Extracted batches 10 times bigger than the images and annotations

[OmarAshkar]

Hello,

I am trying to train a hovernet. I have WSI with 40X resolution. My images are about 500GB in total. When extracting the patches, I get terabytes returned and I am out of storage!

I think this is because of the parameter 'mirror' in the patching script. I am wondering if it would affect the training if If switched that to 'valid'

Also, I am wondering if I choose a lower resolution level like 10X resolution, would that affect hovernet design?

Thanks in advance!