6800, training accuracy 0.601562 step 7000, training accuracy 0.640625 step 7200, training accuracy 0.640625 step 7400, training accuracy 0.609375 step 7600, training accuracy 0.617188 step 7800, training accuracy 0.609375 step 8000, training accuracy 0.640625 step 8200, training accuracy 0.578125 step 8400, training accuracy 0.609375 step 8600, training accuracy 0.59375 step 8800, training accuracy 0.625 step 9000, training accuracy 0.617188 step 9200, training accuracy 0.648438 step 9400, training accuracy 0.617188 step 9600, training accuracy 0.648438 step 9800, training accuracy 0.617188 step 10000, training accuracy 0.632812 step 10200, training accuracy 0.65625 step 10400, training accuracy 0.679688 step 10600, training accuracy 0.609375 step 10800, training accuracy 0.664062 step 11000, training accuracy 0.734375 step 11200, training accuracy 0.632812 step 11400, training accuracy 0.6875 step 11600, training accuracy 0.71875 step 11800, training accuracy 0.585938 step 12000, training accuracy 0.679688 step 12200, training accuracy 0.671875 step 12400, training accuracy 0.640625 step 12600, training accuracy 0.609375 step 12800, training accuracy 0.679688 step 13000, training accuracy 0.664062 step 13200, training accuracy 0.5625 step 13400, training accuracy 0.625 step 13600, training accuracy 0.65625 step 13800, training accuracy 0.640625 step 14000, training accuracy 0.6875 step 14200, training accuracy 0.664062 step 14400, training accuracy 0.695312 step 14600, training accuracy 0.703125 step 14800, training accuracy 0.671875 finished！ test accuracy 0.664062
MNIST卷積和CIFAR卷積分類
# -*- coding: utf-8 -*-
import tensorflow as tf
# 導入 MINST 數據集
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/data/", one_hot=True)
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')
def avg_pool_7x7(x):
return tf.nn.avg_pool(x, ksize=[1, 7, 7, 1],
strides=[1, 7, 7, 1], padding='SAME')
# tf Graph Input
x = tf.placeholder(tf.float32, [None, 784]) # mnist data維度 28*28=784
y = tf.placeholder(tf.float32, [None, 10]) # 0-9 數字=> 10 classes
W_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
x_image = tf.reshape(x, [-1,28,28,1])
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)
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)
#########################################################new
W_conv3 = weight_variable([5, 5, 64, 10])
b_conv3 = bias_variable([10])
h_conv3 = tf.nn.relu(conv2d(h_pool2, W_conv3) + b_conv3)
nt_hpool3=avg_pool_7x7(h_conv3)#64
nt_hpool3_flat = tf.reshape(nt_hpool3, [-1, 10])
y_conv=tf.nn.softmax(nt_hpool3_flat)
cross_entropy = -tf.reduce_sum(y*tf.log(y_conv))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
# 啓動session
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
for i in range(200):#20000
batch = mnist.train.next_batch(50)#50
if i%20 == 0:
train_accuracy = accuracy.eval(feed_dict={
x:batch[0], y: batch[1]})
print( "step %d, training accuracy %g"%(i, train_accuracy))
train_step.run(feed_dict={x: batch[0], y: batch[1]})
print ("test accuracy %g"%accuracy.eval(feed_dict={
x: mnist.test.images, y: mnist.test.labels}))
step 0, training accuracy 0.26 step 20, training accuracy 0.14 step 40, training accuracy 0.14 step 60, training accuracy 0.26 step 80, training accuracy 0.18 step 100, training accuracy 0.22 step 120, training accuracy 0.22 step 140, training accuracy 0.3 step 160, training accuracy 0.12 step 180, training accuracy 0.34 test accuracy 0.2344
實例45：演示反捲積操作
# -*- coding: utf-8 -*-
import numpy as np
import tensorflow as tf
img = tf.Variable(tf.constant(1.0,shape = [1, 4, 4, 1]))
filter = tf.Variable(tf.constant([1.0,0,-1,-2],shape = [2, 2, 1, 1]))
conv = tf.nn.conv2d(img, filter, strides=[1, 2, 2, 1], padding='VALID')
cons = tf.nn.conv2d(img, filter, strides=[1, 2, 2, 1], padding='SAME')
print(conv.shape)
print(cons.shape)
contv= tf.nn.conv2d_transpose(conv, filter, [1,4,4,1],strides=[1, 2, 2, 1], padding='VALID')
conts = tf.nn.conv2d_transpose(cons, filter, [1,4,4,1],strides=[1, 2, 2, 1], padding='SAME')
with tf.Session() as sess:
sess.run(tf.global_variables_initializer() )
print("conv:\n",sess.run([conv,filter]))
print("cons:\n",sess.run([cons]))
print("contv:\n",sess.run([contv]))
print("conts:\n",sess.run([conts]))
實例46：演示反池化操作
# -*- coding: utf-8 -*-
import tensorflow as tf
import numpy as np
def max_pool_with_argmax(net, stride):
_, mask = tf.nn.max_pool_with_argmax(net, ksize=[1, stride, stride, 1], strides=[1, stride, stride, 1], padding='SAME')
mask = tf.stop_gradient(mask)
net = tf.nn.max_pool(net, ksize=[1, stride, stride, 1],strides=[1, stride, stride, 1], padding='SAME')
return net, mask
def unpool(net, mask, stride):
ksize = [1, stride, stride, 1]
input_shape = net.get_shape().as_list()
# calculation new shape
output_shape = (input_shape[0], input_shape[1] * ksize[1], input_shape[2] * ksize[2], input_shape[3])
# calculation indices for batch, height, width and feature maps
one_like_mask = tf.ones_like(mask)
batch_range = tf.reshape(tf.range(output_shape[0], dtype=tf.int64), shape=[input_shape[0], 1, 1, 1])
b = one_like_mask * batch_range
y = mask // (output_shape[2] * output_shape[3])
x = mask % (output_shape[2] * output_shape[3]) // output_shape[3]
feature_range = tf.range(output_shape[3], dtype=tf.int64)
f = one_like_mask * feature_range
# transpose indices & reshape update values to one dimension
updates_size = tf.size(net)
indices = tf.transpose(tf.reshape(tf.stack([b, y, x, f]), [4, updates_size]))
values = tf.reshape(net, [updates_size])
ret = tf.scatter_nd(indices, values, output_shape)
return ret
img=tf.constant([
[[0.0,4.0],[0.0,4.0],[0.0,4.0],[0.0,4.0]],
[[1.0,5.0],[1.0,5.0],[1.0,5.0],[1.0,5.0]],
[[2.0,6.0],[2.0,6.0],[2.0,6.0],[2.0,6.0]],
[[3.0,7.0],[3.0,7.0], [3.0,7.0],[3.0,7.0]]
])
img=tf.reshape(img,[1,4,4,2])
pooling2=tf.nn.max_pool(img,[1,2,2,1],[1,2,2,1],padding='SAME')
encode, mask = max_pool_with_argmax(img, 2)
img2 = unpool(encode,mask,2)
print(img.shape)
print(encode.shape)
print(mask.shape)
print(img2.shape)
with tf.Session() as sess:
print("image:")
print (sess.run(img))
result=sess.run(pooling2)
print ("pooling2:\n",result)
result,mask2=sess.run([encode, mask])
print ("encode:\n",result,mask2)
result=sess.run(img2)
print ("reslut:\n",result)
實例47：演示gradients的基本用法
# -*- coding: utf-8 -*-
import tensorflow as tf
w1 = tf.Variable([[1.,2]])
w2 = tf.Variable([[3.,4]])
y = tf.matmul(w1, [[9.],[10]])
#grads = tf.gradients(y,[w1,w2])#w2不相干，會報錯
grads = tf.gradients(y,[w1])
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
gradval = sess.run(grads)
print(gradval)
[array([[ 9., 10.]], dtype=float32)]
實例48：使用gradients對多個式子求多變量求導
# -*- coding: utf-8 -*-
import tensorflow as tf
tf.reset_default_graph()
w1 = tf.get_variable('w1', shape=[2])
w2 = tf.get_variable('w2', shape=[2])
w3 = tf.get_variable('w3', shape=[2])
w4 = tf.get_variable('w4', shape=[2])
y1 = w1 + w2+ w3
y2 = w3 + w4
a = w1+w2
a_stoped = tf.stop_gradient(a)
y3= a_stoped+w3
gradients = tf.gradients([y1, y2], [w1, w2, w3, w4], grad_ys=[tf.convert_to_tensor([1.,2.]),
tf.convert_to_tensor([3.,4.])])
