Not able to replicate NYUv2 KD-ORI result

[kaushal-py]

Hi, I was not able to replicate the KD-ORI results for NYUv2 dataset. Is it possible for you to release code for that? If not, can you just give a code snippet of the forward pass and loss functions used?

[Lyan-ing]

the teacher accuracy of NYUv2 offer in the github not fit the accuacy in paper, can you give a correct teacher model weight. Thank you