[Question] Relation between results and varying architecture and data augmentation

[kirk86]

Hi,

Given that the underlying network architecture varies among resnet18 & resnet50 between datasets, in addition to the data augmentations, I was wondering if you have any insights on what the impact would be in terms of the results presented on table 2, if we were to keep the same underlying architecture and same augmentations for all datasets?

Also, what was the optimizer used? Adam?

Thanks!