Searching here and You Tube tuts I see others face the same issue with messy scan data misbehaving in ZB when converting to Dynamesh. As I understand it, wall thinness, overlapping faces, and pinched polys are responsible for Swiss cheese and poly islands for shattering. I’ve come across numerous sources advising a workflow that uses similar techniques, extruding thin walls, inflating, smoothing, deleting polys, filling in with clay brush, etc. all highly manual work that would never work for me. While these artist’s are mainly focussed on modeling objects and sometimes limited views of scenes, my work with large (complex) sets and also objects (but different ballgame from others) requires a procedural approach. I’m good with pulling sliders to grow/shrink selections of polys, use a lasso to grab stray islands, but my meshes don’t lend themselves to going in with brushes to tackle the variety of issues plaguing scan data. Here’s what I’ve learned, big thanks for absorbing and questioning with an eye on optimizing for efficiency and results.
Objectives: Scan data is noisy and suffers from an issue related to how point clouds are tessellated to form polygonal meshes. I’m largely fine with how my scan data behaves with natural subject matter, complex cave systems, but I’m now dealing with manmade environments that dish up straight lines, hard edges, homogenous textures, all scanning methodologies sharing problems here but photogrammetry having its own issues using texture information to produce geometry. I’m turning to ZBrush to remove artifacts associated with surface noise while preserving edge detail. The edges themselves need smoothing, but only in the direction of the edge, not to lose the edge detail running perpendicular to that.
Strategy: Convert tris to quads with Dynamesh> generate Polygroups> blur edges of mask per Polygroup (preserve edge detail)> use large SmoothStronger brush to zap all bumps on planar shapes (sometimes curved) to preserve overall form.
Workflow: If I try converting a 20 million tris mesh to Dynamesh (same polycount, Polish = 0, Project enabled), I get Swiss cheese and/or shattering. I’ve successfully produced a Dynamesh by coming down a tad on polycount with Decimation Master (19 M) and then using Geometry> Polish Crisp Edges with low levels of Sharp/RSharp and Smooth/RSmooth. I can then ratchet up iterations of Polish Crisp Edges until I see I’m in the ballpark with something Polygroupit can deal with to break apart the model along so many edges. From there it’s about using masks, which I may have to clean up with the lasso or rectangle, add blur to protect edges, then hit everything inside the field with SmoothStronger brush. Once there, I can use ZRemesher to project detail hi to lo, bake normal maps, etc. (I’m leaving out how I deal with textures.)
Questions: I question if I’m throwing out useful data. Are there tools I’m not aware of that are more appropriate to getting me over the hump producing the Dynamesh w/o Swiss cheese or shattering, this while protecting what’s needed throughout to preserve edge detail? Are there other workflows downstream of producing the first workable Dynamesh that better preserve edge detail, dig deeper to grab more edges, act more global on the mesh? Is there a yet more procedural approach, fewer clicks and drags, to zapping all artifacts, e.g. bumps and depressions, within the planar gross facets of the mesh?
I realize my quest, my very real need, for a highly procedural workflow runs up against the reality of most CG artists, who will argue there’s no getting around highly manual intervention, this defining the CG artist’ role/purpose. I’m encouraged to believe, based simply on my own success thus far, that a highly procedural approach can win the day for using scan data, absolving the artist of the tedious and less creative task dealing with overlapping faces and pinched polys, freeing them to apply their limited resources toward the creative end. Invention of the camera didn’t kill art or the artist. Some painters were freed to present more imaginative subject matter, just as CG artist will focus on fantasy characters and environments. Other artists put down the paint brush and easel, picked up a camera. Art has been, is, and will be on the move.
Big thanks for your insights!
BvC