Do you selecet 'difficult'=0 and 'difficult'=1 samples for each category, when spliting VOC datasets?

[Parsifal133]

Thanks for your amazing works and open-source repositories from ORE, META-IOD to ELI.

According to your code https://github.com/JosephKJ/ELI/blob/93ddce7c35b1b38a55cf237fc3f51e36ba92fbc2/detection/detectron2/data/datasets/pascal_voc.py#LL56C1-L57C70
I assume that you select both 'difficult'=0 and 'difficult'=1 samples for each category, when spliting VOC datasets.
Therefore, there are 3002 training imgs in first10 dataset, and 3340 training imgs in last 10 dataset.

Please let me know if i was wrong.