Skip to content

[prototype] Adjust solarize threshold on input type #6874

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.

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

Merged
merged 5 commits into from
Nov 1, 2022

Conversation

datumbox
Copy link
Contributor

@datumbox datumbox commented Oct 31, 2022

Fixes #6840

This PR makes it possible for float images to use AutoAugmentation methods. Given that now all operators support float images, the only change necessary is to adjust the thresholds of solarize to depend on the max value of the input type (1.0 for floats, 255 for integers or PIL).

cc @vfdev-5 @bjuncek @pmeier

@@ -137,7 +137,8 @@ def _apply_image_or_video_transform(
elif transform_id == "Posterize":
return F.posterize(image, bits=int(magnitude))
elif transform_id == "Solarize":
return F.solarize(image, threshold=magnitude)
bound = 1.0 if image.is_floating_point() else 255.0
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@pmeier On your #6830, this should grab the value from _max_value(image.dtype)

@datumbox datumbox changed the title Adjust solarize threshold on input type [prototype] Adjust solarize threshold on input type Oct 31, 2022
Copy link
Collaborator

@pmeier pmeier left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

One question inline. Otherwise LGTM. Thanks Vasilis!

@datumbox datumbox merged commit c35e054 into pytorch:main Nov 1, 2022
@datumbox datumbox deleted the prototype/aa_solarize branch November 1, 2022 08:36
facebook-github-bot pushed a commit that referenced this pull request Nov 4, 2022
Summary:
* Adjust solarize threshold on input type

* Handle PIL images

* minor refactor

* Fix linter

Reviewed By: datumbox

Differential Revision: D41020549

fbshipit-source-id: 100cf505acd93e86d46e193addc99fbb4c23e3d0
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Support all integer and floating point dtypes in prototype transform kernels?
3 participants