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
print(sess.run(gradients))
[array([1., 2.], dtype=float32), array([1., 2.], dtype=float32), array([4., 6.], dtype=float32), array([3., 4.], dtype=float32)]
實例49：演示gradients停止
# -*- coding: utf-8 -*-
import tensorflow as tf
tf.reset_default_graph()
w1 = tf.get_variable('w1', shape=[2])
w2 = tf.get_variable('w2', shape=[2])
w3 = tf.get_variable('w3', shape=[2])
w4 = tf.get_variable('w4', shape=[2])
y1 = w1 + w2+ w3
y2 = w3 + w4
a = w1+w2
a_stoped = tf.stop_gradient(a)
y3= a_stoped+w3
gradients = tf.gradients([y1, y2], [w1, w2, w3, w4], grad_ys=[tf.convert_to_tensor([1.,2.]),
tf.convert_to_tensor([3.,4.])])
gradients2 = tf.gradients(y3, [w1, w2, w3], grad_ys=tf.convert_to_tensor([1.,2.]))
print(gradients2)
gradients3 = tf.gradients(y3, [ w3], grad_ys=tf.convert_to_tensor([1.,2.]))
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
print(sess.run(gradients))
#print(sess.run(gradients2))#報錯
print(sess.run(gradients3))
[None, None, <tf.Tensor 'gradients_1/grad_ys_0:0' shape=(2,) dtype=float32>] [array([1., 2.], dtype=float32), array([1., 2.], dtype=float32), array([4., 6.], dtype=float32), array([3., 4.], dtype=float32)] [array([1., 2.], dtype=float32)]
實例50：用反捲積復原CNN各層圖像
# -*- coding: utf-8 -*-
import cifar10_input
import tensorflow as tf
import numpy as np
#最大池化
def max_pool_with_argmax(net, stride):
_, mask = tf.nn.max_pool_with_argmax( net,ksize=[1, stride, stride, 1], strides=[1, stride, stride, 1],padding='SAME')
mask = tf.stop_gradient(mask)
net = tf.nn.max_pool(net, ksize=[1, stride, stride, 1],strides=[1, stride, stride, 1], padding='SAME')
return net, mask
#4*4----2*2--=2*2 【6，8，12，16】
#反池化
def unpool(net, mask, stride):
ksize = [1, stride, stride, 1]
input_shape = net.get_shape().as_list()
output_shape = (input_shape[0], input_shape[1] * ksize[1], input_shape[2] * ksize[2], input_shape[3])
one_like_mask = tf.ones_like(mask)
batch_range = tf.reshape(tf.range(output_shape[0], dtype=tf.int64), shape=[input_shape[0], 1, 1, 1])
b = one_like_mask * batch_range
y = mask // (output_shape[2] * output_shape[3])
x = mask % (output_shape[2] * output_shape[3]) // output_shape[3]
feature_range = tf.range(output_shape[3], dtype=tf.int64)
f = one_like_mask * feature_range
updates_size = tf.size(net)
indices = tf.transpose(tf.reshape(tf.stack([b, y, x, f]), [4, updates_size]))
values = tf.reshape(net, [updates_size])
ret = tf.scatter_nd(indices, values, output_shape)
return ret
batch_size = 128
data_dir = '../cifar-10-batches-bin'
print("begin")
images_train, labels_train = cifar10_input.inputs(eval_data = False,data_dir = data_dir, batch_size = batch_size)
images_test, labels_test = cifar10_input.inputs(eval_data = True, data_dir = data_dir, batch_size = batch_size)
print("begin data")
def weight_variable(shape):
initial = tf.truncated_normal(shape, stddev=0.01)
return tf.Variable(initial)
def bias_variable(shape):
initial = tf.constant(0.01, 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')
def avg_pool_6x6(x):
return tf.nn.avg_pool(x, ksize=[1, 6, 6, 1], strides=[1, 6, 6, 1], padding='SAME')
# tf Graph Input
x = tf.placeholder(tf.float32, [batch_size, 24,24,3]) # cifar data image of shape 24*24*3
y = tf.placeholder(tf.float32, [batch_size, 10]) # 0-9 數字=> 10 classes
W_conv1 = weight_variable([5, 5, 3, 64])
b_conv1 = bias_variable([64])
x_image = tf.reshape(x, [-1,24,24,3])
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
#h_pool1 = max_pool_2x2(h_conv1)
h_pool1, mask1 = max_pool_with_argmax(h_conv1, 2)
W_conv2 = weight_variable([5, 5, 64, 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)
#############################################################
h_pool2, mask = max_pool_with_argmax(h_conv2, 2)#(128, 6, 6, 64)
print(h_pool2.shape)
t_conv2 = unpool(h_pool2, mask, 2)#(128, 12, 12, 64)
t_pool1 = tf.nn.conv2d_transpose(t_conv2-b_conv2, W_conv2, h_pool1.shape,[1,1,1,1])#(128, 24, 24, 64)
print(t_conv2.shape,h_pool1.shape,t_pool1.shape)
t_conv1 = unpool(t_pool1, mask1, 2)
t_x_image = tf.nn.conv2d_transpose(t_conv1-b_conv1, W_conv1, x_image.shape,[1,1,1,1])
#第一層卷積還原
t1_conv1 = unpool(h_pool1, mask1, 2)
t1_x_image = tf.nn.conv2d_transpose(t1_conv1-b_conv1, W_conv1, x_image.shape,[1,1,1,1])
# 生成最終圖像
stitched_decodings = tf.concat((x_image, t1_x_image,t_x_image), axis=2)
decoding_summary_op = tf.summary.image('source/cifar', stitched_decodings)
#############################################################
W_conv3 = weight_variable([5, 5, 64, 10])
b_conv3 = bias_variable([10])
h_conv3 = tf.nn.relu(conv2d(h_pool2, W_conv3) + b_conv3)
nt_hpool3=avg_pool_6x6(h_conv3)#10
nt_hpool3_flat = tf.reshape(nt_hpool3, [-1, 10])
y_conv=tf.nn.softmax(nt_hpool3_flat)
cross_entropy = -tf.reduce_sum(y*tf.log(y_conv)) +(tf.nn.l2_loss(W_conv1)+tf.nn.l2_loss(W_conv2)+tf.nn.l2_loss(W_conv3))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
sess = tf.Session()
sess.run(tf.global_variables_initializer())
summary_writer = tf.summary.FileWriter('./log/', sess.graph)
tf.train.start_queue_runners(sess=sess)
for i in range(15000):#20000
image_batch, label_batch = sess.run([images_train, labels_train])
label_b = np.eye(10,dtype=float)[label_batch] #one hot
train_step.run(feed_dict={x:image_batch, y: label_b},session=sess)
#_, decoding_summary = sess.run([train_step, decoding_summary_op],feed_dict={x:image_batch, y: label_b})
if i%200 == 0:
train_accuracy = accuracy.eval(feed_dict={
x:image_batch, y: label_b},session=sess)
print("step %d, training accuracy %g"%(i, train_accuracy))
print("cross_entropy",cross_entropy.eval(feed_dict={x:image_batch, y: label_b},session=sess))
#summary_writer.add_summary(decoding_summary)
image_batch, label_batch = sess.run([images_test, labels_test])
label_b = np.eye(10,dtype=float)[label_batch]#one hot
print ("finished！ test accuracy %g"%accuracy.eval(feed_dict={
x:image_batch, y: label_b},session=sess))
decoding_summary = sess.run(decoding_summary_op,feed_dict={x:image_batch, y: label_b})
summary_writer.add_summary(decoding_summary)
step 0, training accuracy 0.109375 cross_entropy 299.33258 step 200, training accuracy 0.265625 cross_entropy 255.26352 step 400, training accuracy 0.328125 cross_entropy 220.17691 step 600, training accuracy 0.304688 cross_entropy 239.9032 step 800, training accuracy 0.4375 cross_entropy 215.55 step 1000, training accuracy 0.34375 cross_entropy 240.33012 step 1200, training accuracy 0.507812 cross_entropy 208.6071 step 1400, training accuracy 0.414062 cross_entropy 215.67905 step 1600, training accuracy 0.484375 cross_entropy 213.92612 step 1800, training accuracy 0.335938 cross_entropy 233.1239 step 2000, training accuracy 0.5 cross_entropy 208.82675 step 2200, training accuracy 0.492188 cross_entropy 191.40732 step 2400, training accuracy 0.460938 cross_entropy 193.65083 step 2600, training accuracy 0.492188 cross_entropy 186.18613 step 2800, training accuracy 0.484375 cross_entropy 189.64601 step 3000, training accuracy 0.539062 cross_entropy 183.80658 step 3200, training accuracy 0.421875 cross_entropy 204.8297 step 3400, training accuracy 0.476562 cross_entropy 193.53462 step 3600, training accuracy 0.515625 cross_entropy 204.49474 step 3800, training accuracy 0.523438 cross_entropy 194.06113 step 4000, training accuracy 0.53125 cross_entropy 194.76062 step 4200, training accuracy 0.476562 cross_entropy 191.30899 step 4400, training