keras—神经网络CNN—CIFAR_10图像识别


 1 from keras.datasets import cifar10
 2 from keras.utils import np_utils
 3 import matplotlib.pyplot as plt
 4 from keras.models import load_model
 5 import numpy as np
 6 np.random.seed(10)
 7 (x_img_train,y_label_train),(x_img_test,y_label_test)=cifar10.load_data()
 8 print(\'train data=\',len(x_img_train))
 9 print(\'test data=\',len(x_img_test))
10 print(\'x_train_image:\',x_img_train.shape)
11 print(\'x_test_image:\',x_img_test.shape)
12 x_img_train_4D=x_img_train.reshape(x_img_train.shape[0],32,32,3).astype(\'float32\')
13 x_img_test_4D=x_img_test.reshape(x_img_test.shape[0],32,32,3).astype(\'float32\')
14 x_img_train_normalize=x_img_train_4D/255.0
15 x_img_test_normalize=x_img_test_4D/255.0
16 print(x_img_train_normalize[0][0][0])
17 y_label_train_OneHot=np_utils.to_categorical(y_label_train)
18 y_label_test_OneHot=np_utils.to_categorical(y_label_test)
19 print(y_label_train_OneHot[:5])
20 from keras.models import Sequential
21 from keras.layers import Dense,Dropout,Flatten,Conv2D,MaxPooling2D,ZeroPadding2D,Activation
22 model=Sequential()
23 model.add(Conv2D(filters=32,
24                  kernel_size=(3,3),
25                  padding=\'same\',
26                  input_shape=(32,32,3),
27                  activation=\'relu\'))
28 model.add(Dropout(0.25))
29 model.add(MaxPooling2D(pool_size=(2,2)))
30 model.add(Conv2D(filters=64,
31                  kernel_size=(3,3),
32                  padding=\'same\',
33                  activation=\'relu\'))
34 model.add(Dropout(0.25))
35 model.add(MaxPooling2D(pool_size=(2,2)))
36 model.add(Flatten())
37 model.add(Dropout(0.25))
38 model.add(Dense(1024,activation=\'relu\'))
39 model.add(Dropout(0.25))
40 model.add(Dense(10,activation=\'softmax\'))
41 print(model.summary())
42 model.compile(loss=\'categorical_crossentropy\',
43               optimizer=\'adam\',metrics=[\'accuracy\'])
44 try:
45     model=load_model("CnnModel.h5")
46     print("Load model successfully!Continuous training model!......")
47 except :
48     print("Failure of loading model!Start training a new model......")
49 
50 train_history=model.fit(x=x_img_train_normalize,
51                         y=y_label_train_OneHot,validation_split=0.2,
52                         epochs=5,batch_size=128,verbose=1)
53 model.save("CnnModel.h5")
54 print("Saved model to disk")
55 def show_train_history(train_history,train,validation):
56     plt.plot(train_history.history[train])
57     plt.plot(train_history.history[validation])
58     plt.title(\'Train History\')
59     plt.ylabel(train)
60     plt.xlabel(\'Epoch\')
61     plt.legend([\'train\',\'validation\'],loc=\'upper left\')    #显示左上角标签
62     plt.show()
63 show_train_history(train_history,\'acc\',\'val_acc\')
64 show_train_history(train_history,\'loss\',\'val_loss\')
65 scores=model.evaluate(x_img_test_normalize,y_label_test_OneHot)
66 print()
67 print(\'accuracy\',scores[1])
68 prediction=model.predict_classes(x_img_test_normalize)
69 print("prediction[:10]",prediction[:10])
70 import matplotlib.pyplot as plt
71 label_dict={0:"airplane",1:"automobile",2:"bird",3:"cat",4:"deer",5:"dog",6:"frog",7:"horse",8:"ship",9:"truck"}
72 def plot_image_labels_prediction_1(image,labels,prediction,idx,num=10):
73     fig=plt.gcf()
74     fig.set_size_inches(12,14)
75     if num>25:num=25
76     for i in range(0,num):
77         ax=plt.subplot(5,5,i+1)
78         ax.imshow(image[idx],cmap=\'binary\')
79         title=str(i)+\',\'+label_dict[labels[i][0]]
80         if len(prediction)>0:
81             title+="=>"+label_dict[prediction[i]]
82         ax.set_title(title,fontsize=10)
83         ax.set_xticks([]);ax.set_yticks([])
84         idx+=1
85     plt.show()
86 plot_image_labels_prediction_1(x_img_test,y_label_test,prediction,0,10)
87 Predicted_Probability=model.predict(x_img_test_normalize)
88 def show_Predicted_Probability(y,prediction,x_img_test,Predicted_Probability,i):
89     print(\'label:\',label_dict[y[i][0]],\'predict:\',label_dict[prediction[i]])
90     plt.figure(figsize=(2,2))
91     plt.imshow(np.reshape(x_img_test[i],(32,32,3)))
92     plt.show()
93     for j in range(10):
94         print(label_dict[j]+
95               \'Probability:%1.9f\'%(Predicted_Probability[i][j]))
96 show_Predicted_Probability(y_label_test,prediction,x_img_test,Predicted_Probability,0)
97 show_Predicted_Probability(y_label_test,prediction,x_img_test,Predicted_Probability,3)