-
Couldn't load subscription status.
- Fork 7.2k
Closed
Description
In some scenarios (like semantic segmentation), we might want to apply the same random transform to both the input and the GT labels (cropping, flip, rotation, etc).
I think we can get this behaviour emulated in a segmentation dataset class by resetting the random seed before calling the transform for the labels.
This sound a bit fragile though.
One other possibility is to have the transforms accept both inputs and targets as arguments.
Do you have any better solutions?
mbarnes1, jizongFox, nd26, ababycat, krumo and 44 more
Metadata
Metadata
Assignees
Labels
No labels