TTA Segmentation Fastaiv2

if we use dsets for segmentation, then will we have to take care of affine transformations on masks for corresponding ones over images ?