Skip to content

Remove option for specific interpolation mode and antialias for segmentation kernels #5370

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
pmeier opened this issue Feb 3, 2022 · 2 comments · Fixed by #5406
Closed

Remove option for specific interpolation mode and antialias for segmentation kernels #5370

pmeier opened this issue Feb 3, 2022 · 2 comments · Fixed by #5406
Assignees
@pmeier
Labels
enhancement module: transforms prototype

Comments

@pmeier
Copy link
Collaborator

pmeier commented Feb 3, 2022 •
edited by pytorch-bot bot
Loading

We typically use nearest interp mode for segmenation masks resize and there is no antialias option for nearest mode. I think we can remove antialias argument.

Originally posted by @vfdev-5 in #5323 (comment)

cc @vfdev-5 @datumbox @bjuncek
@pmeier pmeier self-assigned this Feb 3, 2022
@pmeier pmeier mentioned this issue Feb 10, 2022
Merged
@datumbox
Copy link
Contributor

Yes correct. masks are a geographical positioning of class Ids. We can't do any math operation that includes averaging or numeric functions on these pixel values. We can do only nearest-interpolation logic.

@pmeier pmeier linked a pull request Feb 11, 2022 that will close this issue
cleanup features / transforms feature branch #5406
Merged
@pmeier
Copy link
Collaborator Author

pmeier commented Feb 15, 2022

Closed in #5407.

@pmeier pmeier closed this as completed Feb 15, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@pmeier pmeier

Labels
enhancement module: transforms prototype
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

cleanup features / transforms feature branch pmeier/vision
2 participants
@datumbox @pmeier