Styjun

Is expanding the research of a group into machine learning as a PhD student risky?

Where does the Z80 processor start executing from?

What Brexit proposals are on the table in the indicative votes on the 27th of March 2019?

How can I kill an app using Terminal?

Risk of infection at the gym?

What does the word "Atten" mean?

Is HostGator storing my password in plaintext?

Is there a good way to store credentials outside of a password manager?

Anatomically Correct Strange Women In Ponds Distributing Swords

Do all network devices need to make routing decisions, regardless of communication across networks or within a network?

Detecting if an element is found inside a container

Applicability of Single Responsibility Principle

Is `x >> pure y` equivalent to `liftM (const y) x`

Customer Requests (Sometimes) Drive Me Bonkers!

Purchasing a ticket for someone else in another country?

Hostile work environment after whistle-blowing on coworker and our boss. What do I do?

Is there a problem with hiding "forgot password" until it's needed?

How do I go from 300 unfinished/half written blog posts, to published posts?

How easy is it to start Magic from scratch?

How does it work when somebody invests in my business?

How did Arya survive the stabbing?

Why escape if the_content isnt?

Pole-zeros of a real-valued causal FIR system

Implement the Thanos sorting algorithm

How does the second convolutional layer in tensorflow keras work?

Calculating size of output of a Conv layer in CNN modelImplement Character Convolution in KerasKeras Conv2D layer outputs array filled with NaNConvolutional layer output sizeOutput of conv2d in kerasHow to use Padding in conv2d layer of specific sizeConvolutional layer interpolation scaling in keras & tensorflowHow to mimic Caffe's max pooling behavior in Keras/Tensorflow?keras conv2d layers does not include second channel into trainingHow to compute the number of parameters in the second conversional layer?

0

I have the fllowing model in keras.

model = Sequential()

model.add(Conv2D(4, (3, 3), input_shape=input_shape, name='Conv2D_0', padding = 'same', use_bias=False, activation=None))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Conv2D(8, (3, 3), name='Conv2D_1', padding='same', use_bias=False, activation=None))
model.add(MaxPooling2D(pool_size=(2, 2)))

input_shape is (32, 32). So, for the first layer, if I have a an image of size (32, 32), I get 4 images of size (32, 32). So the input image is convoluted with 4 diffrent kernels. After the pooling layer, I get 4 images of size (16, 16).

The second convolutional layer gives me 8 images of size (16, 16). This layer has

4*8 kernels. The kernels have the size (3, 3, 4, 8). But I don't get, how the 8 output images are computed.

I thought for example for the first image I can do sth like:
H_i : i-th output image of the first Pooling layer
Ker_i : i-th kernel. (:, :, i, 0)

So the first output image of the second convolutional layer could be:
conv(H_0, ker_0) + conv(H_1, ker_1) + conv(H_2, ker_2) + conv(H_3, ker_3)

But this seems to be wrong.

Can anyone explaine me, how the second conv-layer computes the output images?
Thank you for your help.

python tensorflow keras

share|improve this question

asked Mar 21 at 15:54

5yn4x5yn4x

62

• 
  
  
  

  
  

  
  
  
  
  
  

  
  
  the same way as in first layer. number of filters becomes number of channels in the output. take a look cs231n.github.io/convolutional-networks
  
  – Sharky
  Mar 21 at 16:13
  

  
  
  
  


• 
  
  
  

  
  

  
  
  
  
  
  

  
  
  This layer has 4*8 kernels. - How do you define kernel?
  
  – Vlad
  Mar 21 at 16:49
  

  
  
  
  


• 
  
  
  

  
  

  
  
  
  
  
  

  
  
  I mean convolutional kernel (en.wikipedia.org/wiki/Kernel_(image_processing)).
  
  – 5yn4x
  Mar 21 at 19:36
  
  

  
  
  
  


• 
  
  
  

  
  

  
  
  
  
  
  

  
  
  "the same way as in first layer. number of filters becomes number of channels in the output. take a look" But this time I have 4 input images. How are the filters applied on these 4 images.
  
  – 5yn4x
  Mar 21 at 19:37
  
  

  
  
  
  


• 
  
  
  

  
  

  
  
  
  
  
  

  
  
  I found the answer. The filter kernels are flipped!
  
