search

Jeff-sjtu / HybrIK

Official code of "HybrIK: A Hybrid Analytical-Neural Inverse Kinematics Solution for 3D Human Pose and Shape Estimation", CVPR 2021
MIT License
1.17k stars 142 forks source link

about using the twist angle #119

Closed llpspark closed 1 year ago

llpspark commented 1 year ago

great work for 3D human. And I wonder to know where it is a mistake about using the twist angle code:https://github.com/Jeff-sjtu/HybrIK/blob/e62363c192b370fc671d2e289c338acda0c0c8c7/hybrik/models/layers/smpl/lbs.py#L742 why using i-1 to index the phis and not using i ?

Jeff-sjtu commented 1 year ago

Hi @llpspark, that's because we don't predict the twist angle for the root joint. Thus all indices should subtract 1.

llpspark commented 1 year ago

@Jeff-sjtu I'm very glad to communicate with you again here. got it. Because the root node has three child nodes, the unique twist cannot be obtained. close it.