Masks yield unusual black / grey / patches unusual results
I have been using Hugin for a decade now - that doesn't really mean anything except I would consider myself fairly familiar with the tricks needed to get around various stitching issues - I guess this is my way of saying "yes, I rebooted and juggled the handle"
Anyway... I have one project, a lighthouse. Nothing spectacular, but it is proving very difficult to get a stack fused version to render out. Here is the LDR image ( from about 45 5-image brackets ) https:/
Per usual I mask off tripod legs and some of the anomalous stuff like the ATV rolling by and so on. Then to eliminate ghosting in the lighthouse, I will exclude the lighthouse from other brackets, forcing Hugin to use the one I want. Normally this works exceptionally well. But the moment I added a mask, even a trivial one, black splotches started appearing near where ever the mask was added. Here are some examples:
https:/
https:/
https:/
My working theory is that at some point it starts dumping out black intermediate images ( the _000xxx.tiffs ) which are then fused. But I have never really see that. I am not sure if this is nona that does the cutout / mask work, or when enblend builds out the exposure layers.
In terms of the setup. its 2018.0.0 + the stock versions of enblend / enfuse running on SuSE Tumbleweed or Leap. Presuming it was an enfuse / enblend bug I have also tried building enblend / enfuse from source ( routinely do this to get a gomp enabled version as I have a lot of cores. This made no difference. ) I see similar results on my laptop as well as my primary rendering server ( HP DL 380, 32 cores, 64GB RAM, heaps of disk ) as well as its predecessor ( 24 cores, 128GB RAM ) .. in watching the panorama build, I see no resource issues. I have also rebuilt the entire project a couple times starting with a blank project and adding the images in, the moment masks gets involved, I start seeing weird stuff.
Is there a reference project used for QA / regression testing by Hugin developers?
This sort of issue reminds me of the required manual memory tuning of nona / enfuse years back, however I don't see where that is a thing any more. The -m options do not seem to be available.
Anyway, sort for the verbal explosion.
Thanks!
-- Bob
Question information
- Language:
- English Edit question
- Status:
- Solved
- For:
- Hugin Edit question
- Assignee:
- No assignee Edit question
- Solved by:
- tmodes
- Solved:
- Last query:
- Last reply: