[hugin-ptx] Re: Powermac G5 compiling Hugin from svn. No joy. Possible LAPACK problem

2009-04-17 Thread Harry van der Wolf
Hi Seth,

Sorry.

you are completely right. I read the comment and saw the hash mark and
jumped too fast to conclusions.
I'll have a look again this weekend.

Harry

2009/4/16 Seth Berrier seth.berr...@gmail.com

 Are you sure?

 In C/C++ lines are commented out only with '//' or with '/* ... */'.

 Statements pre-pended with a '#' are preprocessor directives.  You have
 things like:

 #include
 #define
 #ifdef
 #else
 #endif
 #pragma

 and many others, including:

 #undef

 '#undef XXX'  will undefine the preprocessor macro 'XXX' if it is defined (
 http://www.cppreference.com/wiki/preprocessor/undef).

 As it stands right now, lm.h will always undefine 'HAVE_LAPACK'.  I assume
 in the original distribution of LevMar this line looked like this:

 //#undef HAVE_LAPACK // uncomment this to force not using LAPACK

 With the extra '//' at the beginning.  For whatever reason, it has already
 been uncommented in the hugin-levmar version forcing levmar to not use
 LAPACK no matter what.  Aside: we might want to ask why ... this seems
 intentional and perhaps those who understand the math better than we had
 good reasons.

 For fun, I went ahead and removed this on my system and added the line
 '#define HAVE_LAPACK' just below it (just a hack).  The system compiled
 happily (since there are links to LAPACK in /usr on OS X) but I haven't
 tried to run Hugin just yet to see if it really has everything it needs for
 LAPACK.

 Note: I think '#' is used for comments in shell scripts and the like but
 not in C/C++.

 Seth

 P.S.  Sorry for hijacking your thread RickyRicky.  None of this is much
 help for your problem I'm afraid.


 On Thu, Apr 16, 2009 at 9:25 AM, Harry van der Wolf hvdw...@gmail.comwrote:

 Hi Seth,

 thanks for looking into it.

 However, your statement that lm.h undefines it, is not correct.
 the line is:

 #undef HAVE_LAPACK // uncomment this to force not using LAPACK

 As soon as you remove the # it will be an undef statement. Currently it is
 only a remark, so currently it will use lapack provided it can be found.

 Hoi,
 Harry


 2009/4/16 Seth Berrier seth.berr...@gmail.com

 Okay, it looks like LevMar (the Levenburg-Marquardt (sp) library) is
 what's looking for LAPACK.  It uses the preprocesser define 'HAVE_LAPACK' to
 indicate that it is available.

 If you look in 'lm.h' the code specifically UN-defines HAVE_LAPACK so
 even if you were to define it, this line would defeat your define (very odd
 choice).  So, perhaps this might be doing it.  You just need to go into lm.h
 and comment out line 23.

 Of course, this by itself won't fix the problem.  You need to find LAPACK
 on the drive and provide the library paths and header paths to cmake but it
 sounds like you have that going already.  If you can find them then you just
 need to add -DHAVE_LAPACK to the compiler flags for the libhuginlevmar.a
 target and comment out that line in lm.h.

 Hope that helps.
 Seth


 On Thu, Apr 16, 2009 at 8:27 AM, Seth Berrier seth.berr...@gmail.comwrote:

 I'll take a look.  No idea if I can help but it sounds like something I
 might be able to figure out.

 Seth


 On Thu, Apr 16, 2009 at 1:45 AM, Harry van der Wolf 
 hvdw...@gmail.comwrote:

 Lapack is simply not correctly detected in hugin (actually in levmar).
 I dug a little deeper and the routines to correctly detect BLAS and Lapack
 are simply not there. Lapack and BLAS are available via the Accelerate
 framework and via the vecLib framework on MacOSx. To maintain 
 compatibility
 they are even linked to in /usr/lib.
 To test I also built hugin on Ubuntu with lapack (BLAS) installed (libs
 and dev) and also there it is not detected.
 I already tried to implement the FindBLAS.cmake and FindLapack.cmake
 modules to detect it correctly but can't make it work (yet?). I know I'm 
 on
 the correct route but I simply lack the programming skills to make it a
 success. I'm a builder, not a programmer.

 The warnings you get when compiling via cmake is what everyone gets on
 every platform (as far as I can tell now). I can only say that all my 
 builds
 are also always using LU instead of Lapack/BLAS.
 I will try to get it working but if I can't get it to work very soon, I
 will simply file a bug and hope a programmer can take a look.

 Harry



 2009/4/15 RickyRicky meles...@gmail.com


 Hello Harry,

 No, I do not have vigra or jhead installed as a part of MacPorts...

 BigDaddy:Application meleschi$ port installed | grep -i vigra
 BigDaddy:Application meleschi$ port installed | grep -i jhead

 Thanks...  Anything else I can check?

