Improved 100% UV Utilization Regardless of Image Size Input for 4K, 8K or 16K Export
Hi Capturing Reality,
This is a post to discuss how Reality Capture under utilizes UV space.
It seems that Maximal texture resolution has a large impact on how UV space is utilized and it always seems to be under utilized. This also means that users cannot force process textures to 8k or 16k because UV is further underutilized.
Unless the users edits the mesh and UV's outside of RC then re-import and retexture, keeping custom UV's.
It seems underutilization is tied to the input image resolution but this should not force restrict users when wanting to generate large texture maps or get optimal UV mapping.
This is using "Maximal texture resolution" as we only want 1 UV map. Not multiple UV maps. Not multiple UV maps tiled into 1 UV map, as other settings can create.
We want to utilize 100% of the UV space available for 1 large texture map.
Typical settings to export the best consistant (across 4D frames) UV utilization. Even though we require 8K we are forced to 4K.
If we set Maximal texture resolution to 4K (73% utilization)
Still under utilized.
If we set Maximal texture resolution to 8K or 16K (29% utilization?)
Lots of wasted UV space.
So we are forced to work at 4K only to get the best results, which makes sense to match the image input from the cameras but still wasteful.
It's understable that RC does intelligent maths with regards to texel size input from the size of the camera sensor but why under utilize UV space?
If we take another 4D scan frame and generate Maximal texture resolution at 4K, we again get under utilized space, which can also vary from frame to frame over time:
To improve this, we can take the data into Houdini, and repack the UV islands to better utilize the UV space
We can then import this mesh back into Reality Capture and bake the texture at 8K or 16K as we desire.
This adds additional processing steps (mesh import) and processing time to the 4D CLI pipeline.
Capturing Reality please can you introduce a method to ignore image size input from the camera and force utilize 100% of the UV space and allow users to choose which texture size they want to texture map with. Regardless if this is wasteful or if the texture results are slightly blurry from smaller image size input data.
Unless there is another way to adjust settings to improve current UV utilization.
Thanks,
INFINITE
-
hello, this seems to be a similar older topic: https://support.capturingreality.com/hc/en-us/community/posts/115001353472-Unwarp-texture-size-improvements-otherwise-RC-is-a-toy
-
Hello.
I guess so? It was a bit vague in their request post.
"Can you try to set the "RECONSTRUCTION \ Settings \ Advanced \ Use legacy unwrap algorithms" to false and try to unwrap again?
There is a new algorithm that should work better.... - Michal 1 Year Ago"
"You can consider it as experimental. We implemented it but had to swith to another tasks so it is not yet approved. It is needed to be tested properly, the tool needs to pass internal reviews etc. it is a process. But we released it in this form so that we can get a feedback meanwhile. So any comments on the new unwrap are welcome.The difference is that it should produce better utilization.
Have you tried it? Do you like the results? - Michal 1 Year Ago"
What was the conclusion from that? That there's an experimental version that customers should test, that's not approved?
Their results still only seemed to achieve below 80% utilization nor does it address being able to generate 16k textures at will with no restrictions. -
Seeing that it takes years to perfect UV algorithms, wouldn't it be a quick solution to allow baking textures at resolution independent of resolution set in Unwrap settings? This way, we'll just use 4k or 8k for unwrap settings if it results in more optimal UV space utilization, and do actual texture baking at 16k or whatever we need.
-Andrew
Please sign in to leave a comment.
Comments
5 comments