inputs = Input(input_size)
conv1 = Conv2D(64, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(inputs)
conv1 = Conv2D(64, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv1)
pool1 = MaxPooling2D(pool_size=(2, 2))(conv1)
# 128->64
conv2 = Conv2D(128, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(pool1)
conv2 = Conv2D(128, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv2)
pool2 = MaxPooling2D(pool_size=(2, 2))(conv2)
# 64->32
conv3 = Conv2D(256, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(pool2)
conv3 = Conv2D(256, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv3)
pool3 = MaxPooling2D(pool_size=(2, 2))(conv3)
# 32->16
conv4 = Conv2D(512, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(pool3)
conv4 = Conv2D(512, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv4)
drop4 = Dropout(0.5)(conv4)
pool4 = MaxPooling2D(pool_size=(2, 2))(drop4)
# 16->8

conv5 = Conv2D(1024, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(pool4)
conv5 = Conv2D(1024, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv5)
drop5 = Dropout(0.5)(conv5)

up6 = Conv2D(512, 2, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(drop5))
merge6 = concatenate([drop4,up6], axis = 3)
conv6 = Conv2D(512, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(merge6)
conv6 = Conv2D(512, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv6)

up7 = Conv2D(256, 2, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(conv6))
merge7 = concatenate([conv3,up7], axis = 3)
conv7 = Conv2D(256, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(merge7)
conv7 = Conv2D(256, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv7)

up8 = Conv2D(128, 2, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(conv7))
merge8 = concatenate([conv2,up8], axis = 3)
conv8 = Conv2D(128, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(merge8)
conv8 = Conv2D(128, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv8)

up9 = Conv2D(64, 2, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(UpSampling2D(size = (2,2))(conv8))
merge9 = concatenate([conv1,up9], axis = 3)
conv9 = Conv2D(64, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(merge9)
conv9 = Conv2D(64, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv9)
conv9 = Conv2D(2, 3, activation = 'relu', padding = 'same', kernel_initializer = 'he_normal')(conv9)
conv10 = Conv2D(1, 1, activation = 'sigmoid')(conv9)

model = Model(inputs = inputs, outputs = conv10)

def get_iou_vector(A, B):
    # Numpy version
    batch_size = A.shape[0]
    metric = 0.0
    for batch in range(batch_size):
        t, p = A[batch], B[batch]
        true = np.sum(t)
        pred = np.sum(p)

        # deal with empty mask first
        if true == 0:
            metric += (pred == 0)
            continue

        # non empty mask case.  Union is never empty
        # hence it is safe to divide by its number of pixels
        intersection = np.sum(t * p)
        union = true + pred - intersection
        iou = intersection / union

        # iou metrric is a stepwise approximation of the real iou over 0.5
        iou = np.floor(max(0, (iou - 0.45) * 20)) / 10

        metric += iou

    # teake the average over all images in batch
    metric /= batch_size
    return metric

def my_iou_metric(label, pred):
    # Tensorflow version
    return tf.py_func(get_iou_vector, [label, pred > 0.5], tf.float64)

git clone -b hexo_local_file git@github.com:LiHuaBai/LiHuaBai.github.io.git
cnpm install
hexo g
hexo s

import urllib.request
import urllib.parse
import re

# url地址，这里选择了江干区二手房的第三页，不过其后天数据有随机性，并不固定
url='http://hz.58.com/jianggan/ershoufang/pn3/?'
data={}
head={}
# 请求头
head['User-Agent']='Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/60.0.3112.78 Safari/537.36'
# 请求数据
data['utm_source']='sem-sales-baidu-pc'
data['utm_campaign']='sell'
data['utm_medium']='cpc'
data['spm']='62854932425.16537920598'
data['PGTID']='0d300000-0000-0a0c-8067-a60b6ee48510'
data['ClickID']=1
data=urllib.parse.urlencode(data).encode('utf-8')
req=urllib.request.Request(url,data,head,method='GET')
# 将请求得到的响应（response）下载下来
req = urllib.request.urlopen(req)
my_data = req.read()