 Ricardo

 On Apr 13, 1:29 pm, Harry van der Wolf hvdw...@gmail.com wrote:
  Hi Ricardo,
 
  Did you install vigra and/or jhead via MacPorts for some reason? If
 you did,
  please deactivate them (sudo port deactivate vigra; sudo port
 deactivate
  jhead) and reconfigure and recompile Hugin.
  If the configure script finds vigra outside  hugin it will use
 that one.
 
  You can find installed packages by running port installed 

[hugin-ptx] Re: Command-line panorama making a Linear pano

2009-04-17 Thread Seb Perez-D

On Fri, Apr 17, 2009 at 15:46, Oskar Sander oskar.san...@gmail.com wrote:
 I ran into the same  problem (evidently) with the max-number of images
 for CP-generation in Hugin as have been discussed here the last few
 days. [...]
 How do you sole these types of CP-problems, or am I alone in getting
 these?

I find that optimizing on all images at the same time is too much for
the optimizer. I do a step-wise approach: I optimize in the beginning
with just a few images, weed out the bad control points, add a few
more connected images, re-optimize. This way, the optimizer only has
to work hard on the images close to the images being added each time.

The latest versions of hugin have a small coloured bar in the image
tabs that indicate how good the control points (in average? in
maximum?) are. I haven't used that yet, but it should give a
indication.

Cheers,

Seb

--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
hugin and other free panoramic software group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
To post to this group, send email to hugin-ptx@googlegroups.com
To unsubscribe from this group, send email to 
hugin-ptx-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/hugin-ptx
-~--~~~~--~~--~--~---



[hugin-ptx] Re: Command-line panorama making a Linear pano

2009-04-17 Thread Tim Nugent

Hi Oskar,

If you have lots of sky/clouds in your pano, you can remove unwanted CPs 
from these areas using celeste:

http://wiki.panotools.org/Using_Celeste_with_hugin

You can use the command line version too which is probably more suitable 
given then number of images you have:

celeste_standalone -i .pto file

Tim

Oskar Sander wrote:
 I ran into the same  problem (evidently) with the max-number of images
 for CP-generation in Hugin as have been discussed here the last few
 days.
 
 Tried to create a project with a line of 180something images which is
 obviously above the limit.
 
 The thing is that the actual panorama is 3 parallel lines of 180
 pictures each.Administrating all these control points and weeding
 out bad ones is quite a task.   What programs can you use for working
 and refining  control points in a project like this, is the best way
 to open it in Hugin when ready and use the control point tools of
 Hugin to try to find bad ones?One thing I found in a smaller Hugin
 project was that I was getting some false control points between some
 images that are not at all connected. It is like I would need some way
 to graphically visualize how the photos are connected to find these.
 How do you sole these types of CP-problems, or am I alone in getting
 these?
 
 The second question I have is if absolute HVFO is at all relevant or
 if I can set it to one arbitrary small number in this case.My
 reasoning: As I am building a linear panorama and Panotools/hugin does
 not support this, I am only ever able to optimize on  d,e  r and HFOV
 and no other movements or distortion parameters.   I am thinking that
 HFOV is only then useful for the optimizer to in effect scale the
 images in relation to each other, the resulting HFOV of the complete
 linear panorama is not relevant as it does not form a sphere or part
 thereof.
 
 
 
 
 Cut from a discussion in December on how to build panoramas from the
 command line, that I'm going to work from
 
 Step 1. I used autopano-sift-c.ext output.pto 1.jpg 2.jpg 3.jpg 4.jpg
 Step 1.5 Manually wash the control points from false and bad points
 step 2. autooptimiser.ext output.pto
 (only on d, e in the first run, and then d, e, HFOV and r in second pass)
 Step 3. PTBatcher to stich
 
 
 Cheers!
 /O
 
  
 

--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
hugin and other free panoramic software group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
To post to this group, send email to hugin-ptx@googlegroups.com
To unsubscribe from this group, send email to 
hugin-ptx-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/hugin-ptx
-~--~~~~--~~--~--~---



[hugin-ptx] Re: Command-line panorama making a Linear pano

2009-04-17 Thread Oskar Sander

Thanks Seb!

How do you do that practically, do you selectively pick the images to
search for control points between somehow?  (How do you do that in
Hugin?)

Do you also uncheck the previously optimized images in the next
optimization run to make sure they stay put, or can Hugin handle it
just doing minor adjustments to them?  I guess tension would build
up in the pano as it grows otherwise?

Cheers

2009/4/17 Seb Perez-D sbprzd+...@gmail.com:

 On Fri, Apr 17, 2009 at 15:46, Oskar Sander oskar.san...@gmail.com wrote:
 I ran into the same  problem (evidently) with the max-number of images
 for CP-generation in Hugin as have been discussed here the last few
 days. [...]
 How do you sole these types of CP-problems, or am I alone in getting
 these?

 I find that optimizing on all images at the same time is too much for
 the optimizer. I do a step-wise approach: I optimize in the beginning
 with just a few images, weed out the bad control points, add a few
 more connected images, re-optimize. This way, the optimizer only has
 to work hard on the images close to the images being added each time.

