How to compute the average motion variance in ablation study?

[zhuoqun-chen]

Hi @zzhang-18 and @waffleboat,

Thank you for releasing the code for your TEDi paper.

I'm trying to train the model and test how the performance varies with different design choice. I'd like to compute the "average motion variance" metric that you report in Fig.9, but it seems it is not mentioned how is it computed? Do you have any reference about it?

Another small question is about the dimension of each frame (283) in the downloaded cmu dataset. What's the order of each bit? For example, I can guess that:
[0:3] for root_pos
[-4:] for contact
(rot) x (num_joints) = 6 x 30 = 180
(pos) x (num_joints) = 3 x 30 = 90

And in total 3 + 4 + 180 +90 = 277. What did I miss?

Any help would be appreciated! Thank you again!