import urllib.request
import urllib.parse
import re
#保存HTML源码
def saveFile(data):
    path = "D:\\python\\Spider\\58\\html.out"
    f = open(path,'wb')
    f.write(data)
    f.close()
#所需要的数据，考虑到正则匹配的水平太差，就挑了好拿的部分
content = {}
content['tingshi'] = []
content['daxiao'] = []
content['chaoxiang'] = []
content['louceng'] = []
content['jiedao'] = []
content['price'] = []
name = ['tingshi','daxiao','chaoxiang','louceng','jiedao']
#正则匹配式
res_tr1 = r"<ul class='house-list-wrap'>(.*?)</ul>"
res_tr2 = r"<span>(.*?)</span>"
res_price = r'<b>(.*?)</b>'
res_jiedao = r'<a>(.*?)</a>'
def getdata(datalist):
    num = -1
    for x in datalist:
        num = num + 1
        num = num%6;
        if num == 5:
            continue
        if num < 4:
            print(x)
            content[name[num]].append(x)
        if num == 4:
            content[name[num]].append(re.findall(res_jiedao,x,re.S|re.M))
url='http://hz.58.com/jianggan/ershoufang/pn3/?'
data={}
head={}
head['User-Agent']='Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/60.0.3112.78 Safari/537.36'
data['utm_source']='sem-sales-baidu-pc'
data['utm_campaign']='sell'
data['utm_medium']='cpc'
data['spm']='62854932425.16537920598'
data['PGTID']='0d300000-0000-0a0c-8067-a60b6ee48510'
data['ClickID']=1
data=urllib.parse.urlencode(data).encode('utf-8')
req=urllib.request.Request(url,data,head,method='GET')
req = urllib.request.urlopen(req)
my_data = req.read()
saveFile(my_data)
#匹配出ul标签内数据
my_strdata = bytes.decode(my_data)
m_list1 =  re.findall(res_tr1,my_strdata,re.S|re.M)
m_tr1 = "".join(m_list1)
content['price'] =  re.findall(res_price,m_tr1,re.S|re.M)
m_list2 =  re.findall(res_tr2,m_tr1,re.S|re.M)
getdata(m_list2)
print(content)

# 可在tensorflow库中导入
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets('MNIST_data', one_hot=True)

# 首先定义设置网络层的函数，输入输出间均为线性关系
def set_layer(inputs,in_size,out_size,layer_name,activate_function = None):
    # 权重和偏差量定义
    W = tf.Variable(tf.random_uniform([in_size, out_size], -1.0, 1.0), name="W" + layer_name)
    bias =  tf.Variable(tf.constant(0.1, shape=[out_size]), name="bias_"+ layer_name)
    # 线性模型
    Wx_Plus_b = tf.matmul(inputs, W) + bias
    # 防止过拟合,keep_prob为每个元素的被保留概率
    Wx_Plus_b = tf.nn.dropout(Wx_Plus_b, keep_prob)
    # 激活函数的使用
    if activate_function is None:
        outputs = Wx_Plus_b
    else:
        outputs = activate_function(Wx_Plus_b)
    return outputs
# 原始数据占位符
xs = tf.placeholder(tf.float32, [None, n_input], name="input")
ys = tf.placeholder(tf.float32, [None, n_classes], name="output")
# 保留概率
keep_prob = tf.placeholder(tf.float32)
# 神经网络的构建
h1 = set_layer(xs, n_input, hidden_units1, 'hidden_layer_1', activate_function=tf.nn.tanh)
# softmax和tf.nn.tanh为激活函数，都可用,常见的还有relu函数
h2 = set_layer(h1, hidden_units1, hidden_units2, 'hidden_layer_2', activate_function=tf.nn.tanh)
prediction = set_layer(h2, hidden_units2, n_classes, 'prediction_layer', activate_function=None)
# 预测输出在用None无激活函数效果较好，也可用tf.nn.softmax，tf.nn.tanh等激活函数

# cross_entropy = tf.reduce_mean(-tf.reduce_sum(ys * tf.log(prediction)))
# 有看到资料说和下面一句意义相同，但实际上无法使用，
# tf.nn.softmax_cross_entropy_with_logits将预测值转化为总和为一的概率值，并与实际值的计算交叉熵代价函数
# tf.reduce_mean取均值
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=ys, logits=prediction))
# 训练目标为使cross_entropy最小，使用梯度下降法进行训练
train_op = tf.train.GradientDescentOptimizer(learning_rate).minimize(cross_entropy)
tf.summary.scalar('loss', cross_entropy)
# 准确率
correct_prediction = tf.equal(tf.argmax(prediction, 1), tf.argmax(ys, 1))
# 得到true false数据
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
# 求均值