accuracy 0.484375 cross_entropy 198.17056 step 4600, training accuracy 0.523438 cross_entropy 176.11752 step 4800, training accuracy 0.5 cross_entropy 185.19417 step 5000, training accuracy 0.570312 cross_entropy 188.2881 step 5200, training accuracy 0.515625 cross_entropy 184.50743 step 5400, training accuracy 0.632812 cross_entropy 175.47253 step 5600, training accuracy 0.515625 cross_entropy 193.34174 step 5800, training accuracy 0.554688 cross_entropy 181.11494 step 6000, training accuracy 0.53125 cross_entropy 172.79439 step 6200, training accuracy 0.65625 cross_entropy 153.55139 step 6400, training accuracy 0.523438 cross_entropy 183.32246 step 6600, training accuracy 0.570312 cross_entropy 166.75311 step 6800, training accuracy 0.539062 cross_entropy 192.8255 step 7000, training accuracy 0.585938 cross_entropy 172.40381 step 7200, training accuracy 0.625 cross_entropy 154.44427 step 7400, training accuracy 0.523438 cross_entropy 173.51205 step 7600, training accuracy 0.617188 cross_entropy 158.45212 step 7800, training accuracy 0.515625 cross_entropy 194.20056 step 8000, training accuracy 0.59375 cross_entropy 177.66724 step 8200, training accuracy 0.476562 cross_entropy 190.3621 step 8400, training accuracy 0.59375 cross_entropy 173.1125 step 8600, training accuracy 0.53125 cross_entropy 181.5401 step 8800, training accuracy 0.570312 cross_entropy 160.98604 step 9000, training accuracy 0.59375 cross_entropy 169.50435 step 9200, training accuracy 0.65625 cross_entropy 159.84276 step 9400, training accuracy 0.53125 cross_entropy 185.61572 step 9600, training accuracy 0.609375 cross_entropy 163.57227 step 9800, training accuracy 0.539062 cross_entropy 172.54062 step 10000, training accuracy 0.570312 cross_entropy 182.86372 step 10200, training accuracy 0.5625 cross_entropy 175.23633 step 10400, training accuracy 0.625 cross_entropy 158.63727 step 10600, training accuracy 0.570312 cross_entropy 183.32611 step 10800, training accuracy 0.539062 cross_entropy 186.68808 step 11000, training accuracy 0.554688 cross_entropy 180.54099 step 11200, training accuracy 0.523438 cross_entropy 187.641 step 11400, training accuracy 0.601562 cross_entropy 157.28458 step 11600, training accuracy 0.5625 cross_entropy 177.28255 step 11800, training accuracy 0.601562 cross_entropy 180.65999 step 12000, training accuracy 0.601562 cross_entropy 155.04745 step 12200, training accuracy 0.625 cross_entropy 164.97495 step 12400, training accuracy 0.554688 cross_entropy 160.26328 step 12600, training accuracy 0.609375 cross_entropy 169.59125 step 12800, training accuracy 0.578125 cross_entropy 181.78036 step 13000, training accuracy 0.65625 cross_entropy 155.21802 step 13200, training accuracy 0.507812 cross_entropy 183.50308 step 13400, training accuracy 0.617188 cross_entropy 162.36247 step 13600, training accuracy 0.585938 cross_entropy 172.93718 step 13800, training accuracy 0.617188 cross_entropy 167.09555 step 14000, training accuracy 0.664062 cross_entropy 153.10864step 14200, training accuracy 0.648438 cross_entropy 154.17062 step 14400, training accuracy 0.546875 cross_entropy 185.39635 step 14600, training accuracy 0.640625 cross_entropy 152.71246 step 14800, training accuracy 0.617188 cross_entropy 162.1098 finished！ test accuracy 0.601562
實例51: CIFAR封裝代碼
使用tf.contrib.layers重構
# -*- coding: utf-8 -*-
import cifar10_input
import tensorflow as tf
import numpy as np
batch_size = 128
data_dir = '../cifar-10-batches-bin'
print("begin")
images_train, labels_train = cifar10_input.inputs(eval_data = False,data_dir = data_dir, batch_size = batch_size)
images_test, labels_test = cifar10_input.inputs(eval_data = True, data_dir = data_dir, batch_size = batch_size)
print("begin data")
# tf Graph Input
x = tf.placeholder(tf.float32, [None, 24,24,3]) # cifar data image of shape 24*24*3
y = tf.placeholder(tf.float32, [None, 10]) # 0-9 數字=> 10 classes
x_image = tf.reshape(x, [-1,24,24,3])
h_conv1 =tf.contrib.layers.conv2d(x_image,64,[5,5],1,'SAME',activation_fn=tf.nn.relu)
h_pool1 = tf.contrib.layers.max_pool2d(h_conv1,[2,2],stride=2,padding='SAME')
h_conv2 =tf.contrib.layers.conv2d(h_pool1,64,[5,5],1,'SAME',activation_fn=tf.nn.relu)
h_pool2 = tf.contrib.layers.max_pool2d(h_conv2,[2,2],stride=2,padding='SAME')
nt_hpool2 = tf.contrib.layers.avg_pool2d(h_pool2,[6,6],stride=6,padding='SAME')
nt_hpool2_flat = tf.reshape(nt_hpool2, [-1, 64])
y_conv = tf.contrib.layers.fully_connected(nt_hpool2_flat,10,activation_fn=tf.nn.softmax)
cross_entropy = -tf.reduce_sum(y*tf.log(y_conv))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
sess = tf.Session()
sess.run(tf.global_variables_initializer())
tf.train.start_queue_runners(sess=sess)
for i in range(15000):#20000
image_batch, label_batch = sess.run([images_train, labels_train])
label_b = np.eye(10,dtype=float)[label_batch] #one hot
train_step.run(feed_dict={x:image_batch, y: label_b},session=sess)
if i%200 == 0:
train_accuracy = accuracy.eval(feed_dict={
x:image_batch, y: label_b},session=sess)
print( "step %d, training accuracy %g"%(i, train_accuracy))
image_batch, label_batch = sess.run([images_test, labels_test])
label_b = np.eye(10,dtype=float)[label_batch]#one hot
print ("finished！ test accuracy %g"%accuracy.eval(feed_dict={
x:image_batch, y: label_b},session=sess))
step 0, training accuracy 0.148438 step 200, training accuracy 0.328125 step 400, training accuracy 0.421875 step 600, training accuracy 0.34375 step 800, training accuracy 0.382812 step 1000, training accuracy 0.414062 step 1200, training accuracy 0.5 step 1400, training accuracy 0.414062 step 1600, training accuracy 0.476562 step 1800, training accuracy 0.398438 step 2000, training accuracy 0.554688 step 2200, training accuracy 0.40625 step 2400, training accuracy 0.5625 step 2600, training accuracy 0.523438 step 2800, training accuracy 0.46875 step 3000, training accuracy 0.53125 step 3200, training accuracy 0.40625 step 3400, training accuracy 0.523438 step 3600, training accuracy 0.53125 step 3800, training accuracy 0.515625 step 4000, training accuracy 0.492188 step 4200, training accuracy 0.515625 step 4400, training accuracy 0.5 step 4600, training accuracy 0.617188 step 4800, training accuracy 0.546875 step 5000, training accuracy 0.539062 step 5200, training accuracy 0.578125 step 5400, training accuracy 0.5625 step 5600, training accuracy 0.492188 step 5800, training accuracy 0.554688 step 6000, training accuracy 0.632812 step 6200, training accuracy 0.523438 step 6400, training accuracy 0.523438 step 6600, training accuracy 0.523438 step 6800, training accuracy 0.492188 step 7000, training accuracy 0.570312 step 7200, training accuracy 0.523438 step 7400, training accuracy 0.601562 step 7600, training accuracy 0.601562 step 7800, training accuracy 0.570312 step 8000, training accuracy 0.625 step 8200, training accuracy 0.5 step 8400, training accuracy 0.609375 step 8600, training accuracy 0.601562 step 8800, training accuracy 0.539062 step 9000, training accuracy 0.578125 step 9200, training accuracy 0.609375 step 9400, training accuracy 0.546875 step 9600, training accuracy 0.609375 step 9800, training accuracy 0.601562 step 10000, training accuracy 0.601562 step 10200, training accuracy 0.5625 step 10400, training accuracy 0.6875 step 10600, training accuracy 0.609375 step 10800, training accuracy 0.570312 step 11000, training accuracy 0.578125 step 11200, training accuracy 0.53125 step 11400, training accuracy 0.65625 step 11600, training accuracy 0.609375 step 11800, training accuracy 0.5625 step 12000, training accuracy 0.679688 step 12200, training accuracy 0.632812 step 12400, training accuracy 0.625 step 12600, training accuracy 0.65625 step 12800, training accuracy 0.632812 step 13000, training accuracy 0.585938 step 13200, training accuracy 0.539062 step 13400, training accuracy 0.640625 step 13600, training accuracy 0.617188 step 13800, training accuracy 0.679688 step 14000, training accuracy 0.664062 step 14200, training accuracy 0.609375 step 14400, training accuracy 0.585938 step 14600, training accuracy 0.65625 step 14800, training accuracy 0.609375 finished！ test accuracy 0.617188
