Can i use images of varying sizes to train a unet_learner

i have a dataset of images of varying sizes. i need to do pixel wise annotation for training a segmentation model. but if a use a resize transform on the images and its masks, wouldnt it screw up the masks?

In that case, do i have to resize all the images to a fixed size image and then do the annotation?

Thanks in advance :slight_smile:

1 Like