TensorFlow 代码

从Anaconda中转出
[网易TensorFlow 入门实操课程链接](https://www.icourse163.org/learn/youdao-1460578162?tid=1461280442#/learn/content?type=detail&id=1239078006&sm=1 "网易课程链接")
[课程代码原文下载](https://note.youdao.com/coshare/index.html?token=6AD21133F93449538732CAFE13303CF2&gid=117583840#/906295617 "课程代码原文下载")

## 1.线性拟合预测
```python
#!/usr/bin/env python
# coding: utf-8

# In[2]:

get_ipython().run_line_magic('config', 'IPCompleter.greedy=True')

# In[12]:

from tensorflow import keras   #keras是tensorflow中的一个API
import numpy as np
model=keras.Sequential([keras.layers.Dense(units=1,input_shape=[1])])  #layers代表一层神经元，其中units也只有1，input_shape是输入值
model.compile(optimizer='sgd',loss='mean_squared_error')   #根据什么来优化，根据什么来检测损失函数，loss值越来越小证明训练靠谱

xs=np.array([-1.0,0.0,1.0,2.0,3.0,4.0],dtype=float)
ys=np.array([-3.0,-1.0,1.0,3.0,5.0,7.0],dtype=float)

model.fit(xs,ys,epochs=500)    #训练次数500

# In[13]:

model.predict([10.0])  #输出x=10时的预测值

```

## 2.加载数据库
```python
#!/usr/bin/env python
# coding: utf-8

# In[2]:

#加载Fashion MNIST数据集

from tensorflow import keras
fashion_mnist = keras.datasets.fashion_mnist
(train_images,train_labels),(test_images,test_labels) = fashion_mnist.load_data()

# In[6]:

print(train_images.shape) #60000张图片，28*28像素，其中都是灰度值
print(test_images.shape) #测试集10000张，要是没见过的，用于测试准确度

# In[5]:

import matplotlib.pyplot as plt
plt.imshow(train_images[0])    #显示第0张图片

```

## 2.构建网络神经元，通过数据库来训练

```python
#!/usr/bin/env python
# coding: utf-8

# In[1]:

#加载Fashion MNIST数据集
import tensorflow as tf
from tensorflow import keras
fashion_mnist = keras.datasets.fashion_mnist
(train_images,train_labels),(test_images,test_labels) = fashion_mnist.load_data()

# In[2]:

#构造神经元模型
model = keras.Sequential([
    keras.layers.Flatten(input_shape=(28,28)),   #输入层
    keras.layers.Dense(128,activation=tf.nn.relu),    #中间层，128个神经元，relu/softmax-激活函数
    keras.layers.Dense(10,activation=tf.nn.softmax)    #输出层,10个类别
])

# In[3]:

model.summary()

# In[4]:

train_images=train_images/255  #normalization,把像素数值变成0~1之间的数
model.compile(optimizer="adam",loss="sparse_categorical_crossentropy",metrics=['accuracy']) #优化和损失函数

# In[5]:

model.fit(train_images,train_labels,epochs=5)  #训练，所有数据跑五次

# In[6]:

test_images=test_images/255
model.evaluate(test_images,test_labels)  #评估成效 ,loss和accuracy跟训练的时候差不多

# In[7]:

#类别 0_Tshirt/top,1_trouser/pants,2_pullover shirt,3_Dress,4_coat,5__sandal,6_shirt,7_sneaker,8_bag,9_anklle boot

# In[15]:

import numpy as np
np.argmax(model.predict((test_images[0]/255).reshape(1,28,28,1))) #预测第一张图片的类别，输出可能属于几个类别的概率,用argmax直接输出概率最大的标签编号

# In[16]:

import matplotlib.pyplot as plt
print(test_labels[0])
plt.imshow(test_images[0])  #核对答案

