Hi, Do you have the list of "common public segmentation datasets" used to train SAM (mentioned in Page 6 of the paper)?