實例52：cifar卷積核優化
# -*- coding: utf-8 -*-
import cifar10_input
import tensorflow as tf
import numpy as np
batch_size = 128
data_dir = '../cifar-10-batches-bin'
print("begin")
images_train, labels_train = cifar10_input.inputs(eval_data = False,data_dir = data_dir, batch_size = batch_size)
images_test, labels_test = cifar10_input.inputs(eval_data = True, data_dir = data_dir, batch_size = batch_size)
print("begin data")
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')
def avg_pool_6x6(x):
return tf.nn.avg_pool(x, ksize=[1, 6, 6, 1],
strides=[1, 6, 6, 1], padding='SAME')
# tf Graph Input
x = tf.placeholder(tf.float32, [None, 24,24,3]) # cifar data image of shape 24*24*3
y = tf.placeholder(tf.float32, [None, 10]) # 0-9 數字=> 10 classes
W_conv1 = weight_variable([5, 5, 3, 64])
b_conv1 = bias_variable([64])
x_image = tf.reshape(x, [-1,24,24,3])
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)
#########################################################new########
W_conv21 = weight_variable([5, 1, 64, 64])
b_conv21 = bias_variable([64])
h_conv21 = tf.nn.relu(conv2d(h_pool1, W_conv21) + b_conv21)
W_conv2 = weight_variable([1, 5, 64, 64])
b_conv2 = bias_variable([64])
h_conv2 = tf.nn.relu(conv2d(h_conv21, W_conv2) + b_conv2)
###########################################################old#########
#W_conv2 = weight_variable([5, 5, 64, 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)
W_conv3 = weight_variable([5, 5, 64, 10])
b_conv3 = bias_variable([10])
h_conv3 = tf.nn.relu(conv2d(h_pool2, W_conv3) + b_conv3)
nt_hpool3=avg_pool_6x6(h_conv3)#10
nt_hpool3_flat = tf.reshape(nt_hpool3, [-1, 10])
y_conv=tf.nn.softmax(nt_hpool3_flat)
cross_entropy = -tf.reduce_sum(y*tf.log(y_conv))
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
sess = tf.Session()
sess.run(tf.global_variables_initializer())
tf.train.start_queue_runners(sess=sess)
for i in range(15000):#20000
image_batch, label_batch = sess.run([images_train, labels_train])
label_b = np.eye(10,dtype=float)[label_batch] #one hot
train_step.run(feed_dict={x:image_batch, y: label_b},session=sess)
if i%200 == 0:
train_accuracy = accuracy.eval(feed_dict={
x:image_batch, y: label_b},session=sess)
print( "step %d, training accuracy %g"%(i, train_accuracy))
image_batch, label_batch = sess.run([images_test, labels_test])
label_b = np.eye(10,dtype=float)[label_batch]#one hot
print ("finished！ test accuracy %g"%accuracy.eval(feed_dict={
x:image_batch, y: label_b},session=sess))
step 0, training accuracy 0.125 step 200, training accuracy 0.34375 step 400, training accuracy 0.375 step 600, training accuracy 0.367188 step 800, training accuracy 0.507812 step 1000, training accuracy 0.382812 step 1200, training accuracy 0.429688 step 1400, training accuracy 0.5 step 1600, training accuracy 0.4375 step 1800, training accuracy 0.390625 step 2000, training accuracy 0.515625 step 2200, training accuracy 0.59375 step 2400, training accuracy 0.554688 step 2600, training accuracy 0.539062 step 2800, training accuracy 0.53125 step 3000, training accuracy 0.5 step 3200, training accuracy 0.414062 step 3400, training accuracy 0.515625 step 3600, training accuracy 0.515625 step 3800, training accuracy 0.476562 step 4000, training accuracy 0.585938 step 4200, training accuracy 0.5625 step 4400, training accuracy 0.570312 step 4600, training accuracy 0.523438 step 4800, training accuracy 0.523438 step 5000, training accuracy 0.5625 step 5200, training accuracy 0.53125 step 5400, training accuracy 0.59375 step 5600, training accuracy 0.53125 step 5800, training accuracy 0.601562 step 6000, training accuracy 0.585938 step 6200, training accuracy 0.617188 step 6400, training accuracy 0.59375 step 6600, training accuracy 0.617188 step 6800, training accuracy 0.570312 step 7000, training accuracy 0.585938 step 7200, training accuracy 0.664062 step 7400, training accuracy 0.648438 step 7600, training accuracy 0.65625 step 7800, training accuracy 0.632812 step 8000, training accuracy 0.65625 step 8200, training accuracy 0.585938 step 8400, training accuracy 0.578125 step 8600, training accuracy 0.585938 step 8800, training accuracy 0.578125 step 9000, training accuracy 0.695312 step 9200, training accuracy 0.570312 step 9400, training accuracy 0.546875 step 9600, training accuracy 0.742188 step 9800, training accuracy 0.6875 step 10000, training accuracy 0.59375 step 10200, training accuracy 0.5625 step 10400, training accuracy 0.679688 step 10600, training accuracy 0.617188 step 10800, training accuracy 0.65625 step 11000, training accuracy 0.578125 step 11200, training accuracy 0.679688 step 11400, training accuracy 0.617188 step 11600, training accuracy 0.609375 step 11800, training accuracy 0.632812 step 12000, training accuracy 0.671875 step 12200, training accuracy 0.664062 step 12400, training accuracy 0.609375 step 12600, training accuracy 0.710938 step 12800, training accuracy 0.617188 step 13000, training accuracy 0.679688 step 13200, training accuracy 0.625 step 13400, training accuracy 0.632812 step 13600, training accuracy 0.554688 step 13800, training accuracy 0.65625 step 14000, training accuracy 0.679688 step 14200, training accuracy 0.710938 step 14400, training accuracy 0.554688 step 14600, training accuracy 0.75 step 14800, training accuracy 0.632812 finished！ test accuracy 0.609375
實例53:cifar多通道卷
# -*- coding: utf-8 -*-
import cifar10_input
import tensorflow as tf
import numpy as np
batch_size = 128
data_dir = '../cifar-10-batches-bin'
print("begin")
images_train, labels_train = cifar10_input.inputs(eval_data = False,data_dir = data_dir, batch_size = batch_size)
images_test, labels_test = cifar10_input.inputs(eval_data = True, data_dir = data_dir, batch_size = batch_size)
print("begin data")
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')
def avg_pool_6x6(x):
return tf.nn.avg_pool(x, ksize=[1, 6, 6, 1],
strides=[1, 6, 6, 1], padding='SAME')
# tf Graph Input
x = tf.placeholder(tf.float32, [None, 24,24,3]) # cifar data image of shape 24*24*3
y = tf.placeholder(tf.float32, [None, 10]) # 0-9 數字=> 10 classes
W_conv1 = weight_variable([5, 5, 3, 64])
b_conv1 = bias_variable([64])
x_image = tf.reshape(x, [-1,24,24,3])
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)
#######################################################多卷積核
W_conv2_5x5 = weight_variable([5, 5, 64, 64])
b_conv2_5x5 = bias_variable([64])
W_conv2_7x7 = weight_variable([7, 7, 64, 64])
b_conv2_7x7 = bias_variable([64])
W_conv2_3x3 = weight_variable([3, 3, 64, 64])
b_conv2_3x3 = bias_variable([64])
W_conv2_1x1 = weight_variable([1, 1, 64, 64])
