Hi Thomas,
Thank you for the reply and thanks for Hugin! I have replied inline.
On 26/06/2024 17:06, 'T. Modes' via hugin and other free panoramic
software wrote:
[email protected] schrieb am Dienstag, 25. Juni 2024 um 20:00:49 UTC+2:
The Python code that I have written can be found here [2].
Sorry, but this unnecessary complicated. I had to dig deep into your
script to found the Hugin commands you use. It would be easier if you
would have only cited your used commands.
My apologies, I'll try to just include the commands next time, I thought
this was maybe some useful context. In the next few weeks/months I also
hope to put the code as a whole (including the ORB control point finder)
online some where in case it could maybe benefit others.
The main problem is that you optimize yaw of all images:
pto_var -o project2 --opt y0,r,y,..
Because you have not other constraints like vertical control point the
optimizer result is ambiguous and the pano can easily slide to left or
right.
I see, that makes sense then. The material that is being photographed
moves around ever so slightly, and because one pixel is about 3.76 μm I
found that I had to also adjust for the roll and yaw. Would there be a
way to tell it to not change/optimise the yaw by more than (say) a few
degrees? That could prove to be quite helpful.
Would you mind elaborating a little bit on what it is that you mean when
you mention vertical control point constraints?
The best advice is to look at the intermediate project files in the GUI
and see the result.
Do you mean to look at the intermediate project files to see what
variables need to be optimised for a good result? This is how I came up
with the current workflow (as well as some online resources) and why I
split the steps up into different "project" files (project.pto,
project1.pto, project2.pto, etc).
PS: Hugin has also possibilities to handle such use cases.
First assign rough image positions according to your layout. Either with
a template (pto_template) or with numbers/expressions (pto_var).
Then run cpfind --prealigned ... on this project file. Then cpfind looks
only at overlapping images. So no need to reinvent the wheel twice.
The reasons I made a custom control pointer finder is that (1) cpfind
wouldn't find control points in cases where I really expected it to find
control points (also with different sieve1/sieve2 sizes) and (2) because
cpfind is relatively slow for the images (151 megapixel times 16) that I
am working with, even if things are prealigned (as I understand it).
After automatically stitching 10000+ images I'm very close to a 100%
success rate, but I couldn't quite get there with just cpfind.
I didn't figure out how to use pto_template make a template with
pre-aligned images. Is there any documentation available on how to do this?
I did manage to run cpfind with the --prealigned option on a project
file where I had already done some control point finding and
optimisating step, but I recall that is was still quite a bit slower -
which is why the code that I provided above falls back to cpfind
(although not with --prealigned flag currently) when it cannot find
control points on its own.
If I could figure out how to use this template option that would help a
lot and then I could also adjust my ORB control point finder to read the
orientation/layout from the project files as well - which would make the
code more generic, and as you say, prevent having to partially reinvent
the wheel.
Regards,
Merlijn
--
A list of frequently asked questions is available at:
http://wiki.panotools.org/Hugin_FAQ
---
You received this message because you are subscribed to the Google Groups "hugin and other free panoramic software" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/hugin-ptx/57b2d6f4-22b2-42ae-8e5a-d084e17856f0%40archive.org.
