Thanks for the amazing work and especially for open-sourcing the code.
I read from the paper that SMPLicit also trained an auto-encoder to generate realistic pose dependent deformation of clothes when draped over the SMPL body. However, I don't really see this part reflected in the code. Instead, from my understanding, the code finds the correspondence of garment mesh with SMPL body model and then transform the mesh accordingly.
Is this intended? Or did you find simply deform the clothes meshes with skinning function and weights from SMPL good enough?
Hi,
Thanks for the amazing work and especially for open-sourcing the code.
I read from the paper that SMPLicit also trained an auto-encoder to generate realistic pose dependent deformation of clothes when draped over the SMPL body. However, I don't really see this part reflected in the code. Instead, from my understanding, the code finds the correspondence of garment mesh with SMPL body model and then transform the mesh accordingly.
Is this intended? Or did you find simply deform the clothes meshes with skinning function and weights from SMPL good enough?
Thanks.