OMG!! MD garment with decent quad mesh!



  • I was happily doodling tonight in the latest version of Marvelous Designer on a simple 'cut your basics' for PE, so a sort of fitting geometry with edges the way you would cut the patterns for clothing.
    I decided that a quad mesh would be better to recognize the general shape. I switched on the mesh view to see the tesult and, behold!!

    0_1534366354196_Knipsel.JPG



  • Let me see it on PE fully rendered



  • @F_Verbaas interesting to note the meshing asymmetry. Not that clothes have to be symmetrical, of course.


  • Poser Ambassadors

    Yep, no symmetry was my first thought also.



  • And over 50 pattern pieces. Sheesh. My mother was a seamstress who mostly used Vogue patterns, and she never worked with patterns that looked anything like that, no matter how much detail there was.



  • @Miss-B I'd certainly agree, in the sense that a human seamstress needs to optimise her workflow, and accurately cutting patterns from bulk cloth is time consuming, but in the digital world, that's not necessarily a consideration of the same magnitude, since the application won't spend hours hand or physical machine stitching extra seams, but mere microseconds.

    Without direct MD experience, I also can't judge whether that's a setting that could be adjusted for actual manufacturing of designs into real world clothing.



  • @anomalaus My MD experience is limited, but I still can't see why some of those pieces can't be "stitched" together to make larger pieces so you could wind up with say half the amount.

    I haven't seen a render of what this is supposed to look like, but I've seen pieces of clothing with seams that weren't necessary. For instance, a seam at the waist when the texturing didn't show a seam at the waist, and there's no belt to cover it. The same "might" also be true for a seam down the middle of the front AND the back. I can see the back if you want the option to put in a digital zipper, but what's the purpose of a front center seam, unless you want an open neck. Having both of those options in one top just doesn't seem realistic to me.

    I can do more with texturing in details, than actually having a lot of pieces in a mesh itself. Then again, I was a 2D graphic artist long before I got into 3D, and my first attempts at working with a 3D mesh a friend had created, was to create a whole bunch of textures for it.

    That said, I know a lot of folks would prefer to add details, to some extent, in the mesh itself, but where clothing is concerned, I'd rather do it with the proper set of textures, and that doesn't always mean a 50+ piece UV Map. Just my 2¢ FWIW.



  • @Miss-B fully agree. It's pretty obvious that MD's unwrapping is the exact opposite reverse of physical clothing construction, where the (usually pre-textured) cloth is minimally cut and stitched, always starting with the flat, bulk cloth and applying intelligence and experience to get the best use of cloth and minimise offcuts. Pattern matching at seams is also a consideration. With UV unwrapping, the flattening process has different priorities, trying to minimise texture shearing and distortion, hence the multiple panels intended to surround the neck, where perhaps a single, more or less rectangular strip would be the choice in a physical cloth tailoring process. Tailors and seamstresses know that there will be a certain amount of warping of the cloth when the garment is completed (perhaps steaming or ironing to de-stress or stretch the cloth later)



  • @F_Verbaas I'm adding this reply since I didn't ever intend my responses to producing a quad mesh within MD as being totally negative. I think it's absolutely terrific that the program can produce a mesh that will most certainly perform better within Poser than a pure tri-mesh, especially with dynamics. Everything else is just nit-picking, in the sense of running through my mental checklist for good Poser figure design and noting that symmetry breaking makes it harder to develop morphs. :-)



  • @anomalaus Well I never UV unwrapped in MD when I've used it. I export to OBJ, and import to Blender, my preference for UV unwrapping. What that looks like is how the pieces were actually created, but FVerbaas would have to specify if that's how the pieces are laid out as he created them.



  • @anomalaus Oh being able to create a Quad-based mesh is a plus, especially since the old MD3 I've used didn't have that ability, and I'm not crazy about converting Tris to Quads in Blender, so I usually import into ZBrush and use ZRemesher, which does a wonderful job of the conversion.

    I happen to be much happier working with Quads than Tris, so adding the ability to export to Quads in MD is a vast improvement.



  • Good to hear you all agree about the quad mesh. That was what the post is about.
    A few notes: some of you commented on the number of panels. The project at hand required close body fit for all but the under-bust area and as little strain as practical. It is not a 'dress' and indeed for actual garment one could merge some parts together and straighten up edges, but at the expense of extra strain. The present split was a compromise. The body, arms and legs are standard kimono cut. The neck and head head are developed for the project.
    About symmetry: If one needs symmetry in a computer model you should build one side only and mirror it. MD does not (yet) offer a 'pin vertex at x=0.0', so I used standard symmetric copy editing instead. Mesh symmetry is not a requirement for my project.
    I will add a few shots later on.



  • As promised.
    Do not look at the chest. I added a lot of crssing lines there I need for my project.

    0_1534440716370_Knipsel.JPG
    0_1534440726893_Knipsel2.JPG


  • Poser Ambassadors

    Having a quad option in MD is good & very useful for some tasks, but I have to say (based on my understanding & experience) that triangulated meshes will perform better than quad meshes when used as dynamic cloth because triangular polygons are always co-planar i.e. all vertices are in the same geometric plane.

    A regular quad mesh is great for something that deforms in specific ways at specific locations like a figure, but cloth has to be highly non-specific - & when you deform a quad mesh against its edge flow some of the polygons will become non-planar resulting in jagged sawtooth artefacts (how noticeable depends on the polygon density), Then the options are to smooth the mesh (losing detail) or to reduce visibility of the problem using subdivision.

    @fverbaas - MD has done a pretty good job for an auto solution, though my preference would be to do some cleanup work to reduce the amount of poles. I don't see any spirals which has been a shortcoming with Zbrush's Zremesher auto retopo function. Does MD maintain an even number of vertices around the boundaries? (Keeping to even numbers is one of the key lessons I’ve learned in modelling.)


  • Poser Ambassadors

    Quads are very good for general modeling but tris are better at dynamics.

    That being said, and these are my personal opinions:

    • As soon as symmetry is lost, the obj file is lost.
    • Always, and I repeat always, and let me underline always once more. . . For most if not ALL 3D object files, try to get as few seams as possible.

    So sorry to say, but both of the above rules are violated by the app, and if I look at the way the cutting-up is done, I am again so sorry to say : Thank you, but no.


  • Poser Ambassadors

    It would be very weird to have symmetrical folds in clothing.
    So I don't have a problem with that the MD OBJ output isn't symmetrical.

    But I do agree for Dynamics Triangle Polygons give a much better flow in the simulation then Quads.



  • @vilters
    I agree it would be good to have a pin to x=0.0 option and a 'save as symmetric copy' option. That should not be too difficult to make by the way.
    But were me, me and, again, me violating your Rule 2 Tony. We conspired and we committed it all three us us. Flagrantly!
    You cannot blame MD for that.
    LOL!



  • @caisson
    I do not know the secrets of MD. The koreans are not very talkative.
    If you want something filled with 4-sided facets, starting with an odd number of vertices would not help.

    I am with you on most other things. quads are much more 'pre-determined' in their bending than tris. I tend to use quads only for elements with a predefined bending like straps and belts. when you model these with tris the suface gets wobbly.