retinamask
retinamask copied to clipboard
Why need include ground truth masks during training?
As mentioned in the paper "Thus, the number of mask proposals is (100+Gt) during training.", I am wondering whether including the GT mask for training mask head could increase accuracy?