问题描述
我正在关注 tensorflow 的卷积自动编码器教程,使用 tensorflow 2.0 和 keras,发现 here。
使用提供的代码构建 CNN,但在编码器和解码器中再添加一个卷积层会导致代码损坏:
class Denoise(Model):
def __init__(self):
super(Denoise,self).__init__()
self.encoder = tf.keras.Sequential([
layers.Input(shape=(28,28,1)),layers.Conv2D(16,(3,3),activation='relu',padding='same',strides=2),layers.Conv2D(8,## New Layer ##
layers.Conv2D(4,strides=2)
## --------- ##
])
self.decoder = tf.keras.Sequential([
## New Layer ##
layers.Conv2DTranspose(4,kernel_size=3,strides=2,padding='same'),## --------- ##
layers.Conv2DTranspose(8,layers.Conv2DTranspose(16,layers.Conv2D(1,kernel_size=(3,activation='sigmoid',padding='same')
])
def call(self,x):
encoded = self.encoder(x)
decoded = self.decoder(encoded)
return decoded
autoencoder = Denoise()
运行 autoencoder.encoder.summary()
和 autoencoder.decoder.summary()
,我可以看到这是一个形状问题:
Encoder:
Layer (type) Output Shape Param #
=================================================================
conv2d_124 (Conv2D) (None,14,16) 160
_________________________________________________________________
conv2d_125 (Conv2D) (None,7,8) 1160
_________________________________________________________________
conv2d_126 (Conv2D) (None,4,4) 292
=================================================================
Total params: 1,612
Trainable params: 1,612
Non-trainable params: 0
_________________________________________________________________
Decoder:
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_transpose_77 (Conv2DT (32,8,4) 148
_________________________________________________________________
conv2d_transpose_78 (Conv2DT (32,16,8) 296
_________________________________________________________________
conv2d_transpose_79 (Conv2DT (32,32,16) 1168
_________________________________________________________________
conv2d_127 (Conv2D) (32,1) 145
=================================================================
Total params: 1,757
Trainable params: 1,757
Non-trainable params: 0
_________________________________________________________________
为什么解码端的前导维度是32
?如果输入是从编码器传递的,为什么传入层的维度不是 None,4
?我该如何解决这个问题?
在此先感谢您的帮助!
解决方法
Keras 使用 32 作为默认的 batch_size。可能和这个有关系。但要解决此问题,您可以在解码器中包含 input_shape
参数。
self.decoder = tf.keras.models.Sequential([
## New Layer ##
layers.Conv2DTranspose(4,kernel_size=3,strides=2,activation='relu',padding='same',input_shape=(4,4,4)),## --------- ##
layers.Conv2DTranspose(8,padding='same'),layers.Conv2DTranspose(16,layers.Conv2D(1,kernel_size=(3,3),activation='sigmoid',padding='same')
])
此外,为了避免潜在问题和一致性,我将 tf.keras.models
模块用于模型(而不是 tf.keras.Model
)和 tf.keras.layers
用于图层。这可能不是问题..但可能会导致问题。
在最后一个编码器层中删除 stride=2
,并在最后一个解码器层中添加 stride=2
。
from tensorflow.keras import layers
from tensorflow.keras import Model
class Denoise(Model):
def __init__(self):
super(Denoise,self).__init__()
self.encoder = tf.keras.Sequential([
layers.Input(shape=(28,28,1)),layers.Conv2D(16,(3,strides=2),layers.Conv2D(8,## New Layer ##
layers.Conv2D(4,padding='same')
## --------- ##
])
self.decoder = tf.keras.Sequential([
## New Layer ##
layers.Conv2DTranspose(4,strides=2)
])
def call(self,x):
encoded = self.encoder(x)
decoded = self.decoder(encoded)
return decoded
autoencoder = Denoise()
autoencoder.build(input_shape=(1,1))
autoencoder.summary()