LeNet-5模型的keras实现

 1 import keras
 2 from keras.models import Sequential
 3 from keras.layers import Input,Dense,Activation,Conv2D,MaxPooling2D,Flatten
 4 from keras.datasets import mnist
 5 
 6 
 7 (x_train,y_train),(x_test,y_test) = mnist.load_data()
 8 x_train = x_train.reshape(-1, 28, 28, 1)    #######
 9 x_train = x_train.astype("float32")
10 print(x_train.shape)
11 y_train = y_train.astype("float32")
12 x_test = x_test.reshape(-1,28,28,1)
13 x_test = x_test.astype("float32")
14 y_test = y_test.astype("float32")
15 
16 print(y_train)
17 x_train /= 255
18 x_test /= 255
19 
20 from keras.utils import np_utils
21 y_train_new = np_utils.to_categorical(num_classes=10,y=y_train)
22 print(y_train_new)
23 y_test_new = np_utils.to_categorical(num_classes=10,y=y_test)
24 
25 def LeNet_5():
26     model = Sequential()
27     model.add(Conv2D(filters=6,kernel_size=(5,5),padding="valid",activation="tanh",input_shape=[28, 28, 1]))
28     model.add(MaxPooling2D(pool_size=(2,2)))
29     model.add(Conv2D(filters=16,kernel_size=(5,5),padding="valid",activation="tanh"))
30     model.add(MaxPooling2D(pool_size=(2,2)))
31     model.add(Flatten())
32     model.add(Dense(120,activation="tanh"))
33     model.add(Dense(84,activation="tanh"))
34     model.add(Dense(10,activation="softmax"))
35     return model
36 
37 def train_model():
38     model = LeNet_5()
39     model.compile(optimizer="adam",loss="categorical_crossentropy",metrics=["accuracy"])
40     model.fit(x_train,y_train_new,batch_size=64,epochs=1,verbose=1,validation_split=0.2,shuffle=True)
41     return model
42 
43 model = train_model()
44 
45 loss,accuracy = model.evaluate(x_test,y_test_new)
46 print(loss,accuracy)