Potentially update transformation handling to be in line with pytorch transforms. The suggestion being that A.Compose([]) objects are called using a lambda function before being passed into the data loader. Something like: transform = lambda x : A.Compose([...])[image=x]['image'].
Potentially update transformation handling to be in line with pytorch transforms. The suggestion being that
A.Compose([])
objects are called using a lambda function before being passed into the data loader. Something like:transform = lambda x : A.Compose([...])[image=x]['image']
.