b_conv2_1x1 = bias_variable([64])
h_conv2_1x1 = tf.nn.relu(conv2d(h_pool1, W_conv2_1x1) + b_conv2_1x1)
h_conv2_3x3 = tf.nn.relu(conv2d(h_pool1, W_conv2_3x3) + b_conv2_3x3)
h_conv2_5x5 = tf.nn.relu(conv2d(h_pool1, W_conv2_5x5) + b_conv2_5x5)
h_conv2_7x7 = tf.nn.relu(conv2d(h_pool1, W_conv2_7x7) + b_conv2_7x7)
h_conv2 = tf.concat([h_conv2_5x5,h_conv2_7x7,h_conv2_3x3,h_conv2_1x1],3)
#######################################################
#W_conv2 = weight_variable([5, 5, 64, 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)
#######################################################
W_conv3 = weight_variable([5, 5, 256, 10])
b_conv3 = bias_variable([10])
h_conv3 = tf.nn.relu(conv2d(h_pool2, W_conv3) + b_conv3)
nt_hpool3=avg_pool_6x6(h_conv3)#10
nt_hpool3_flat = tf.reshape(nt_hpool3, [-1, 10])
y_conv=tf.nn.softmax(nt_hpool3_flat)
cross_entropy = -tf.reduce_sum(y*tf.log(y_conv))
#不同的優化方法測測效果
#train_step = tf.train.GradientDescentOptimizer(1e-3).minimize(cross_entropy)
#train_step = tf.train.AdagradOptimizer(1e-5).minimize(cross_entropy)
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
sess = tf.Session()
sess.run(tf.global_variables_initializer())
tf.train.start_queue_runners(sess=sess)
for i in range(15000):#20000
image_batch, label_batch = sess.run([images_train, labels_train])
label_b = np.eye(10,dtype=float)[label_batch] #one hot
train_step.run(feed_dict={x:image_batch, y: label_b},session=sess)
if i%200 == 0:
train_accuracy = accuracy.eval(feed_dict={
x:image_batch, y: label_b},session=sess)
print( "step %d, training accuracy %g"%(i, train_accuracy))
image_batch, label_batch = sess.run([images_test, labels_test])
label_b = np.eye(10,dtype=float)[label_batch]#one hot
print ("finished！ test accuracy %g"%accuracy.eval(feed_dict={
x:image_batch, y: label_b},session=sess))
step 0, training accuracy 0.09375 step 200, training accuracy 0.304688 step 400, training accuracy 0.398438 step 600, training accuracy 0.4375 step 800, training accuracy 0.507812 step 1000, training accuracy 0.46875 step 1200, training accuracy 0.421875 step 1400, training accuracy 0.585938 step 1600, training accuracy 0.585938 step 1800, training accuracy 0.484375 step 2000, training accuracy 0.546875 step 2200, training accuracy 0.539062 step 2400, training accuracy 0.625 step 2600, training accuracy 0.570312 step 2800, training accuracy 0.570312 step 3000, training accuracy 0.546875 step 3200, training accuracy 0.601562 step 3400, training accuracy 0.609375 step 3600, training accuracy 0.609375 step 3800, training accuracy 0.625 step 4000, training accuracy 0.671875 step 4200, training accuracy 0.625 step 4400, training accuracy 0.601562 step 4600, training accuracy 0.609375 step 4800, training accuracy 0.570312 step 5000, training accuracy 0.65625 step 5200, training accuracy 0.640625 step 5400, training accuracy 0.671875 step 5600, training accuracy 0.5625 step 5800, training accuracy 0.65625 step 6000, training accuracy 0.695312 step 6200, training accuracy 0.757812 step 6400, training accuracy 0.648438 step 6600, training accuracy 0.734375 step 6800, training accuracy 0.632812 step 7000, training accuracy 0.6875 step 7200, training accuracy 0.679688 step 7400, training accuracy 0.695312 step 7600, training accuracy 0.75 step 7800, training accuracy 0.65625 step 8000, training accuracy 0.71875 step 8200, training accuracy 0.617188 step 8400, training accuracy 0.632812 step 8600, training accuracy 0.679688 step 8800, training accuracy 0.679688 step 9000, training accuracy 0.625 step 9200, training accuracy 0.710938 step 9400, training accuracy 0.625 step 9600, training accuracy 0.765625 step 9800, training accuracy 0.679688 step 10000, training accuracy 0.710938 step 10200, training accuracy 0.710938 step 10400, training accuracy 0.71875 step 10600, training accuracy 0.679688 step 10800, training accuracy 0.671875 step 11000, training accuracy 0.75 step 11200, training accuracy 0.609375 step 11400, training accuracy 0.6875 step 11600, training accuracy 0.695312 step 11800, training accuracy 0.6875 step 12000, training accuracy 0.75 step 12200, training accuracy 0.773438 step 12400, training accuracy 0.71875 step 12600, training accuracy 0.671875 step 12800, training accuracy 0.679688 step 13000, training accuracy 0.664062 step 13200, training accuracy 0.710938 step 13400, training accuracy 0.773438 step 13600, training accuracy 0.734375 step 13800, training accuracy 0.6875 step 14000, training accuracy 0.75 step 14200, training accuracy 0.765625 step 14400, training accuracy 0.757812 step 14600, training accuracy 0.796875 step 14800, training accuracy 0.773438 finished！ test accuracy 0.625
實例54：BN
# -*- coding: utf-8 -*-
import cifar10_input
import tensorflow as tf
import numpy as np
from tensorflow.contrib.layers.python.layers import batch_norm
batch_size = 128
data_dir = '../cifar-10-batches-bin'
print("begin")
images_train, labels_train = cifar10_input.inputs(eval_data = False,data_dir = data_dir, batch_size = batch_size)
images_test, labels_test = cifar10_input.inputs(eval_data = True, data_dir = data_dir, batch_size = batch_size)
print("begin data")
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')
def avg_pool_6x6(x):
return tf.nn.avg_pool(x, ksize=[1, 6, 6, 1],
strides=[1, 6, 6, 1], padding='SAME')
def batch_norm_layer(value,train = None, name = 'batch_norm'):
if train is not None:
return batch_norm(value, decay = 0.9,updates_collections=None, is_training = True)
else:
return batch_norm(value, decay = 0.9,updates_collections=None, is_training = False)
# tf Graph Input
x = tf.placeholder(tf.float32, [None, 24,24,3]) # cifar data image of shape 24*24*3
y = tf.placeholder(tf.float32, [None, 10]) # 0-9 數字=> 10 classes
train = tf.placeholder(tf.float32)
W_conv1 = weight_variable([5, 5, 3, 64])
b_conv1 = bias_variable([64])
x_image = tf.reshape(x, [-1,24,24,3])
h_conv1 = tf.nn.relu(batch_norm_layer((conv2d(x_image, W_conv1) + b_conv1),train))
h_pool1 = max_pool_2x2(h_conv1)
W_conv2 = weight_variable([5, 5, 64, 64])
b_conv2 = bias_variable([64])
h_conv2 = tf.nn.relu(batch_norm_layer((conv2d(h_pool1, W_conv2) + b_conv2),train))
h_pool2 = max_pool_2x2(h_conv2)
W_conv3 = weight_variable([5, 5, 64, 10])
b_conv3 = bias_variable([10])
h_conv3 = tf.nn.relu(conv2d(h_pool2, W_conv3) + b_conv3)
nt_hpool3=avg_pool_6x6(h_conv3)#10
nt_hpool3_flat = tf.reshape(nt_hpool3, [-1, 10])
y_conv=tf.nn.softmax(nt_hpool3_flat)
cross_entropy = -tf.reduce_sum(y*tf.log(y_conv))
global_step = tf.Variable(0, trainable=False)
decaylearning_rate = tf.train.exponential_decay(0.04, global_step,1000, 0.9)
train_step = tf.train.AdamOptimizer(decaylearning_rate).minimize(cross_entropy,global_step=global_step)
correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
sess = tf.Session()
sess.run(tf.global_variables_initializer())
tf.train.start_queue_runners(sess=sess)
for i in range(20000):
image_batch, label_batch = sess.run([images_train, labels_train])
label_b = np.eye(10,dtype=float)[label_batch] #one hot
train_step.run(feed_dict={x:image_batch, y: label_b,train:1},session=sess)
if i%200 == 0:
train_accuracy = accuracy.eval(feed_dict={
x:image_batch, y: label_b},session=sess)
print( "step %d, training accuracy %g"%(i, train_accuracy))
image_batch, label_batch = sess.run([images_test, labels_test])
label_b = np.eye(10,dtype=float)[label_batch]#one hot
