问题描述
我正在使用Python 3.7.7和Tensorflow 2.1.0。
我想创建一个VGG16自动编码器网络，向其中加载权重文件，然后获取其编码器和解码器。
def get_vgg16_encoder(input_size=(200,200,1)):
inputs = Input(input_size,name='input')
conv1 = Conv2D(64,(3,3),activation='relu',padding='same',name='conv1_1')(inputs)
conv1 = Conv2D(64,name='conv1_2')(conv1)
pool1 = MaxPooling2D(pool_size=(2,2),strides=(2,name='pool_1')(conv1)
conv2 = Conv2D(128,name='conv2_1')(pool1)
conv2 = Conv2D(128,name='conv2_2')(conv2)
pool2 = MaxPooling2D(pool_size=(2,name='pool_2')(conv2)
conv3 = Conv2D(256,name='conv3_1')(pool2)
conv3 = Conv2D(256,name='conv3_2')(conv3)
conv3 = Conv2D(256,name='conv3_3')(conv3)
pool3 = MaxPooling2D(pool_size=(2,name='pool_3')(conv3)
conv4 = Conv2D(512,name='conv4_1')(pool3)
conv4 = Conv2D(512,name='conv4_2')(conv4)
conv4 = Conv2D(512,name='conv4_3')(conv4)
pool4 = MaxPooling2D(pool_size=(2,name='pool_4')(conv4)
conv5 = Conv2D(512,name='conv5_1')(pool4)
conv5 = Conv2D(512,name='conv5_2')(conv5)
conv5 = Conv2D(512,name='conv5_3')(conv5)
pool5 = MaxPooling2D(pool_size=(2,name='pool_5')(conv5)
return pool5,inputs
def get_vgg16_decoder(pool5):
upsp1 = UpSampling2D(size = (2,name = 'upsp1')(pool5)
conv1 = Conv2D(512,3,activation = 'relu',padding = 'same',name = 'conv1_1_dec')(upsp1)
conv1 = Conv2D(512,name = 'conv1_2_dec')(conv1)
conv1 = Conv2D(512,name = 'conv1_3_dec')(conv1)
upsp2 = UpSampling2D(size = (2,name = 'upsp2')(conv1)
conv2 = Conv2D(512,name = 'conv2_1_dec')(upsp2)
conv2 = Conv2D(512,name = 'conv2_2_dec')(conv2)
conv2 = Conv2D(512,name = 'conv2_3_dec')(conv2)
zero1 = ZeroPadding2D(padding=((1,0),(1,0)),data_format='channels_last',name='zero1')(conv2)
upsp3 = UpSampling2D(size = (2,name = 'upsp3')(zero1)
conv3 = Conv2D(256,name = 'conv3_1_dec')(upsp3)
conv3 = Conv2D(256,name = 'conv3_2_dec')(conv3)
conv3 = Conv2D(256,name = 'conv3_3_dec')(conv3)
upsp4 = UpSampling2D(size = (2,name = 'upsp4')(conv3)
conv4 = Conv2D(128,name = 'conv4_1_dec')(upsp4)
conv4 = Conv2D(128,name = 'conv4_2_dec')(conv4)
upsp5 = UpSampling2D(size = (2,name = 'upsp5')(conv4)
conv5 = Conv2D(64,name = 'conv5_1_dec')(upsp5)
conv5 = Conv2D(64,name = 'conv5_2_dec')(conv5)
conv6 = Conv2D(1,name = 'conv6_dec')(conv5)
return conv6
def get_vgg16(img_shape=(200,1)):
enc = get_vgg16_encoder(img_shape)
dec = get_vgg16_decoder(enc[0])
model = Model(inputs=enc[-1],outputs=dec)
return model
主要功能是：
def get_vgg_encoder_decoder_trained(img_shape):
# Get the model
model = get_vgg16(img_shape)
# Compile it.
model.compile(tf.keras.optimizers.Adam(lr=(1e-4) * 2),loss='binary_crossentropy',metrics=['accuracy'])
# Add the weights.
path_to_weights = os.path.join(weights_path,weights_filename)
model.load_weights(path_to_weights)
# Get the encoder
first_encoder_layer = 0
last_encoder_layer = 18
encoder = Model(inputs=model.layers[first_encoder_layer].input,outputs=model.layers[last_encoder_layer].output,name='encoder')
# extract decoder fitted weights
restored_w = []
for w in model.layers[last_encoder_layer + 1:]:
restored_w.extend(w.get_weights())
# reconstruct decoder architecture setting the fitted weights
new_inp = [Input(l.shape[1:]) for l in get_vgg16_encoder(img_shape)]
new_dec = get_vgg16_decoder(new_inp[-1])
decoder = Model(new_inp,new_dec)
decoder.set_weights(restored_w)
return encoder,decoder
但是当我尝试恢复权重时出现以下错误：
ValueError: Layer weight shape (3,1,512) not compatible with provided weight shape (3,512,512)
new_inp = [Input(l.shape[1:]) for l in get_vgg16_encoder(img_shape)]
new_dec = get_vgg16_decoder(new_inp[-1])
有什么建议吗？
解决方法
我已解决了修改最后一段代码的错误：
# reconstruct decoder architecture setting the fitted weights
new_inp = [Input(l.shape[1:]) for l in get_vgg16_encoder(img_shape)]
new_dec = get_vgg16_decoder(new_inp[0])
decoder = Model(new_inp[0],new_dec)
decoder.set_weights(restored_w)