```

## 3.自动终止训练（回调函数）
```python
#!/usr/bin/env python
# coding: utf-8

# In[1]:

import tensorflow as tf

# In[8]:

class myCallback(tf.keras.callbacks.Callback):
  def on_epoch_end(self,epoch,logs={}):
    if(logs.get('loss')<0.4):
      print('\nloss is low so cancelling training!')
      self.model.stop_training = True

callbacks = myCallback()
mnist = tf.keras.datasets.fashion_mnist
(training_images,training_labels),(test_images,test_labels)=mnist.load_data()
training_images=training_images/255.0
test_images=test_images/255.0
model=tf.keras.models.Sequential([
  tf.keras.layers.Flatten(input_shape=(28, 28)),
  tf.keras.layers.Dense(512,activation=tf.nn.relu),
  tf.keras.layers.Dense(10,activation=tf.nn.softmax)
])
model.compile(optimizer='adam',loss='sparse_categorical_crossentropy',metrics=['accuracy'])
model.fit(training_images,training_labels,epochs=5,callbacks=[callbacks])

```

## 4.卷积神经网络
```python
#!/usr/bin/env python
# coding: utf-8

# In[3]:

#加载Fashion MNIST数据集
import tensorflow as tf
from tensorflow import keras
fashion_mnist = keras.datasets.fashion_mnist
(train_images,train_labels),(test_images,test_labels) = fashion_mnist.load_data()

# In[5]:

#构造神经元模型+卷积网络
model = keras.Sequential([
    keras.layers.Conv2D(64,(3,3),activation='relu',input_shape=(28,28,1)),   #64个3x3大小的过滤器
    keras.layers.MaxPooling2D(2,2),    #卷积后，用maxpooling增强它的特征，减少数据，2x2四个像素
    keras.layers.Conv2D(64,(3,3),activation='relu'),   #64个3x3大小的过滤器
    keras.layers.MaxPooling2D(2,2),   
    keras.layers.Flatten(input_shape=(28,28)),   #输入层
    keras.layers.Dense(128,activation=tf.nn.relu),    #中间层，128个神经元，relu/softmax-激活函数
    keras.layers.Dense(10,activation=tf.nn.softmax)    #输出层,10个类别
])

# In[8]:

model.compile(optimizer='adam',loss='sparse_categorical_crossentropy',metrics=['accuracy'])
model.fit(train_images,train_labels,epochs=5)  #训练结果更加准确，时间更长

# In[10]:

model.summary()  #查看神经元网络结构

# In[14]:

#把每层分开
import matplotlib.pyplot as plt
layer_outputs = [layer.output for layer in model.layers]
activation_model = tf.keras.models.Model(inputs= model.input,outputs = layer_outputs)
pred = activation_model.predict(test_images[0].reshape(1,28,28,1))
pred   #可以看到七层输出

# In[21]:

plt.imshow(pred[0][0,:,:,4])  #以图片形式查看第一层，所有行所有列第二个过滤器的卷积结果

