I have been looking to use the fastai image augmentation pipeline for images that have six channels (ie biological flourescent images). I can see that most of the image transforms work on PILImages and know that PIL does not support multi-channel images. I have seen some work to use the albumentations library to create a pipeline, as in this thread Normalizing 6-channel images for pretrained models, but wondered whether the is any way to use at least the after batch transformations since that will allow on GPU operation and will likely be much faster.
Thanks for any suggestions