Hi after pulling down the latest master branch using the apply_tfms function on my masks, after transforming the input images seem to return the mask associated with the last transformed input image.
That’s normal since do_resolve=False doesn’t change the state of tfms[0] and leaves it at the same state it was last (all your masks correspond to the last image).
You should apply the transform to your mask with do_resolve=False immediately after doing it to the image (like in the docs).
1 Like