control point generator help needed

66 views
Skip to first unread message

Emad ud din Btt

unread,
Dec 25, 2010, 2:18:26 AM12/25/10
to hugi...@googlegroups.com
Dear all,

I need urgent help for a control point generator that can handle 300 images 360x180 architectural panoramas. Please! Mention best possible CP generator and also share it's parameters or steps to follow. 


Emaad
www.flickr.com/emaad

Andres

unread,
Dec 25, 2010, 9:10:33 AM12/25/10
to hugi...@googlegroups.com
If you're in a hurry, you could check out Microsoft ICE (google for links)...

Emad ud din Btt

unread,
Dec 25, 2010, 10:42:21 AM12/25/10
to hugi...@googlegroups.com
Andres I have tried ICE before but i dont think it can handle 300 image. 


On Sat, Dec 25, 2010 at 7:10 PM, Andres <andre...@gmail.com> wrote:
If you're in a hurry, you could check out Microsoft ICE (google for links)...
--
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 hugi...@googlegroups.com
To unsubscribe from this group, send email to hugin-ptx+...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/hugin-ptx



--


Emaad
www.flickr.com/emaad

Yuval Levy

unread,
Dec 25, 2010, 12:24:31 PM12/25/10
to hugi...@googlegroups.com
On December 25, 2010 02:18:26 am Emad ud din Btt wrote:
> I need urgent help for a control point generator that can handle 300 images
> 360x180 architectural panoramas. Please! Mention best possible CP generator
> and also share it's parameters or steps to follow.

if you live in a jurisdiction that is not covered by the SURF patent, you may
want to give Panomatic a try.

To build it [0]

To use it, assuming all your input images are JPG and are in the same folder:

panomatic -o project.pto *.jpg

go have lunch and when you are back open the resulting project.pto file in
Hugin and continue working on it.

Yuv

[0] http://wiki.panotools.org/Hugin_Compiling_Ubuntu#Pan-o-matic

signature.asc

Tom Sharpless

unread,
Dec 25, 2010, 12:38:39 PM12/25/10
to hugin and other free panoramic software
Hi Emad,

I'm not expert on the problem of aligning large sets of images. But I
can tell you the Hugin CP finders will not handle this either. It is
because they try to find CPs between every possible pair of images,
and the number of possibilities grows too large too fast.

The way to do it with APSCP, etc, would be to process separate groups
of images that actually do overlap, then add all the found CPs to a
project file that names all the images (this would involve changing
the image numbers on the CPs to match the ones in the project). For
example if your pano is in rows and columns, each group might be a
central image and 4 neighbors.
I believe PTGui partially automates this, in that you can select some
images and ask it to just find CPs between those. But I'm not sure it
works for more than 2 images at a time, I've never tried more.

There must be CP finders that can handle your problem, because
stitching gigapixel images from hundreds of photos is now quite a
popular activity. I would start by checking the GigaPan website.
Also I would look for a script (perhaps by Bruno Postle) for doing it
with the Hugin tools.

I keep suggesting that someone should make Hugin do CP finding
intelligently, when the arrangement of the source images is already
known. But so far nobody has volunteered.

Regards, Tom

On Dec 25, 10:42 am, Emad ud din Btt <xyzt...@gmail.com> wrote:
> Andres I have tried ICE before but i dont think it can handle 300 image.
>
> On Sat, Dec 25, 2010 at 7:10 PM, Andres <andres.l...@gmail.com> wrote:
> > If you're in a hurry, you could check out Microsoft ICE (google for
> > links)...
> > --
> > 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 hugi...@googlegroups.com
> > To unsubscribe from this group, send email to
> > hugin-ptx+...@googlegroups.com<hugin-ptx%2Bunsu...@googlegroups.com>
> > For more options, visit this group at
> >http://groups.google.com/group/hugin-ptx
>
> --
>
> *Emaad*www.flickr.com/emaad

Emad ud din Btt

unread,
Dec 25, 2010, 1:16:39 PM12/25/10
to hugi...@googlegroups.com
Thanks Yuv, 

 I did try to use panomatic but it failed for such high number of images. I will try Panomatic with parameters you sent. I have no patent restrictions here. 

Thanks.
--


Emaad
www.flickr.com/emaad

Yuval Levy

unread,
Dec 25, 2010, 1:41:35 PM12/25/10
to hugi...@googlegroups.com
On December 25, 2010 12:38:39 pm Tom Sharpless wrote:
> I keep suggesting that someone should make Hugin do CP finding
> intelligently, when the arrangement of the source images is already
> known. But so far nobody has volunteered.

when did you last check Hugin? Thomas Modes has added the multi-row strategy
a while ago...

Yuv

signature.asc

Emad ud din Btt

unread,
Dec 25, 2010, 1:42:25 PM12/25/10
to hugi...@googlegroups.com
Sir,

Many thanks. 

Control point generation is a very important step in panorama creation. Hugin is an excellent software but its not a CP finder. 

