This question could also be applied to SMPL.
I'm not an expert in human skeleton, and I got stuck in applying existing joint angle annotations from other human pose datasets(e.g. Human3.6M) to STAR.
When all zero angles are applied to the model, it returns T-pose.
From my understandings, joint index 1, 2, 3 have joint index 0(pelvis) as a parent.
So, when we measure the axis-angle using the vector from the position of joint 0 to the joint 1(in T-pose), for example, the angle would not be zero.
Is there some kind of reference vector and we measure the angle between the reference vector and our pose vector in axis-angle format?
Sorry for the weird term. they are just made-ups by myself.
This question could also be applied to SMPL. I'm not an expert in human skeleton, and I got stuck in applying existing joint angle annotations from other human pose datasets(e.g. Human3.6M) to STAR. When all zero angles are applied to the model, it returns T-pose. From my understandings, joint index 1, 2, 3 have joint index 0(pelvis) as a parent. So, when we measure the axis-angle using the vector from the position of joint 0 to the joint 1(in T-pose), for example, the angle would not be zero. Is there some kind of reference vector and we measure the angle between the reference vector and our pose vector in axis-angle format? Sorry for the weird term. they are just made-ups by myself.