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
import time
mnist = input_data.read_data_sets('MNIST_data', one_hot=True)
def set_layer(inputs,in_size,out_size,layer_name,activate_function = None):
    W = tf.Variable(tf.random_uniform([in_size, out_size], -1.0, 1.0), name="W" + layer_name)
    bias =  tf.Variable(tf.constant(0.1, shape=[out_size]), name="bias_"+ layer_name)
    Wx_Plus_b = tf.matmul(inputs, W) + bias
    Wx_Plus_b = tf.nn.dropout(Wx_Plus_b, keep_prob)#防止过拟合,keep_prob为每个元素的被保留概率
    if activate_function is None:
        outputs = Wx_Plus_b
    else:
        outputs = activate_function(Wx_Plus_b)
    return outputs
# 参数设定
hidden_layers = 1
hidden_units1 = 200
hidden_units2 = 50
n_input = 784
n_classes = 10
learning_rate = 0.8
# 神经网络的构建
xs = tf.placeholder(tf.float32, [None, n_input], name="input")
ys = tf.placeholder(tf.float32, [None, n_classes], name="output")
keep_prob = tf.placeholder(tf.float32)
h1 = set_layer(xs, n_input, hidden_units1, 'hidden_layer_1', activate_function=tf.nn.tanh)
h2 = set_layer(h1, hidden_units1, hidden_units2, 'hidden_layer_2', activate_function=tf.nn.tanh)
prediction = set_layer(h2, hidden_units2, n_classes, 'prediction_layer', activate_function=None)
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=ys, logits=prediction))
train_op = tf.train.GradientDescentOptimizer(learning_rate).minimize(cross_entropy)
tf.summary.scalar('loss', cross_entropy)
# 训练结果准确性
correct_prediction = tf.equal(tf.argmax(prediction, 1), tf.argmax(ys, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
# 训练
init = tf.initialize_all_variables()
n_epochs = 40
batch_size = 100
with tf.Session() as sess:
    st = time.time()
    write = tf.summary.FileWriter('logs/', sess.graph)
    sess.run(init)
    for epoch in range(n_epochs):
        n_batch = int(mnist.train.num_examples / batch_size)
        for i in range(n_batch):
            batch_xs, batch_ys = mnist.train.next_batch(batch_size)
            sess.run(train_op, feed_dict={xs: batch_xs, ys: batch_ys, keep_prob:0.75})
        print ('epoch', epoch, 'accuracy:', sess.run(accuracy, feed_dict={keep_prob:1.0, xs: mnist.test.images, ys: mnist.test.labels}))
    end = time.time()
    print ('*' * 30)
    print ('training finish. cost time:', int(end-st) , 'seconds; accuracy:', sess.run(accuracy, feed_dict={keep_prob:1.0, xs: mnist.test.images, ys: mnist.test.labels}))

# 普通的python数组
a = [[1., 2., 3.], [4., 5., 6.]]

# tensorflow常量的赋值只能为定值，dtype定义了其数据为浮点型,name为其命名
b1 = tf.constant(3,dtype=tf.float32,name = 'b1')
b2 = tf.constant(a,dtype=tf.float32)
# 也可用b1 = tf.constant(3,"float")定义
# 测试后发现tf.zeros([5]),tf.ones([5,5]),tf.random_uniform([5, 5], -2.0, 2.0)创建的也为常量数据
# 其实质和constant相同，zeros等为常量的名称
# 我理解为constant常量可用于给变量或占位符进行赋值

c1 = tf.Variable(a, dtype = tf.float32,name = 'c1')
c2 = tf.Variable(b1, dtype = tf.float32)
c3 = tf.Variable(tf.ones([5,5]))
c4 = tf.Variable(tf.random_uniform([5, 5], -2.0, 2.0))

# tensorflow变量可直接使用python普通数组赋值，也可用tensorflow常量
# 若所赋的初始化值已定义数据类型，则可以不再定义
# 使用变量需先初始化，在计算中其值是不断变换的，一般形状大小设置固定，用作被训练的权重等

# 占位符是事先设定好，用来接受外部输入的一个值，一般用作训练样本和标签的数据
# shape定义的形状大小，None代表任意大小，也可先不定义，与输入数据相同
x1 = tf.placeholder(tf.float32)
x2 = tf.placeholder(tf.float32,[None,3])
x3 = tf.placeholder(dtype = tf.float32,shape = (2,2))

# 在进行运算时，将数据从外部输入
print(sess.run([x1,x2,x3], feed_dict = {x1:[[1,2],[1,2],[3,4]],x2:[[1,2,3],[2,6,4]],x3:[[1,2],[3,4]]}))

import tensorflow as tf
a = [[1., 2., 3.], [4., 5., 6.]]
b1 = tf.constant(3,dtype=tf.float32,name = 'b1')
b2 = tf.constant(a,dtype=tf.float32)
print(b1)
print(b2)
c1 = tf.Variable(a, dtype = tf.float32,name = 'c1')
c2 = tf.Variable(b1, dtype = tf.float32)
c3 = tf.Variable(tf.ones([5,5]))
c4 = tf.Variable(tf.random_uniform([5, 5], -2.0, 2.0))
x1 = tf.placeholder(tf.float32)
x2 = tf.placeholder(tf.float32,[None,3])
x3 = tf.placeholder(dtype = tf.float32,shape = (2,2))
sess = tf.Session()
# 创建图
init = tf.global_variables_initializer()
sess.run(init)
# 将所有变量初始化
print(sess.run([b1,b2]))
print(sess.run([c1,c2,c3,c4]))
print(sess.run([x1,x2,x3], feed_dict = {x1:[[1,2],[1,2],[3,4]],x2:[[1,2,3],[2,6,4]],x3:[[1,2],[3,4]]}))