Hi,
I found a problem when calculating L1 loss and saw your TODO in l1_loss.py.
# TODO Handling an exception so that it can be handled even when the target data is empty
for a clean dataset, the main reason why target becomes to None is going through random crop augmention, so I simplely add a if ( ) in RandomCrop3D. When crop region is empty of target, just return original data and give up this augmention.
Hi, I found a problem when calculating L1 loss and saw your TODO in l1_loss.py.
for a clean dataset, the main reason why target becomes to None is going through random crop augmention, so I simplely add a if ( ) in RandomCrop3D. When crop region is empty of target, just return original data and give up this augmention.