Search Issue Tracker
Fixed in 2019.3.X
Fixed in 2020.3.X
Votes
2
Found in
2019.1
2019.3.0a2
Issue ID
1152382
Regression
No
[Denoising] Denoising using OpenImage or Optix is introducing dark borders to UV charts in intensity
Repro:
01 - Downloadload and install the required editor version: -r 7bd0b0ebb78d
02 - Pull the content repo from https://ono.unity3d.com/test-data/gi-tests/ArchVizPRO-Interior-Vol6
03 - Update to content revision: 6f47bfe08488
04 - Generate lighting using the settings stored in the Scene. All this to complete (3-4 mins)
05 - Navigate in the Scene to the upstairs bedroom area shown in the screenshot.
06 - Select the wall mesh shown in the screenshots.
07 -Open the baked lightmap view for this mesh by double clicking lightmap preview in the inspector.
08 - Repeat all this using either Optix or OpenImage denoisers for comparison.
Note:
Around the edges of UV shells in the baked lightmap view there are dark edges with a kind of sawtooth patina. These are seemingly introduced by the denoisers. Padding wont fix the issue. This corruption is different between the denoisers, which is super odd.
Expected:
There should not be dark edges where walls meet in this Scene. We kinda get away with it because it looks like AO, but is actually a bug. Denoised results should ideally be consistent between denoisers.
Add comment
All about bugs
View bugs we have successfully reproduced, and vote for the bugs you want to see fixed most urgently.
Latest issues
- [Render Graph Viewer] Resource list icons cropped and in inconsistent positions
- Scene View with Debug Draw Mode enabled is gray when GPU Resident Drawer is enabled
- The UI is not visible when resizing out of focus built Player window with Application.runInBackground set to false
- [Linux] "Error while reading movie" errors are thrown when opening templates
- Standalone Profiler process is hidden in macOS dock
Bovine
Jan 15, 2020 15:38
If this is won't fix, can we be given a view as to how it will be fixed, i.e. are we waiting for a 3rd party to resolve this?
I am concerned that this is being put down to the way 2 different denoisers work, yet it would seem off that 2 separate denoisers produce the same errors. Occam's Razor might suggest that the error is not outside of unity's control..