keras—神经网络CNN—MNIST手写数字识别

 1 from keras.datasets import mnist
 2 from keras.utils import np_utils
 3 from plot_image_1 import plot_image_1
 4 from plot_prediction_1 import plot_image_labels_prediction_1
 5 from show_train_history import show_train_history
 6 import numpy as np
 7 import pandas as pd
 8 from keras.models import Sequential
 9 from keras.layers import Dense,Dropout,Flatten,Conv2D,MaxPooling2D
10 np.random.seed(10)
11 (x_Train,y_Train),(x_Test,y_Test)=mnist.load_data()
12 print('train data=',len(x_Train))
13 print('test data=',len(x_Test))
14 print('x_train_image:',x_Train.shape)
15 print('y_train_label:',y_Train.shape)
16 x_Train4D=x_Train.reshape(x_Train.shape[0],28,28,1).astype('float32')
17 x_Test4D=x_Test.reshape(x_Test.shape[0],28,28,1).astype('float32')
18 x_Train4D_normalize=x_Train4D/255
19 x_Test4D_normalize=x_Test4D/255
20 y_TrainOneHot=np_utils.to_categorical(y_Train)
21 y_TestOneHot=np_utils.to_categorical(y_Test)
22 model=Sequential()
23 model.add(Conv2D(filters=16,
24                  kernel_size=(5,5),
25                  padding='same',
26                  input_shape=(28,28,1),
27                  activation='relu'))
28 model.add(MaxPooling2D(pool_size=(2,2)))
29 model.add(Conv2D(filters=36,
30                  kernel_size=(5,5),
31                  padding='same',
32                  activation='relu'))
33 model.add(MaxPooling2D(pool_size=(2,2)))
34 model.add(Dropout(0.25))
35 model.add(Flatten())
36 model.add(Dense(128,activation='relu'))
37 model.add(Dropout(0.5))
38 model.add(Dense(10,activation='softmax'))
39 print(model.summary())
40 model.compile(loss='categorical_crossentropy',
41               optimizer='adam',metrics=['accuracy'])
42 train_history=model.fit(x=x_Train4D_normalize,
43                         y=y_TrainOneHot,validation_split=0.2,
44                         epochs=5,batch_size=300,verbose=2)
45 show_train_history(train_history,'acc','val_acc')
46 show_train_history(train_history,'loss','val_loss')
47 scores=model.evaluate(x_Test4D_normalize,y_TestOneHot)
48 print()
49 print('accuracy',scores[1])
50 prediction=model.predict_classes(x_Test4D_normalize)
51 print("prediction[:10]",prediction[:10])
52 plot_image_labels_prediction_1(x_Test,y_Test,prediction,idx=0)
53 pd.crosstab(y_Test,prediction,rownames=['label'],colnames=['predict'])

1 import matplotlib.pyplot as plt
2 def plot_image_1(image):
3     fig=plt.gcf()
4     fig.set_size_inches(2,2)
5     plt.imshow(image,cmap='binary')
6     plt.show()

 1 import matplotlib.pyplot as plt
 2 def plot_image_labels_prediction_1(image,labels,prediction,idx,num=10):
 3     fig=plt.gcf()
 4     fig.set_size_inches(12,14)
 5     if num>25:num=25
 6     for i in range(0,num):
 7         ax=plt.subplot(5,5,i+1)
 8         ax.imshow(image[idx],cmap='binary')
 9         title="label="+str(labels[idx])
10         if len(prediction)>0:
11             title+=",predict="+str(prediction[idx])
12         ax.set_title(title,fontsize=10)
13         ax.set_xticks([]);ax.set_yticks([])
14         idx+=1
15     plt.show()

1 import matplotlib.pyplot as plt
2 def show_train_history(train_history,train,validation):
3     plt.plot(train_history.history[train])
4     plt.plot(train_history.history[validation])
5     plt.title('Train History')
6     plt.ylabel(train)
7     plt.xlabel('Epoch')
8     plt.legend(['train','validation'],loc='upper left')    #显示左上角标签
9     plt.show()