I've extracted the weights from the filters of a pretrained model (AlexNet). I wish to represent these weights visually, this works fine for the first layer as there is only 3 input channels so I can represent the filters as color images with channels (RGB).

However this breaks down in later layers as they often have more than 3 input channels, I can represent them as many greyscale kernels but was hoping someone has found a better solution.

  • $\begingroup$ Seldom does it help due to the fact that they are small. You cannot find many meaningfull patterns that can help you. For instance, a $5\times5$ filter is very small. The papers usually visualise the outputs of a convolution. $\endgroup$ – Media Mar 21 at 21:03
  • $\begingroup$ Thank you am I correct in reffering to the outptus as a feature map? $\endgroup$ – Fraser Hamilton Mar 22 at 8:33
  • $\begingroup$ The outputs of convolutional layers are called activation maps or feature maps. $\endgroup$ – Media Mar 22 at 9:21
  • $\begingroup$ Thanks for the assistance I'll try those instead $\endgroup$ – Fraser Hamilton Mar 22 at 9:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.