  – 5yn4x
  Mar 22 at 15:49
  
  

  
  
  
  

add a comment | 

0

I have the fllowing model in keras.

model = Sequential()

model.add(Conv2D(4, (3, 3), input_shape=input_shape, name='Conv2D_0', padding = 'same', use_bias=False, activation=None))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Conv2D(8, (3, 3), name='Conv2D_1', padding='same', use_bias=False, activation=None))
model.add(MaxPooling2D(pool_size=(2, 2)))

input_shape is (32, 32). So, for the first layer, if I have a an image of size (32, 32), I get 4 images of size (32, 32). So the input image is convoluted with 4 diffrent kernels. After the pooling layer, I get 4 images of size (16, 16).

The second convolutional layer gives me 8 images of size (16, 16). This layer has

4*8 kernels. The kernels have the size (3, 3, 4, 8). But I don't get, how the 8 output images are computed.

I thought for example for the first image I can do sth like:
H_i : i-th output image of the first Pooling layer
Ker_i : i-th kernel. (:, :, i, 0)

So the first output image of the second convolutional layer could be:
conv(H_0, ker_0) + conv(H_1, ker_1) + conv(H_2, ker_2) + conv(H_3, ker_3)

But this seems to be wrong.

Can anyone explaine me, how the second conv-layer computes the output images?
Thank you for your help.

python tensorflow keras

share|improve this question

asked Mar 21 at 15:54

5yn4x5yn4x

62

• 
  
  
  

  
  

  
  
  
  
  
  

  
  
  the same way as in first layer. number of filters becomes number of channels in the output. take a look cs231n.github.io/convolutional-networks
  
  – Sharky
  Mar 21 at 16:13
  

  
  
  
  


• 
  
  
  

  
  

  
  
  
  
  
  

  
  
  This layer has 4*8 kernels. - How do you define kernel?
  
  – Vlad
  Mar 21 at 16:49
  

  
  
  
  


• 
  
  
  

  
  

  
  
  
  
  
  

  
  
  I mean convolutional kernel (en.wikipedia.org/wiki/Kernel_(image_processing)).
  
  – 5yn4x
  Mar 21 at 19:36
  
  

  
  
  
  


• 
  
  
  

  
  

  
  
  
  
  
  

  
  
  "the same way as in first layer. number of filters becomes number of channels in the output. take a look" But this time I have 4 input images. How are the filters applied on these 4 images.
  
  – 5yn4x
  Mar 21 at 19:37
  
  

  
  
  
  


• 
  
  
  

  
  

  
  
  
  
  
  

  
  
  I found the answer. The filter kernels are flipped!
  
  – 5yn4x
  Mar 22 at 15:49
  
  

  
  
  
  

add a comment | 

I have the fllowing model in keras.

model = Sequential()

model.add(Conv2D(4, (3, 3), input_shape=input_shape, name='Conv2D_0', padding = 'same', use_bias=False, activation=None))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Conv2D(8, (3, 3), name='Conv2D_1', padding='same', use_bias=False, activation=None))
model.add(MaxPooling2D(pool_size=(2, 2)))

input_shape is (32, 32). So, for the first layer, if I have a an image of size (32, 32), I get 4 images of size (32, 32). So the input image is convoluted with 4 diffrent kernels. After the pooling layer, I get 4 images of size (16, 16).

The second convolutional layer gives me 8 images of size (16, 16). This layer has

4*8 kernels. The kernels have the size (3, 3, 4, 8). But I don't get, how the 8 output images are computed.

I thought for example for the first image I can do sth like:
H_i : i-th output image of the first Pooling layer
Ker_i : i-th kernel. (:, :, i, 0)

So the first output image of the second convolutional layer could be:
conv(H_0, ker_0) + conv(H_1, ker_1) + conv(H_2, ker_2) + conv(H_3, ker_3)

But this seems to be wrong.

Can anyone explaine me, how the second conv-layer computes the output images?
Thank you for your help.

python tensorflow keras

share|improve this question

asked Mar 21 at 15:54

5yn4x5yn4x

62

model = Sequential()

model.add(Conv2D(4, (3, 3), input_shape=input_shape, name='Conv2D_0', padding = 'same', use_bias=False, activation=None))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Conv2D(8, (3, 3), name='Conv2D_1', padding='same', use_bias=False, activation=None))
model.add(MaxPooling2D(pool_size=(2, 2)))

share|improve this question

asked Mar 21 at 15:54

5yn4x5yn4x

62

share|improve this question

asked Mar 21 at 15:54

5yn4x5yn4x

62

asked Mar 21 at 15:54

5yn4x5yn4x

62

asked Mar 21 at 15:54

5yn4x5yn4x

62