I noticed that DexYCB has the left or the right hand. How did you do the cross dataset evaluation experiment compared to HO3D, which only has the right hand?
There are two ways that I can think of:
Mirror all left hand images to the right and only train a single right hand network.
Train a network that predicts both left and right hand on DexYCB, and compare with the right-handed network in HO3D.
I noticed that DexYCB has the left or the right hand. How did you do the cross dataset evaluation experiment compared to HO3D, which only has the right hand?
There are two ways that I can think of:
Could you help clarify? Thanks!