I'm trying to understand the deconv referenced in the paper Visualizing and Understanding Convolutional Networks

The paper states (section 2, p. 3):

the deconvnet uses transposed versions of the same filters, but applied to the rectified maps

Is it possible to implement this step in a short code example? Given an unpooled, rectified map; how would the transposed filter be applied against it?

I did try looking at the referenced paper Adaptive Deconvolutional Networks for Mid and High Level Feature Learning. However, I'm not wrapping my head around its explanations too well; and it references a third paper with regard to its work on "layers of convolutional sparse coding" (deconvolution [M. Zeiler, D. Krishnan, G. Taylor, and R. Fergus. Deconvolutional networks. In CVPR, 2010]), but this 2010 paper appears to require access to download.

  • $\begingroup$ Hi. Please, put your specific question in the title. "Understand “Visualizing and Understanding Convolutional Networks”" is not a question and it would be too broad anyway. It seems that you're asking here "How does deconvolution work?". Have you tried to read something about the topic before asking this question? If no, it's probably a good idea to read something about the topic before proceeding. If yes, please, edit your post to explain what you have not understood. $\endgroup$
    – nbro
    Feb 25 at 9:33
  • 1
    $\begingroup$ Thanks for the suggestions for clarifying the question and edits! $\endgroup$
    – brent
    Mar 1 at 8:09

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.