 The latest versions of hugin have a small coloured bar in the image
 tabs that indicate how good the control points (in average? in
 maximum?) are. I haven't used that yet, but it should give a
 indication.

 Cheers,

 Seb

 




-- 
/O

--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
hugin and other free panoramic software group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
To post to this group, send email to hugin-ptx@googlegroups.com
To unsubscribe from this group, send email to 
hugin-ptx-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/hugin-ptx
-~--~~~~--~~--~--~---



[hugin-ptx] Re: Command-line panorama making a Linear pano

2009-04-17 Thread Oskar Sander


 Do you also uncheck the previously optimized images in the next
 optimization run to make sure they stay put, or can Hugin handle it
 just doing minor adjustments to them?  I guess tension would build
 up in the pano as it grows otherwise?

 No, I keep the previously optimized images as well. Since these
 parameters are already close to the minimum, it does not add much
 tension (although I probably used the optimization based on the
 last known position, or however it is called).


Probably you used the optimize positions incrementally starting from
anchor, right?

Actually before Hugin/panotools can handle linear panorama, would it
be a feasible hack to make an:   optimize positions incrementally
from anchor  for   e, d, r   and   hFOV instead?anyone with
insight in autooptimize, is that heart surgery or piece of cake?

I think it would be useful!

-- 
/O

--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
hugin and other free panoramic software group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
To post to this group, send email to hugin-ptx@googlegroups.com
To unsubscribe from this group, send email to 
hugin-ptx-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/hugin-ptx
-~--~~~~--~~--~--~---



[hugin-ptx] Re: How to install Hugin for G5 Mac?

2009-04-17 Thread Harry van der Wolf
Hi Bob,

With regard to the 0.7 version. That one doesn't work correctly on the G4/G5
in the final stitching stage depending whether you are on Tiger or Leopard.
The early 0.8 versions don't stitch either on G4/G5 depedning whether you
have Leopard or Tiger.
If you pick one of the later 0.8 versions, like the 0.8-rc3, it should work.
Early February I managed to create a patch for the stitching problem (with
help from testers) that works on ppc, both for Leopard and Tiger.
You can find these builds via my website [1].

W.r.t. plug and play and getting you started: Hugin comes in a compressed
.dmg image. Just open it from Finder and drag Hugin to your
Programs/Applications folder. Inside the Hugin applications all neccessary
tools are available apart from the control point generators.
Hugin for OSX uses a plugin structure for the auto control point
generators (for licensing reasons). You can find all plugins also on my
website [1]. The plugins also come as .dmg images. Inside them you will find
installers (and the plugins itself off course). Just double-click the
installer and the plugin will be installed [2].

As Seth already mentioned: If your projects are not too complex and
carefully photographed, you can stay on the Assistant  tab and create
nice pano's from there.

A litte bit of plug and a lot of play

Hoi,
Harry



[1]: 
http://panorama.dyndns.org/index.php?lang=ENsubject=Hugintexttag=Hugin
[2]: For those who are interested: The plugins are installed in your home
directory in ~/Library/Application Support/Hugin/Autopano. Replace
Library with the name for it in your language.

--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
hugin and other free panoramic software group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
To post to this group, send email to hugin-ptx@googlegroups.com
To unsubscribe from this group, send email to 
hugin-ptx-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/hugin-ptx
-~--~~~~--~~--~--~---



[hugin-ptx] Visualize imageGraph of a project?

2009-04-17 Thread Oskar Sander

I think this may be somthing for Pablo to comment on.

I'm browsing the code to understand atooptimize modes (thinking about
the pairwise optimization, but for other parametes than y,p,r).

And came to think whether visualizing or searching the imagegraph
(that is used in the  optimize) is a good idea to verifying
CP-integrity between images in large projects?

-- 
/O

--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
hugin and other free panoramic software group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
To post to this group, send email to hugin-ptx@googlegroups.com
To unsubscribe from this group, send email to 
hugin-ptx-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/hugin-ptx
-~--~~~~--~~--~--~---



[hugin-ptx] Autooptimzer hack for linear panoramas?

2009-04-17 Thread Oskar Sander

I've been browsing the code now for a while, and can't really see that
It would be a problem to make a variant of  Pairwise, which invokes
autoOptimize  with the parameters  r  p and y   in which I would
like to call   d and e  in one case and d e r and v in
another.

Reading autoOptimize in  PTOptimizer.h and PTOptimizer.ccp  I can only
see that it brakes down the panorama problem in a graph and optimizes
each image pair/branch in the graph with the given parameters.

So is my assumption right or am I over-simplifying it?

-- 
/O

--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
hugin and other free panoramic software group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
To post to this group, send email to hugin-ptx@googlegroups.com
To unsubscribe from this group, send email to 
hugin-ptx-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/hugin-ptx
-~--~~~~--~~--~--~---