PTGUI and Autopano giga both create CPs for this project. But again we cant blame Hugin. CP generation is not its feature.

I am not into gigapixel race. I have to shoot high number of images due to my equipment restrictions. So anyone here please help me out. 


Details of my project are:

5 (360) degree rows
38mm lens
40%-50% overlap between images.
40%-50% overlap between rows. Row 1 overlaps with row2, row2 with row3, row3 with row4, row4 with row5.

Looking for help.

Regards,

Emaad
Emaad
www.flickr.com/emaad

Emad ud din Btt

unread,
Dec 25, 2010, 1:45:28 PM12/25/10
to hugi...@googlegroups.com
I have latest hugin windows build downloaded from thepanz.
--


Emaad
www.flickr.com/emaad

Yuval Levy

unread,
Dec 25, 2010, 7:35:28 PM12/25/10
to hugi...@googlegroups.com
On December 25, 2010 01:45:28 pm Emad ud din Btt wrote:
> I have latest hugin windows build downloaded from thepanz.

you'll find the official ones at
http://sourceforge.net/projects/hugin/files/hugin/

Matthew has been very fast and efficient to produce binaries for Windows, like
Harry for OSX.

Yuv

signature.asc

Emad ud din Btt

unread,
Dec 25, 2010, 9:40:13 PM12/25/10
to hugi...@googlegroups.com
Thanks Yuv, 

I have downloaded official release. Its better. It has all CPGs. I installed it in clean state. Now I have started my testing. I will test all CPGs with my images. At the moment CPfind is doing its job. Any tip for parameters. Like I think adding %s parameter is important for huge number of images. But now I am testing with default parameters I got with official version.

Regards,

Emaad
--


Emaad
www.flickr.com/emaad

Tom Sharpless

unread,
Dec 26, 2010, 10:11:23 AM12/26/10
to hugin and other free panoramic software
Hi Yuv
I have 2010.2.0.d8ce0ba947cc open in front of me. Like most other
recent releases it is unable to generate any control points, and
offers no help to understand why. It does present under preferences/
control point detectors a promising-looking option called "Autopano-
SIFT-C (multrow/stacked)". But of course that is not supported by the
files in the release.

Is that what you are talking about? If so, please state clearly:
-- where do I get it?
-- how do I install it?
-- what do I need to know to run it?

What I was talking about would be Hugin actually controlling CP
finding in an intelligent way; behind an intelligible UI that would
let the user specify a 'shooting grid'; the bottom line being
automatic alignment of gigapixel (and smaller) image sets.

Until Hugin ships with the ablility to auto-align a pano, it will
continue to be ignored by a lot of people who could benefit from it.

Andreas Metzler

unread,
Dec 26, 2010, 1:16:55 PM12/26/10
to hugi...@googlegroups.com
Tom Sharpless <tksha...@gmail.com> wrote:
> On Dec 25, 1:41 pm, Yuval Levy <goo...@levy.ch> wrote:
>> On December 25, 2010 12:38:39 pm Tom Sharpless wrote:

>>> I keep suggesting that someone should make Hugin do CP finding
>>> intelligently, when the arrangement of the source images is already
>>> known.  But so far nobody has volunteered.

>> when did you last check Hugin?  Thomas Modes has added the
>> multi-row strategy a while ago...

> I have 2010.2.0.d8ce0ba947cc open in front of me. Like most other
> recent releases it is unable to generate any control points, and
> offers no help to understand why. It does present under preferences/
> control point detectors a promising-looking option called "Autopano-
> SIFT-C (multrow/stacked)". But of course that is not supported by the
> files in the release.

[...]

Hello Tom,

Upgrading does not automatically make the new control point detector
presets available. You'd need to either add them manually or use the
[reset to default] button.

If you have autopano-sift-c available, choosing Autopano-SIFT-C
(multirow/stacked) as cp-detector instead of regular Autopano-SIFT-C
should give better (i.e. faster) results.

The heuristics used is described on
http://wiki.panotools.org/Hugin_Parameters_for_Control_Point_Detectors_dialog

Afaict there is no way for the user to simply change the way it works
(except manually for selecting a subset of images in which cps are to
be searched). The heuristics assume that images are sorted this way:

1 2 3 4
5 6 7 8
...

If autopano-sift-c is not available to you the setting obvously will
not work for you. Especially in that case you really should upgrade to
2010.4.0 rc1, which *includes* cpfind.

All this does not fulfill ...

> What I was talking about would be Hugin actually controlling CP
> finding in an intelligent way; behind an intelligible UI that would
> let the user specify a 'shooting grid'; the bottom line being
> automatic alignment of gigapixel (and smaller) image sets.

... but should be an improvement over the
compare-all-images-with-each-other-approach.
hth, cu andreas
--
`What a good friend you are to him, Dr. Maturin. His other friends are
so grateful to you.'
`I sew his ears on from time to time, sure'

Jan Martin

unread,
Dec 26, 2010, 1:37:10 PM12/26/10
to hugi...@googlegroups.com
For aerial photography, is there a mode that uses the latitude/longitude info of geotagged images to place them instead of just assuming a pattern like this?


