global positions and orientations exhibit large deviations from the ground truth,

[gh786137611]

We would like to conduct a quantitative comparison with your approach. After visualizing the weights provided in your GitHub repository, we noticed that the global positions and orientations exhibit large deviations from the ground truth, while the MPJPE values reported in your paper are quite small. We were wondering if you could kindly clarify how the evaluation was performed? We look forward to your reply.

[gh786137611]

May I ask whether you first perform a global translation and rotation alignment before computing the MPJPE?

[zzk273]

The MPJPE in paper is computed in local root frame. We do not emphasize global tracking accuracy in this work.

[gh786137611]

Thanks for your replay. it really helps