```

## 项目实战(人和马)
```python
#!/usr/bin/env python
# coding: utf-8

# In[9]:

#下载文件到当前目录，并且进行解压，此处省略代码
get_ipython().system('wget --no-check-certificate     https://storage.googleapis.com/laurencemoroney-blog.appspot.com/validation-horse-or-human.zip     -o /tmp/validation-horse-or-human.zip')

# In[10]:

import tensorflow as tf
import scipy

from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.optimizers import RMSprop

#创建两个数据生成器，指定scaling范围0~1
train_datagen = ImageDataGenerator(rescale=1/255)
validation_datagen = ImageDataGenerator(rescale=1/255)

#指向训练数据文件夹
train_generator = train_datagen.flow_from_directory(
        'C:/Users/82399/validation-horse-or-human/',#训练数据所在文件夹
        target_size=(150,150), #指定输出尺寸,设置的小防止内存不够
        batch_size=32,         #指定每一批取多少
        class_mode='binary')   #指定为二分类

#指向测试数据文件夹
validation_generator = validation_datagen.flow_from_directory(
        'C:/Users/82399/horse-or-human/',#训练数据所在文件夹
        target_size=(150,150), #指定输出尺寸
        batch_size=32,         #指定每一批取多少
        class_mode='binary')   #指定为二分类

# In[18]:

pip install keras-tuner --upgrade

# In[11]:

#构建神经网络
#引入kerastuner中的hperband方法,对于一些网络里的参数可以实现循环计算以探究最合适的数字
from keras_tuner.tuners import Hyperband
from keras_tuner.engine.hyperparameters import HyperParameters

hp = HyperParameters() #定义一种参数类型

def build_model(hp):
    model = tf.keras.models.Sequential()
    model.add(tf.keras.layers.Conv2D(hp.Choice('num_filters_layer0',values=[16,64],default=16),(3,3),activation='relu',input_shape=(150,150,3)))
    model.add(tf.keras.layers.MaxPooling2D(2,2))
    for i in range(hp.Int(f'num_conv_layers',1,3)): #1-3之间取整
        model.add(tf.keras.layers.Conv2D(hp.Choice('num_filters_layer0',values=[16,64],default=16),(3,3),activation='relu'))
        model.add(tf.keras.layers.MaxPooling2D(2,2))
    model.add(tf.keras.layers.Flatten())#将结果压平，反馈给深度神经网络DNN
    model.add(tf.keras.layers.Dense(hp.Int('hidden_units',128,512,step=32),activation='relu'))#512个神经元
    model.add(tf.keras.layers.Dense(1,activation='sigmoid'))#只有一个输出神经元，因为结果是二分类的，0-马，1-人
    model.compile(loss='binary_crossentropy',optimizer=RMSprop(learning_rate=0.001),metrics=['acc'])
    return model
tuner =Hyperband(
    build_model,
    objective='val_acc',
    max_epochs=15,
    directory='horse_human_params',
    hyperparameters=hp,
    project_name='my_horse_human-project'
)

# In[12]:

#开始计算，根据参数的多少，计算时间较长
tuner.search(train_generator,epochs=10,validation_data=validation_generator)
#结果Trial 30 Complete [00h 04m 34s]
#val_acc: 0.7994157671928406
#Best val_acc So Far: 0.8938656449317932
#Total elapsed time: 00h 41m 48s
#INFO:tensorflow:Oracle triggered exit

# In[13]:

#查看能使性能最好的参数，并以该组参数建立新的神经网路
best_hps=tuner.get_best_hyperparameters(1)[0]
print(best_hps.values)
model=tuner.hypermodel.build(best_hps)
model.summary()

```

![](/media/202208/2022-08-27_223454_287858.png)

