we are trying to wrap clothes (wrap.latticeDeform) on user scans which have been non rigidly registrated before to an avatar generated with MakeHuman. Sometimes the process works fairly well, sometimes not at all. The process is very sensitive to the “radius” parameter: for some values the latticeDeform generates very strong artifacts, which might go away with larger radius values which in turn might lead to other problems, i.e. “holes” in the output. See attached screenshot. Is there anything we can do to stabilize the latticeDeform process?
Another small problem for which me have a workaround: the latticeDeform removes “s” and “usemtl” statements from the object file. We have to re-insert them after the process to get the textures showing up.
Yes Lattice operator is a very primitive tool. It’s pretty much the same like in Maya and Houdini. Please do not expect much from it.
Cloth transfer is a complicated problem in general case. All the solutions we know (RBF interpolation, Lattice deformer, Cage-based deformation) suffers when there is a significant distance between the cloth and the body especially in case of skirts.
Not sure if it will help but there is another version of Lattice deformer called wrap.latticeDeformAdaptive
For each vertex it chooses adaptive radius as a distance to N’th nearest vertex on target surface. The arguments are the same as in wrap.latticeDeform (http://www.russian3dscanner.com/docs/WrapX/wrapFunctions.html#wrap.latticeDeform) except for the last one. Pass a number of neightbours instead of the radius.
If anyone know better solution we would love to implement it.
Hello Andrew,
thank you very much! This deformer version should be mentioned in the documentation, because for our cases it works much better than the documented “latticeDeform”. I got very good results with last parameter countOfNeighbourhoodControlVertices ~ 20 and using this value for other clothes we prepared so far gave the same good results.
Best regards,
Rüdiger