print ("finished！ test accuracy %g"%accuracy.eval(feed_dict={
x:image_batch, y: label_b},session=sess))
step 0, training accuracy 0.15625 step 200, training accuracy 0.078125 step 400, training accuracy 0.117188 step 600, training accuracy 0.0859375 step 800, training accuracy 0.0625 step 1000, training accuracy 0.125 step 1200, training accuracy 0.109375 step 1400, training accuracy 0.09375 step 1600, training accuracy 0.078125 step 1800, training accuracy 0.0859375 step 2000, training accuracy 0.109375 step 2200, training accuracy 0.0859375 step 2400, training accuracy 0.117188 step 2600, training accuracy 0.09375 step 2800, training accuracy 0.0625 step 3000, training accuracy 0.09375 step 3200, training accuracy 0.109375 step 3400, training accuracy 0.0625 step 3600, training accuracy 0.109375 step 3800, training accuracy 0.078125 step 4000, training accuracy 0.125 step 4200, training accuracy 0.09375 step 4400, training accuracy 0.117188 step 4600, training accuracy 0.125 step 4800, training accuracy 0.101562 step 5000, training accuracy 0.09375 step 5200, training accuracy 0.078125 step 5400, training accuracy 0.132812 step 5600, training accuracy 0.078125 step 5800, training accuracy 0.117188 step 6000, training accuracy 0.132812 step 6200, training accuracy 0.109375 step 6400, training accuracy 0.101562 step 6600, training accuracy 0.078125 step 6800, training accuracy 0.09375 step 7000, training accuracy 0.09375 step 7200, training accuracy 0.117188 step 7400, training accuracy 0.125 step 7600, training accuracy 0.109375 step 7800, training accuracy 0.101562 step 8000, training accuracy 0.109375 step 8200, training accuracy 0.109375 step 8400, training accuracy 0.078125 step 8600, training accuracy 0.078125 step 8800, training accuracy 0.09375 step 9000, training accuracy 0.0859375 step 9200, training accuracy 0.109375 step 9400, training accuracy 0.117188 step 9600, training accuracy 0.0859375 step 9800, training accuracy 0.109375 step 10000, training accuracy 0.09375 step 10200, training accuracy 0.109375 step 10400, training accuracy 0.117188 step 10600, training accuracy 0.09375 step 10800, training accuracy 0.0625 step 11000, training accuracy 0.132812 step 11200, training accuracy 0.078125 step 11400, training accuracy 0.101562 step 11600, training accuracy 0.109375 step 11800, training accuracy 0.117188 step 12000, training accuracy 0.0859375 step 12200, training accuracy 0.0703125 step 12400, training accuracy 0.117188 step 12600, training accuracy 0.078125 step 12800, training accuracy 0.078125 step 13000, training accuracy 0.148438 step 13200, training accuracy 0.101562 step 13400, training accuracy 0.140625 step 13600, training accuracy 0.0859375 step 13800, training accuracy 0.101562 step 14000, training accuracy 0.09375 step 14200, training accuracy 0.09375 step 14400, training accuracy 0.09375 step 14600, training accuracy 0.125 step 14800, training accuracy 0.109375 step 15000, training accuracy 0.09375 step 15200, training accuracy 0.0859375 step 15400, training accuracy 0.109375 step 15600, training accuracy 0.0859375 step 15800, training accuracy 0.078125 step 16000, training accuracy 0.125 step 16200, training accuracy 0.125 step 16400, training accuracy 0.101562 step 16600, training accuracy 0.101562 step 16800, training accuracy 0.132812 step 17000, training accuracy 0.117188 step 17200, training accuracy 0.0859375 step 17400, training accuracy 0.0859375 step 17600, training accuracy 0.0859375 step 17800, training accuracy 0.09375 step 18000, training accuracy 0.109375 step 18200, training accuracy 0.101562 step 18400, training accuracy 0.0703125 step 18600, training accuracy 0.109375 step 18800, training accuracy 0.09375 step 19000, training accuracy 0.09375 step 19200, training accuracy 0.03125 step 19400, training accuracy 0.0546875 step 19600, training accuracy 0.109375 step 19800, training accuracy 0.0625 finished！ test accuracy 0.101562
帶BN的多通道mnist
# -*- coding: utf-8 -*-
import tensorflow as tf
import numpy as np
from tensorflow.contrib.layers.python.layers import batch_norm
# 導入 MINST 數據集
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/data/", one_hot=True)
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')
def avg_pool_7x7(x):
return tf.nn.avg_pool(x, ksize=[1, 7, 7, 1],
strides=[1, 7, 7, 1], padding='SAME')
def batch_norm_layer(value,train = None, name = 'batch_norm'):
if train is not None:
return batch_norm(value, decay = 0.9,updates_collections=None, is_training = True)
else:
return batch_norm(value, decay = 0.9,updates_collections=None, is_training = False)
# tf Graph Input
x = tf.placeholder(tf.float32, [None, 784]) # mnist data維度 28*28=784
y = tf.placeholder(tf.float32, [None, 10]) # 0-9 數字=> 10 classes
train = tf.placeholder(tf.float32)
W_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
x_image = tf.reshape(x, [-1,28,28,1])
h_conv1 = tf.nn.relu(batch_norm_layer((conv2d(x_image, W_conv1) + b_conv1),train))
h_pool1 = max_pool_2x2(h_conv1)
######################################################多卷積核
W_conv2_5x5 = weight_variable([5, 5, 32, 32])
b_conv2_5x5 = bias_variable([32])
W_conv2_7x7 = weight_variable([7, 7, 32, 32])
b_conv2_7x7 = bias_variable([32])
h_conv2_5x5 = tf.nn.relu(batch_norm_layer((conv2d(h_pool1, W_conv2_5x5) + b_conv2_5x5),train))
h_conv2_7x7 = tf.nn.relu(batch_norm_layer((conv2d(h_pool1, W_conv2_7x7) + b_conv2_7x7),train))
h_conv2 = tf.concat([h_conv2_5x5,h_conv2_7x7],3)
h_pool2 = max_pool_2x2(h_conv2)
#########################################################new 池化
W_conv3 = weight_variable([5, 5, 64, 10])
b_conv3 = bias_variable([10])
h_conv3 = tf.nn.relu(conv2d(h_pool2, W_conv3) + b_conv3)
nt_hpool3=avg_pool_7x7(h_conv3)#10
nt_hpool3_flat = tf.reshape(nt_hpool3, [-1, 10])
y_conv=tf.nn.softmax(nt_hpool3_flat)
keep_prob = tf.placeholder("float")
cross_entropy = -tf.reduce_sum(y*tf.log(y_conv))
decaylearning_rate = tf.train.exponential_decay(0.04, 20000,1000, 0.9)
train_step = tf.train.AdamOptimizer(decaylearning_rate).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
# 啓動session
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
for i in range(20000):#20000
batch = mnist.train.next_batch(50)
if i%100 == 0:
train_accuracy = accuracy.eval(feed_dict={
x:batch[0], y: batch[1], keep_prob: 1.0})
print( "step %d, training accuracy %g"%(i, train_accuracy))
train_step.run(feed_dict={x: batch[0], y: batch[1], keep_prob: 0.5})
print ("test accuracy %g"%accuracy.eval(feed_dict={
x: mnist.test.images, y: mnist.test.labels, keep_prob: 1.0}))
