Skip to content

Instantly share code, notes, and snippets.

@kretes

kretes/generate-cldice-examples-willgo.ipynb Secret

Last active September 9, 2021 08:26
Show Gist options
  • Save kretes/550ac2b58260504fb1b586fe5bab7634 to your computer and use it in GitHub Desktop.
Save kretes/550ac2b58260504fb1b586fe5bab7634 to your computer and use it in GitHub Desktop.
Download ZIP
generate-cldice-examples-willgo.ipynb
Display the source blob
Display the rendered blob
Raw
generate-cldice-examples-willgo.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
@jocpae
Copy link

jocpae commented Jul 23, 2021 •
edited

Hi kretes,
can you please change 2 things:

  1. you do not need to change the maxpool kernel
  2. it seems that the channel expansion, we mentioned in the previous comment has not been fixed (in our pytorch code the convention is channels before dimensions.)

I hope this solves the issue :)

@kretes
Copy link
Author

kretes commented Aug 20, 2021 •
edited

Hello, thanks @jocpae for looking at the issue.

I reverted the change in the maxpool kernel, but now the results look even more awkward.

I 've done the 'channels first' for pytorch as well - but it is more tensorflow that I care of first. Anyway - now both methods give same results.

Going back to the original issue - what do you think causes the discontinuity in the result of the soft-skeleton?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment