keras 的 example 文件 mnist_denoising_autoencoder.py 解析

mnist_denoising_autoencoder.py 是一个编解码神经网络，其意义就是如果图片中有噪点的话，可以去除噪点，还原图片

其编码网络为：

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
encoder_input (InputLayer)   (None, 28, 28, 1)         0
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 14, 14, 32)        320
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 7, 7, 64)          18496
_________________________________________________________________
flatten_1 (Flatten)          (None, 3136)              0
_________________________________________________________________
latent_vector (Dense)        (None, 16)                50192
=================================================================
Total params: 69,008
Trainable params: 69,008
Non-trainable params: 0
_________________________________________________________________

 

就是输入一张图片，生成一个16维的向量

 

其解码网络为：

__________________________________________________________________________________________
Layer (type)                            Output Shape                        Param #
==========================================================================================
decoder_input (InputLayer)              (None, 16)                          0
__________________________________________________________________________________________
dense_1 (Dense)                         (None, 3136)                        53312
__________________________________________________________________________________________
reshape_1 (Reshape)                     (None, 7, 7, 64)                    0
__________________________________________________________________________________________
conv2d_transpose_1 (Conv2DTranspose)    (None, 14, 14, 64)                  36928
__________________________________________________________________________________________
conv2d_transpose_2 (Conv2DTranspose)    (None, 28, 28, 32)                  18464
__________________________________________________________________________________________
conv2d_transpose_3 (Conv2DTranspose)    (None, 28, 28, 1)                   289
__________________________________________________________________________________________
decoder_output (Activation)             (None, 28, 28, 1)                   0
==========================================================================================
Total params: 108,993
Trainable params: 108,993
Non-trainable params: 0
__________________________________________________________________________________________

就是输入一个16维的向量，生成一个 28*28 的黑白图片

 

合并之后的网络结构就是

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
encoder_input (InputLayer)   (None, 28, 28, 1)         0
_________________________________________________________________
encoder (Model)              (None, 16)                69008
_________________________________________________________________
decoder (Model)              (None, 28, 28, 1)         108993
=================================================================
Total params: 178,001
Trainable params: 178,001
Non-trainable params: 0
_________________________________________________________________

输入就是有噪音的图片，输出是原图，损失函数是mse，均方差

效果如下：第一行是原图，第二行是加上噪点之后的图，第三行是解码出来的图

 

——————————————————————

总目录

keras的example文件解析