step 0, training accuracy 0.04 step 100, training accuracy 0.78 step 200, training accuracy 0.98 step 300, training accuracy 0.94 step 400, training accuracy 0.98 step 500, training accuracy 0.98 step 600, training accuracy 0.98 step 700, training accuracy 1 step 800, training accuracy 0.98 step 900, training accuracy 0.94 step 1000, training accuracy 0.96 step 1100, training accuracy 0.98 step 1200, training accuracy 0.94 step 1300, training accuracy 0.94 step 1400, training accuracy 1 step 1500, training accuracy 0.98 step 1600, training accuracy 0.98 step 1700, training accuracy 0.98 step 1800, training accuracy 0.98 step 1900, training accuracy 0.98 step 2000, training accuracy 0.98 step 2100, training accuracy 1 step 2200, training accuracy 0.96 step 2300, training accuracy 0.96 step 2400, training accuracy 1 step 2500, training accuracy 0.96 step 2600, training accuracy 0.98 step 2700, training accuracy 0.98 step 2800, training accuracy 1 step 2900, training accuracy 1 step 3000, training accuracy 0.98 step 3100, training accuracy 1 step 3200, training accuracy 0.98 step 3300, training accuracy 0.98 step 3400, training accuracy 0.94 step 3500, training accuracy 1 step 3600, training accuracy 0.98 step 3700, training accuracy 0.96 step 3800, training accuracy 1 step 3900, training accuracy 1 step 4000, training accuracy 0.98 step 4100, training accuracy 0.98 step 4200, training accuracy 1 step 4300, training accuracy 1 step 4400, training accuracy 0.98 step 4500, training accuracy 0.98 step 4600, training accuracy 1 step 4700, training accuracy 1 step 4800, training accuracy 1 step 4900, training accuracy 1 step 5000, training accuracy 0.98 step 5100, training accuracy 0.98 step 5200, training accuracy 0.98 step 5300, training accuracy 1 step 5400, training accuracy 1 step 5500, training accuracy 0.96 step 5600, training accuracy 1 step 5700, training accuracy 0.98 step 5800, training accuracy 0.96 step 5900, training accuracy 1 step 6000, training accuracy 1 step 6100, training accuracy 0.98 step 6200, training accuracy 0.98 step 6300, training accuracy 1 step 6400, training accuracy 0.96 step 6500, training accuracy 1 step 6600, training accuracy 1 step 6700, training accuracy 0.98 step 6800, training accuracy 1 step 6900, training accuracy 1 step 7000, training accuracy 1 step 7100, training accuracy 1 step 7200, training accuracy 0.94 step 7300, training accuracy 1 step 7400, training accuracy 0.96 step 7500, training accuracy 0.98 step 7600, training accuracy 1 step 7700, training accuracy 1 step 7800, training accuracy 0.98 step 7900, training accuracy 0.98 step 8000, training accuracy 1 step 8100, training accuracy 0.98 step 8200, training accuracy 1 step 8300, training accuracy 1 step 8400, training accuracy 1 step 8500, training accuracy 0.98 step 8600, training accuracy 1 step 8700, training accuracy 1 step 8800, training accuracy 1 step 8900, training accuracy 1 step 9000, training accuracy 1 step 9100, training accuracy 1 step 9200, training accuracy 1 step 9300, training accuracy 0.98 step 9400, training accuracy 1 step 9500, training accuracy 1 step 9600, training accuracy 1 step 9700, training accuracy 1 step 9800, training accuracy 1 step 9900, training accuracy 0.96 step 10000, training accuracy 1 step 10100, training accuracy 1 step 10200, training accuracy 1 step 10300, training accuracy 1 step 10400, training accuracy 0.98 step 10500, training accuracy 0.96 step 10600, training accuracy 0.98 step 10700, training accuracy 0.98 step 10800, training accuracy 0.98 step 10900, training accuracy 1 step 11000, training accuracy 0.98 step 11100, training accuracy 1 step 11200, training accuracy 1 step 11300, training accuracy 1 step 11400, training accuracy 1 step 11500, training accuracy 1 step 11600, training accuracy 0.98 step 11700, training accuracy 1 step 11800, training accuracy 0.98 step 11900, training accuracy 1 step 12000, training accuracy 1 step 12100, training accuracy 1 step 12200, training accuracy 1 step 12300, training accuracy 0.98 step 12400, training accuracy 1 step 12500, training accuracy 1 step 12600, training accuracy 0.98 step 12700, training accuracy 1 step 12800, training accuracy 1 step 12900, training accuracy 1 step 13000, training accuracy 0.98 step 13100, training accuracy 1 step 13200, training accuracy 1 step 13300, training accuracy 1 step 13400, training accuracy 1 step 13500, training accuracy 1 step 13600, training accuracy 1 step 13700, training accuracy 1 step 13800, training accuracy 1 step 13900, training accuracy 1 step 14000, training accuracy 1 step 14100, training accuracy 1 step 14200, training accuracy 1 step 14300, training accuracy 1 step 14400, training accuracy 1 step 14500, training accuracy 1 step 14600, training accuracy 1 step 14700, training accuracy 1 step 14800, training accuracy 1 step 14900, training accuracy 0.98 step 15000, training accuracy 1 step 15100, training accuracy 0.98 step 15200, training accuracy 1 step 15300, training accuracy 1 step 15400, training accuracy 0.98 step 15500, training accuracy 1 step 15600, training accuracy 1 step 15700, training accuracy 1 step 15800, training accuracy 1 step 15900, training accuracy 1 step 16000, training accuracy 1 step 16100, training accuracy 1 step 16200, training accuracy 1 step 16300, training accuracy 0.98 step 16400, training accuracy 1 step 16500, training accuracy 1 step 16600, training accuracy 1 step 16700, training accuracy 1 step 16800, training accuracy 1 step 16900, training accuracy 1 step 17000, training accuracy 1 step 17100, training accuracy 1 step 17200, training accuracy 1 step 17300, training accuracy 1 step 17400, training accuracy 1 step 17500, training accuracy 1 step 17600, training accuracy 0.96 step 17700, training accuracy 1 step 17800, training accuracy 1 step 17900, training accuracy 1 step 18000, training accuracy 0.98 step 18100, training accuracy 1 step 18200, training accuracy 1 step 18300, training accuracy 1 step 18400, training accuracy 1 step 18500, training accuracy 1step 18600, training accuracy 1 step 18700, training accuracy 1 step 18800, training accuracy 1 step 18900, training accuracy 1 step 19000, training accuracy 1 step 19100, training accuracy 1 step 19200, training accuracy 1 step 19300, training accuracy 1 step 19400, training accuracy 1 step 19500, training accuracy 1 step 19600, training accuracy 1 step 19700, training accuracy 1 step 19800, training accuracy 1 step 19900, training accuracy 1 test accuracy 0.9933
多通道mnist
# -*- coding: utf-8 -*-
import tensorflow as tf
import numpy as np
# 導入 MINST 數據集
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/data/", one_hot=True)
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')
def avg_pool_7x7(x):
return tf.nn.avg_pool(x, ksize=[1, 7, 7, 1],
strides=[1, 7, 7, 1], padding='SAME')
# tf Graph Input
x = tf.placeholder(tf.float32, [None, 784]) # mnist data維度 28*28=784
y = tf.placeholder(tf.float32, [None, 10]) # 0-9 數字=> 10 classes
W_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
x_image = tf.reshape(x, [-1,28,28,1])
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)
######################################################多卷積核
W_conv2_5x5 = weight_variable([5, 5, 32, 32])
b_conv2_5x5 = bias_variable([32])
W_conv2_7x7 = weight_variable([7, 7, 32, 32])
b_conv2_7x7 = bias_variable([32])
h_conv2_5x5 = tf.nn.relu(conv2d(h_pool1, W_conv2_5x5) + b_conv2_5x5)
h_conv2_7x7 = tf.nn.relu(conv2d(h_pool1, W_conv2_7x7) + b_conv2_7x7)
h_conv2 = tf.concat([h_conv2_5x5,h_conv2_7x7],3)
h_pool2 = max_pool_2x2(h_conv2)
#########################################################new 池化
W_conv3 = weight_variable([5, 5, 64, 10])
b_conv3 = bias_variable([10])
h_conv3 = tf.nn.relu(conv2d(h_pool2, W_conv3) + b_conv3)
nt_hpool3=avg_pool_7x7(h_conv3)#10
nt_hpool3_flat = tf.reshape(nt_hpool3, [-1, 10])
y_conv=tf.nn.softmax(nt_hpool3_flat)
keep_prob = tf.placeholder("float")
cross_entropy = -tf.reduce_sum(y*tf.log(y_conv))
#不同的優化方法測測效果
#train_step = tf.train.GradientDescentOptimizer(1e-3).minimize(cross_entropy)
#train_step = tf.train.AdagradOptimizer(1e-5).minimize(cross_entropy)
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(y_conv,1), tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
