fit smpl+D from raw scans only

[ddlau]

hello, I am appreciate about your excellent work, but something's remain unclear for me and i'm expecting responses: in this project, i see minimal shapes, smpl parameters and so on are used when evaluating, can this project fit smpl+D model from raw scans only? if yes, how? thank you!

[taconite]

Hi, thanks for your interest in our work!

The optimization code does SMPL+D fitting at this line:

https://github.com/taconite/PTF/blob/a8789c9f752aea2944c2a75e04cc2aa21c7e4a00/smpl_registration/fit_SMPLD_PTFs.py#L498

So you can feed your raw scan to the optimization script and get SMPL+D from raw scans.

best, Shaofei

[ddlau]

您好，非常感谢回复和指导！有一些细节想请教一下： 目前我这边测试用的数据，仅有一个含15万个顶点和5万个面片的clothed的mesh；在调用optimize_offsets时这个mesh应该是用来初始化all_posed_cloth_meshes吧？那init_smpl_meshes该怎么办呢？可以用这个mesh算吗？ 再次感谢！

[taconite]

init_smpl_meshes其实就是前面fit SMPL步骤输出的mesh，它们的作用regularize你的SMPL+D优化过程：

https://github.com/taconite/PTF/blob/a8789c9f752aea2944c2a75e04cc2aa21c7e4a00/smpl_registration/fit_SMPLD_PTFs.py#L138

这里的laplacian loss表明，你优化过后的SMPL+D mesh (th_smpl_meshes)相对于你的初始SMPL mesh（init_smpl_meshes）的形变应当尽量小。

所以优化的完整流程应该是：1、先优化SMPL parameter，得到SMPL mesh；2、用第一步得到SMPL parameter和SMPL mesh作为初始化，再去优化SMPL+D的parameter。第一步的SMPL优化很关键，如果不准的话，第二步的结果也不会好