Translate other human model to amp human model

[xizaoqu]

Hi! If I want to translate other human models (such as smplx), how can I get the right T-pose and other parameters like rotation_to_target_skeleton for retargeting motion? By the way, have you tried a more complex human model? Will a different human model affect performance?

[xizaoqu]

I've figured out the T-pose and rotation_to_target_skeleton. the T-pose is stored in the local_translation, the rotation_to_target_skeleton is [0.5, 0.5, 0.5, 0.5]. But how to compute the scale?

[xbpeng]

the scale is to convert between different units. So formats use feet, inch, cm, or m. We use m for our format, so the scale parameter is for converting whatever unit the original format was to m.

[xizaoqu]

the scale is to convert between different units. So formats use feet, inch, cm, or m. We use m for our format, so the scale parameter is for converting whatever unit the original format was to m.

Thanks for your reply. I don't know the unit of smplx from the SAMP dataset, so I directly set scale to 1 and it seems right. It actually controls the height of root joint. But another problem is the feet are not always on the floor. How to address this problem.

[3-body]

the "root_height_offset" of json file maybe what you wanted, or edit the z-axes of root link.

[ghy0626-design]

I've figured out the T-pose and rotation_to_target_skeleton. the T-pose is stored in the local_translation, the rotation_to_target_skeleton is [0.5, 0.5, 0.5, 0.5]. But how to compute the scale?

Hi, it looks like you have already solved this problem, can I ask how to obtain the T-pose?