# 啓動session
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
for i in range(20000):#20000
batch = mnist.train.next_batch(50)
if i%100 == 0:
train_accuracy = accuracy.eval(feed_dict={
x:batch[0], y: batch[1], keep_prob: 1.0})
print( "step %d, training accuracy %g"%(i, train_accuracy))
train_step.run(feed_dict={x: batch[0], y: batch[1], keep_prob: 0.5})
print ("test accuracy %g"%accuracy.eval(feed_dict={
x: mnist.test.images, y: mnist.test.labels, keep_prob: 1.0}))
step 0, training accuracy 0.14 step 100, training accuracy 0.08 step 200, training accuracy 0.44 step 300, training accuracy 0.52 step 400, training accuracy 0.44 step 500, training accuracy 0.36 step 600, training accuracy 0.68 step 700, training accuracy 0.5 step 800, training accuracy 0.74 step 900, training accuracy 0.7 step 1000, training accuracy 0.76 step 1100, training accuracy 0.8 step 1200, training accuracy 0.84 step 1300, training accuracy 0.8 step 1400, training accuracy 0.84 step 1500, training accuracy 0.82 step 1600, training accuracy 0.86 step 1700, training accuracy 0.88 step 1800, training accuracy 0.88 step 1900, training accuracy 0.9 step 2000, training accuracy 0.98 step 2100, training accuracy 0.9 step 2200, training accuracy 0.92 step 2300, training accuracy 0.94 step 2400, training accuracy 0.9 step 2500, training accuracy 0.98 step 2600, training accuracy 0.92 step 2700, training accuracy 0.92 step 2800, training accuracy 0.88 step 2900, training accuracy 0.92 step 3000, training accuracy 0.9 step 3100, training accuracy 0.92 step 3200, training accuracy 0.94 step 3300, training accuracy 0.92 step 3400, training accuracy 0.92 step 3500, training accuracy 0.96 step 3600, training accuracy 0.92 step 3700, training accuracy 0.88 step 3800, training accuracy 0.96 step 3900, training accuracy 0.94 step 4000, training accuracy 0.98 step 4100, training accuracy 0.86 step 4200, training accuracy 0.94 step 4300, training accuracy 0.88 step 4400, training accuracy 0.92 step 4500, training accuracy 0.94 step 4600, training accuracy 0.94 step 4700, training accuracy 0.88 step 4800, training accuracy 0.94 step 4900, training accuracy 0.94 step 5000, training accuracy 0.88 step 5100, training accuracy 0.92 step 5200, training accuracy 0.98 step 5300, training accuracy 0.88 step 5400, training accuracy 0.94 step 5500, training accuracy 0.92 step 5600, training accuracy 0.94 step 5700, training accuracy 0.96 step 5800, training accuracy 0.96 step 5900, training accuracy 0.98 step 6000, training accuracy 1 step 6100, training accuracy 0.94 step 6200, training accuracy 0.98 step 6300, training accuracy 0.96 step 6400, training accuracy 0.92 step 6500, training accuracy 0.94 step 6600, training accuracy 0.98 step 6700, training accuracy 0.92 step 6800, training accuracy 0.94 step 6900, training accuracy 0.96 step 7000, training accuracy 1 step 7100, training accuracy 0.94 step 7200, training accuracy 0.92 step 7300, training accuracy 0.96 step 7400, training accuracy 0.98 step 7500, training accuracy 0.96 step 7600, training accuracy 0.94 step 7700, training accuracy 0.96 step 7800, training accuracy 0.96 step 7900, training accuracy 0.98 step 8000, training accuracy 0.96 step 8100, training accuracy 0.92 step 8200, training accuracy 0.98 step 8300, training accuracy 0.98 step 8400, training accuracy 0.96 step 8500, training accuracy 0.96 step 8600, training accuracy 0.96 step 8700, training accuracy 0.94 step 8800, training accuracy 0.98 step 8900, training accuracy 0.96 step 9000, training accuracy 0.98 step 9100, training accuracy 0.9 step 9200, training accuracy 0.96 step 9300, training accuracy 0.98 step 9400, training accuracy 0.88 step 9500, training accuracy 0.98 step 9600, training accuracy 0.98 step 9700, training accuracy 1 step 9800, training accuracy 0.92 step 9900, training accuracy 0.96 step 10000, training accuracy 0.98 step 10100, training accuracy 0.96 step 10200, training accuracy 0.98 step 10300, training accuracy 0.9 step 10400, training accuracy 0.92 step 10500, training accuracy 0.96 step 10600, training accuracy 0.96 step 10700, training accuracy 0.94 step 10800, training accuracy 0.98 step 10900, training accuracy 0.98 step 11000, training accuracy 0.98 step 11100, training accuracy 0.98 step 11200, training accuracy 1 step 11300, training accuracy 0.92 step 11400, training accuracy 1 step 11500, training accuracy 1 step 11600, training accuracy 1 step 11700, training accuracy 0.94 step 11800, training accuracy 0.98 step 11900, training accuracy 0.98 step 12000, training accuracy 0.96 step 12100, training accuracy 0.94 step 12200, training accuracy 0.96 step 12300, training accuracy 0.98 step 12400, training accuracy 0.94 step 12500, training accuracy 0.92 step 12600, training accuracy 0.92 step 12700, training accuracy 0.98 step 12800, training accuracy 0.98 step 12900, training accuracy 0.98 step 13000, training accuracy 0.96 step 13100, training accuracy 0.96 step 13200, training accuracy 0.98 step 13300, training accuracy 0.96 step 13400, training accuracy 0.92 step 13500, training accuracy 0.96 step 13600, training accuracy 0.98 step 13700, training accuracy 0.94 step 13800, training accuracy 0.98 step 13900, training accuracy 0.98 step 14000, training accuracy 0.98 step 14100, training accuracy 0.9 step 14200, training accuracy 0.98 step 14300, training accuracy 1 step 14400, training accuracy 0.9 step 14500, training accuracy 0.98 step 14600, training accuracy 0.92 step 14700, training accuracy 0.94 step 14800, training accuracy 0.94 step 14900, training accuracy 0.98 step 15000, training accuracy 1 step 15100, training accuracy 1 step 15200, training accuracy 0.98 step 15300, training accuracy 0.94 step 15400, training accuracy 0.96 step 15500, training accuracy 0.94 step 15600, training accuracy 0.98 step 15700, training accuracy 0.98 step 15800, training accuracy 0.92 step 15900, training accuracy 0.94 step 16000, training accuracy 0.98 step 16100, training accuracy 0.98 step 16200, training accuracy 0.98 step 16300, training accuracy 1 step 16400, training accuracy 0.96 step 16500, training accuracy 0.94 step 16600, training accuracy 0.96 step 16700, training accuracy 0.96 step 16800, training accuracy 0.98 step 16900, training accuracy 0.96 step 17000, training accuracy 1 step 17100, training accuracy 0.98 step 17200, training accuracy 0.98 step 17300, training accuracy 0.94 step 17400, training accuracy 1 step 17500, training accuracy 0.94 step 17600, training accuracy 0.96 step 17700, training accuracy 1 step 17800, training accuracy 0.96 step 17900, training accuracy 0.92 step 18000, training accuracy 0.98 step 18100, training accuracy 0.96 step 18200, training accuracy 0.96 step 18300, training accuracy 0.98 step 18400, training accuracy 0.98 step 18500, training accuracy 0.98 step 18600, training accuracy 0.98 step 18700, training accuracy 0.98 step 18800, training accuracy 0.98 step 18900, training accuracy 0.98 step 19000, training accuracy 1 step 19100, training accuracy 0.98 step 19200, training accuracy 0.98 step 19300, training accuracy 1 step 19400, training accuracy 1 step 19500, training accuracy 0.98 step 19600, training accuracy 0.96 step 19700, training accuracy 0.98 step 19800, training accuracy 0.98 step 19900, training accuracy 1 test accuracy 0.9768