I’m following the MarvelousDesigner tutorial, but ran into a problem after the first Rigid alignment…
Traceback (most recent call last):
File “FitMarvelousAvatar.py”, line 33, in
avatarWrapped = wrap.nonRigidRegistration(avatar,scan,controlPointsAvatar,controlPointsScan,freePolygonsAvatar,minNodes = 50,initialRadiusMultiplier = initialRadiusMultiplier,smoothnessFinal = smoothnessFinal,maxIterations = 20)
ValueError: GeomFloating contains 6 disconnected from each other polygon groups, must be 1. One can use wrap.subset() to extract main part.
Successfully used Female #2 avatar from Marvelous Designer 4.5 with non-rigid registration in R3DS::WrapX 1.3.4 now.
Either my first attempt at exporting the AVT file from MD, or UV unpacking with 3ds Max must have been unsuccessful… one step I definitely missed in MD was the deletion of the default cloth objects before exporting the AVT file (initially, I just hid the cloth, but the tutorial explicitly deletes it).
For some reason, my first avatar’s UV map also had a single UV projection when exported from R3DS for some reason (not separated into overlapping parts as expected), so I tried importing with “Import as Single Mesh” unchecked in 3ds Max to get separate UVs, and may have added another UV modifier accidentally. (It was hard to tell which modifier was chosen in the pull down menu as it was cut off during the video… but ultimately it was just Unwrap UVW)… and then I exported to OBJ, reimported that OBJ as a single mesh, and re-exported to OBJ again. That was definitely a bad process… (first time using 3ds Max). The tutorial is actually really good about pointing out potential problems though.
Ultimately, I had to redo the AvatarSrc.avt, AvatarSrc.obj, and AvatarSrcUniqueUVs.obj properly. Luckily this didn’t require using the same male mesh as the tutorial after all.
–
Another problem I’m having is that my RAW scan (from an Artec Showroom Shapify Booth in Palo Alto) has some badly misshapen hands, and the ears are obscured completely by hair. I tried hiding the entire ear regions and back of the head of the basemesh during the non-rigid registration step, but this didn’t seem to help… the ears have been positioned into the hair region anyway — am I using this feature incorrectly?
Would it be better to try patching up the RAW scan by replacing the hands and head regions using the corresponding parts from the basemesh, and starting the wrap process over again?
Overall, I am VERY impressed by WrapX so far. Great work!
In general I would not recommend to keep hair loose during scanning.
Hair is often a problem as it does not represented on basemesh topology. The most common way to solve this is to set all the polygons of the basemesh that lay under the hair as free polygons.
Please check how we do that in the video https://vimeo.com/122286361 (11th minute).
Please note that at the moment you click Ok button all the free polygons should be selected (not hidden). All the hidden polygons will be treated as unselected. If you do everything correctly all the free polygons will be marked as gray during non-rigid registration and they will try to stay as smooth as possible regardless of scan shape around them.
As the head shape is unknown under the hair (at least WrapX does not know how it looks like) WrapX will refer the original shape of the basemesh for those polygons, so please do not expect it to do a trick.