[RandomZoom] Supporting Segmentation Masks #2010
Conversation
| self.get_zoom_matrix(zooms, mask_hd, mask_wd), | ||
| fill_mode=self.fill_mode, | ||
| fill_value=self.fill_value, | ||
| interpolation=self.interpolation, |
There was a problem hiding this comment.
We need to use "nearest" interpolation for segmentation masks to avoid encoding issues for sparse-encoded masks.
|
@ianstenbit how would you want me to add the numeric tests for the segmentation masks? |
A fixed-input test is fine with me (e.g. setting the seed of a random input and verifying the numerics of the output against a known set of values), or we can just rely on the demo results which is also fine with me. |
|
@ianstenbit thanks for the review!
I think this PR is in a good shape to be merged. |
|
@ariG23498 please take a look at the linter error. Once that is fixed I shall merge -- thank you! |
|
@ianstenbit over to you now! |
|
/gcbrun |
* chore: initial commit * chore: adding initial tests * update: readme for preprocessing * fix lint
Related #1992
CC: @ianstenbit
Working Colab Notebook