3DPCK metric in paper

[nicolasugrinovic]

Hi,

Congrats for the great work! I wanted to know how do you compute the metric 3DPCK reported in table 5? Specifically, do you use relative predicted joints (person-centered) or absolute joints (camera-centered) to calculate the results. In some papers they make the difference 3DPCK_rel and 3DPCK_abs. Which one do you report?

Thank you in advanced!