## 项目实战(猫和狗)
```python
#!/usr/bin/env python
# coding: utf-8

# In[2]:

import tensorflow as tf
import os
import random
from shutil import copyfile

# In[12]:

#访问 https://www.microsoft.com/en-us/download/confirmation.aspx?id=54765 获得数据包，解压缩到当前文件夹
print(len(os.listdir('C:/Users/82399/PetImages/Cat')))
print(len(os.listdir('C:/Users/82399/PetImages/Dog')))

# Expected Output:
# 12501
# 12501

# In[21]:

#创建用于训练和测试的文件夹
try:
    os.mkdir('C:/Users/82399/cat-v-dog/')
    os.mkdir('C:/Users/82399/cat-v-dog/training')
    os.mkdir('C:/Users/82399/cat-v-dog/training/cat')
    os.mkdir('C:/Users/82399/cat-v-dog/training/dog')
    os.mkdir('C:/Users/82399/cat-v-dog/testing')
    os.mkdir('C:/Users/82399/cat-v-dog/testing/cat')
    os.mkdir('C:/Users/82399/cat-v-dog/testing/dog')
except OSError:
    pass

# In[ ]:

#定义一个函数，来把数据图片按比例且随机地分给训练集和测试集
#路径加文件名等于下一级路径，所以参数中的问价夹路径应该要加/
def spilt_data(source,training,testing,spilt_size=0.9):
    files=[]
    for filename in os.listdir(source):
        file = source + filename
        if os.path.getsize(file) >0:
            files.append(filename)
        else:
            print(filename + "is zero length,so ignoring.")
    training_length = int(len(files)*spilt_size) #训练集长度
    testing_length = int(len(files)-training_length) #测试集长度
    shuffled_set = random.sample(files,len(files)) #打乱顺序
    training_set = shuffled_set[0:training_length]  #规定训练集
    testing_set = shuffled_set[-testing_length:] #规定测试集
    
    #训练集和测试集数据复制到新建文件夹
    for filename in training_set:  
        this_file = source + filename
        destination = training + filename
        copyfile(this_file,destination)
        
    for filename in testing_set:
        this_file = source + filename
        destination = testing + filename
        copyfile(this_file,destination)
        
cat_source_dir = 'C:/Users/82399/PetImages/Cat/'
dog_source_dir = 'C:/Users/82399/PetImages/Dog/'
cat_training_dir = 'C:/Users/82399/cat-v-dog/training/cat/'
dog_training_dir = 'C:/Users/82399/cat-v-dog/training/dog/'
cat_testing_dir = 'C:/Users/82399/cat-v-dog/testing/cat/'
dog_testing_dir = 'C:/Users/82399/cat-v-dog/testing/dog/'

spilt_data(cat_source_dir,cat_training_dir,cat_testing_dir)
spilt_data(dog_source_dir,dog_training_dir,dog_testing_dir)

# Expected output
# 666.jpg is zero length, so ignoring
# 11702.jpg is zero length, so ignoring

# In[4]:

#查看数据数量
print(len(os.listdir('C:/Users/82399/cat-v-dog/training/cat/')))
print(len(os.listdir('C:/Users/82399/cat-v-dog/training/dog/')))
print(len(os.listdir('C:/Users/82399/cat-v-dog/testing/cat/')))
print(len(os.listdir('C:/Users/82399/cat-v-dog/testing/dog/')))

# Expected output:
# 11250
# 11250
# 1250
# 1250

# In[3]:

#数据预处理，进行归一化，减少运算量
#RGB值都归一化，二分类，每一批取100个，输出150*150像素的照片

from tensorflow.keras.preprocessing.image import ImageDataGenerator
training_dir = 'C:/Users/82399/cat-v-dog/training'
train_datagen = ImageDataGenerator(rescale=1.0/255.0)
train_generator = train_datagen.flow_from_directory(training_dir,batch_size=100,class_mode='binary',target_size=(150,150))

testing_dir = 'C:/Users/82399/cat-v-dog/testing'
test_datagen = ImageDataGenerator(rescale=1.0/255.0)
test_generator = test_datagen.flow_from_directory(testing_dir,batch_size=100,class_mode='binary',target_size=(150,150))

# Expected Output:
# Found 22498 images belonging to 2 classes.
# Found 2500 images belonging to 2 classes.

# In[4]:

#构建CNN模型，Sequential表示每层按顺序链接。共有三对卷积+最大池化，最后一次池化层被展平（Flatten),
from tensorflow.keras.optimizers import RMSprop
model = tf.keras.models.Sequential([
    tf.keras.layers.Conv2D(16,(3,3),activation='relu',input_shape= (150,150,3)), #3表示彩色图片3个通道
    tf.keras.layers.MaxPooling2D(2,2),
    tf.keras.layers.Conv2D(32,(3,3),activation='relu'),
    tf.keras.layers.MaxPooling2D(2,2),
    tf.keras.layers.Conv2D(64,(3,3),activation='relu'),
    tf.keras.layers.MaxPooling2D(2,2),
    tf.keras.layers.Flatten(),
    tf.keras.layers.Dense(512,activation='relu'),
    tf.keras.layers.Dense(1,activation='sigmoid') #二分类问题，输出只需1个神经元
])

#优化方法，步长，损失函数为二元交叉熵
model.compile(optimizer=RMSprop(learning_rate=0.001),loss='binary_crossentropy',metrics=['acc'])

# In[5]:

#进行拟合,5轮（效果好要改多一点），每轮255批（每批100个），verbose显示日志
#history就是一个拟合过程的日志
history = model.fit_generator(train_generator,epochs=5,verbose=1,validation_data=test_generator)

# In[7]:

#%matplotlib inline 可以在Ipython编译器里直接使用，功能是可以内嵌绘图，并且可以省略掉plt.show()这一步
get_ipython().run_line_magic('matplotlib', 'inline')

import matplotlib.image as mping
import matplotlib.pyplot as plt

acc = history.history['acc']
val_acc=history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']
epochs = range(len(acc))

plt.plot(epochs,acc,'r','Training Accuracy') #x,y,颜色，名称
plt.plot(epochs,val_acc,'b','Validation Accuracy')
plt.title('Training and validation accuracy')
plt.figure()

plt.plot(epochs,loss,'r','Training Loss')
plt.plot(epochs,val_loss,'b','validation_Loss')
plt.title('Training and validation Loss')
plt.figure()

```

TensorFlow 代码

【温馨提示：本站文档可配置可见范围，如登录后可见、对特定群组可见等，看不到或进不去就是没权限】

暂时不开放注册