I realized CA exported texture images are usually broken in to pieces(refer to first image) whereas other tools usually output
more in one chunk (refer to second images).
[1] Is there way to export texture more like second image? (Since it is more easy to post process with other tools like photoshop)
[2] What is the general and effective work process to post process 3d models reconstructed by Capturing Reality?
(Since scanned result don’t always give proper result)
If there is a good tutorial, please let me know.
[1] Is there way to export texture more like second image? (Since it is more easy to post process with other tools like photoshop)
NO because UNWRAP is trying to create UV islands without distortion, that is why it cuts the texture into pieces to have all of them without distortion ( curvature of mesh )
Your example shows the standard work that includes custom UVs ( not perfect from the standpoint of distortion-free UVs ) but they are easier to work with. People do this for years…
This could be alternatively solved without custom UVs , but then you need a very good 3D texture-painting solution like Autodesk Mudbox or Mari etc.
The second best workflow is to export it to Zbrush (ZB), clean, retopology, create custom UVs, export it from ZB and import it back to RC for retexturing. In most cases it is done this way as it is a pretty stable solution…
Look for Zbrush scan cleaning tutorials - open YouTube and search 'zbrush scanner scanning workflows"
The second best workflow is to export it to Zbrush (ZB), clean, retopology, create custom UVs, export it from ZB and import it back to RC for retexturing.
I’ve worked this idea through in ZB to the point of reimporting to RC, raised this issue in a separate post, how to control UDIMs in ZB so that RC doesn’t return the error, “A texture coordinate u or v is greater than 1”? Any thoughts on why I’m getting that message and how to avoid it?
I hear your point about the difference between a UV map with lots of tiny islands and presenting no distortions versus clean UVs from a retopologized mesh that’s easier to work with. To what end, though, do clean UVs really count? My understanding is that to support a game engine’s ability to handle lots of geometry and covered with lots of texture detail allowing the user to get up close to anything, that LOD can accomplish only so much. While retopologizing optimizes meshes for rendering, limiting max verts in Parts would seem to bring much in giving a game engine lots of meshes that are small only in relative volume and high in quality with regard to geometry and texture, allowing LOD to go far, possibly rendering the point of retopologizing moot. No? I’m left wondering what’s really gained or lost monkeying with UVs.
At stake is the silhouette of the geometry, distortion of textures, and how lightweight the mesh is. I’d think the transfer detail function in ZB would preserve more detail in the geometry than what’s possible with adaptive decimation in RC, no? Also, don’t we also want to consider what’s possible in leveraging displacement maps or normal maps for providing for finer scale geometry? I’d expect tiny UV islands would preserve much greater detail when baking those maps, but at what cost to polycount/silhouette of the geometry?
I realize there’s not a one size fits all answer to some of these questions. I’m not working with small objects, not with scans of people, am only concerned with hard structures and extensive virtual environments of natural locations.
Thanks for bringing these dynamics into perspective!