keras模拟resnet实现带shortcut模型

1.模型代码

import keras
from keras import Input, Model
from keras.datasets import cifar10, cifar100
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten, Add
from keras.layers import Conv2D, MaxPooling2D
from keras.optimizers import Adam

num_classes = 20
model_name = 'cifar10.h5'

# The data, shuffled and split between train and test sets:
# (x_train, y_train), (x_test, y_test) = cifar10.load_data()
(x_train, y_train), (x_test, y_test) = cifar100.load_data(label_mode='coarse')

print(x_train.shape)
x_train = x_train.astype('float32')/255
x_test = x_test.astype('float32')/255

# Convert class vectors to binary class matrices.
y_train = keras.utils.to_categorical(y_train, num_classes)
y_test = keras.utils.to_categorical(y_test, num_classes)

inputs = Input(shape=x_train.shape[1:], dtype='float32')

cov1 = Conv2D(32, (3, 3), padding='same')(inputs)
act1 = Activation('relu')(cov1)

# max_pool1 = MaxPooling2D(pool_size=(2, 2))(act1)
# drop1 = Dropout(0.25)(max_pool1)

cov2 = Conv2D(3, (3, 3), padding='same')(act1)
act2 = Activation('relu')(cov2)

# max_pool2 = MaxPooling2D(pool_size=(2, 2))(act2)
# drop2 = Dropout(0.25)(max_pool2)

add = Add()([inputs, act2])

flat = Flatten()(add)

dense1 = Dense(512)(flat)
act3 = Activation('relu')(dense1)
drop3 = Dropout(0.5)(act3)

dense2 = Dense(num_classes)(drop3)
outputs = Activation('softmax')(dense2)

model = Model(input=inputs, output=outputs)
model.summary()

# initiate RMSprop optimizer
# opt = keras.optimizers.rmsprop(lr=0.001, decay=1e-6)
opt = Adam(lr=0.01, beta_1=0.9, beta_2=0.999, epsilon=1e-08)
# train the model using RMSprop
model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy'])

model.fit(x_train, y_train, epochs=1000, batch_size=50000, shuffle=True, validation_data=(x_test, y_test))
model.save(model_name)

# evaluate
loss, accuracy = model.evaluate(x_test, y_test)
print('loss: %f, accurcy: %f' % (loss, accuracy))

2.模型结构

Layer (type)                    Output Shape         Param #     Connected to                     
  model = Model(input=inputs, output=outputs)
==================================================================================================
input_1 (InputLayer)            (None, 32, 32, 3)    0                                            
__________________________________________________________________________________________________
conv2d_1 (Conv2D)               (None, 32, 32, 32)   896         input_1[0][0]                    
__________________________________________________________________________________________________
activation_1 (Activation)       (None, 32, 32, 32)   0           conv2d_1[0][0]                   
__________________________________________________________________________________________________
conv2d_2 (Conv2D)               (None, 32, 32, 3)    867         activation_1[0][0]               
__________________________________________________________________________________________________
activation_2 (Activation)       (None, 32, 32, 3)    0           conv2d_2[0][0]                   
__________________________________________________________________________________________________
add_1 (Add)                     (None, 32, 32, 3)    0           input_1[0][0]                    
                                                                 activation_2[0][0]               
__________________________________________________________________________________________________
flatten_1 (Flatten)             (None, 3072)         0           add_1[0][0]                      
__________________________________________________________________________________________________
dense_1 (Dense)                 (None, 512)          1573376     flatten_1[0][0]                  
__________________________________________________________________________________________________
activation_3 (Activation)       (None, 512)          0           dense_1[0][0]                    
__________________________________________________________________________________________________
dropout_1 (Dropout)             (None, 512)          0           activation_3[0][0]               
__________________________________________________________________________________________________
dense_2 (Dense)                 (None, 20)           10260       dropout_1[0][0]                  
__________________________________________________________________________________________________
activation_4 (Activation)       (None, 20)           0           dense_2[0][0]                    
==================================================================================================