hello, I'm doing some work about 2D keypoints regression, want to ask some questions on your experiments，THX

[liangyufz]

In the raw paper, the 2D pck auc of the RHD dataset is about 0.89, what about the auc of yours？

And the loss of p and p-gt, how did you generate your groundtruth? Put gaussian response on the keypoints? Or just calculate difference between the predicted (x, y) and the groundtruth (x, y)?

It would be great if you could reply to me.

Thank you a lot and my email is [email protected]