  1   2   3   4
  5   6   7   8
  ...

Jan


--
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 hugi...@googlegroups.com
To unsubscribe from this group, send email to hugin-ptx+...@googlegroups.com

Yuval Levy

unread,
Dec 26, 2010, 1:47:16 PM12/26/10
to hugi...@googlegroups.com
Hi Tom,

On December 26, 2010 10:11:23 am Tom Sharpless wrote:
> I have 2010.2.0.d8ce0ba947cc open in front of me. Like most other
> recent releases it is unable to generate any control points, and
> offers no help to understand why. It does present under preferences/
> control point detectors a promising-looking option called "Autopano-
> SIFT-C (multrow/stacked)". But of course that is not supported by the
> files in the release.

I guess you are on Windows. "not supported by the files in the release" is a
Windows-specific problem. It may have been solved recently with a better
installer that compensates for Window's lack of package management.


> Is that what you are talking about? If so, please state clearly:

Yes

> -- where do I get it?
> -- how do I install it?

The package manager or installer knows where to get third party CP generators
from and how to install them. From the chatter on this ML I believe the
2010.4 Windows installer has been improved, but I can't tell for sure.

As an extra bonus, 2010.4 will also bring you cpfind. That's Hugin's CP
generator. It is still in its infancy, so you may prefer to use your tried
tested and trusted third party CP generator. icpfind [0] enables the use of
heuristic strategies with third party CP generators.


> -- what do I need to know to run it?

The PTO file is the interface that "tells" cpfind/icpfind about images and
lenses. The strategy is described in [1].


> What I was talking about would be Hugin actually controlling CP
> finding in an intelligent way;

that's the above linked strategy.


> behind an intelligible UI

the *I*nterface is the PTO file. And Hugin is the UI to the PTO file.


> that would let the user specify a 'shooting grid';

I see and understand the need to let the user specify a 'shooting grid'.

Underlying such a 'shooting grid' is a 'shooting sequence'. cpfind/icpfind
builds on that sequence and thus does not really need the grid.

Would a 'shooting grid' optimize CP detection beyond what the 'shooting
sequence' already does? Probably yes, but only marginally. icpfind/cpfind
could go straight on to trying to connect all overlapping images.

Where the grid would help even more is with the optimization; and there it is
already taken into account, again, through the PTO file.

So what is missing is the *UI* part. We already have templates (Menu -> File
-> Apply Template) which covers the most important aspect: repeatabilit.

What is missing is a 'Distribute' function/button on the Images Tab, where the
user can specify the shooting pattern in terms of:
- number of rows
- number of columns
- vertical shooting direction (topdown or bottomup)
- horizontal shooting direction (leftright, rightleft, leftsnake, rightsnake)
- number of brackets
and that function would automatically distribute the hundred of pictures on
the panosphere.

Did I miss something?


> the bottom line being automatic alignment of gigapixel (and smaller)
> image sets.

actually the automatism is already given with the sequence-based heuristic,
and the whole UI would be just added manual intervention to make the process
more robust.


> Until Hugin ships with the ablility to auto-align a pano, it will
> continue to be ignored by a lot of people who could benefit from it.

No intention to stir up controversy but this argument is moot.
Yuv


[0] <http://wiki.panotools.org/Icpfind>
[1]
<http://wiki.panotools.org/Hugin_Parameters_for_Control_Point_Detectors_dialog#Multi-
row_panorama>

signature.asc

Yuval Levy

unread,
Dec 26, 2010, 1:53:14 PM12/26/10
to hugi...@googlegroups.com
On December 26, 2010 01:37:10 pm Jan Martin wrote:
> For aerial photography, is there a mode that uses the latitude/longitude
> info of geotagged images to place them instead of just assuming a pattern

no, but you can file a ticket with a feature request.

and: the current heuristic probably does not need the pre-placement for CP
detection, but it may help the optimizer later on.

actually the airplane example illustrate a point where the current heuristic
is better than any pattern assumption. if your plane flies randomly no pre-
filled pattern will help. but the sequence of adjacent images assumed by the
current heuristic still works, unless a gremlin hidden in your computer
renamed the files out of sequence.

Yuv

signature.asc

Emad ud din Btt

unread,
Dec 26, 2010, 3:46:51 PM12/26/10
to hugi...@googlegroups.com
I have tried CPfind. It worked great for two projects. It added CPs for 217 images in 1008 seconds. I used default 10 cp settings. It even added its max 50 cps per image to these 217 images in very short time. But I dont know why its not working like that now for even smaller number of images. It goes on iterating again and again. Than keep displaying strategies and than back to iteration process. My previous experience with autopano sift c is somewhat same. It also keeps iterating even if you leave it for 24 hours. 

Can anyone, explain how CPfind works. I have allowed CPfind to work for hours but it keeps iterating.
--


Emaad
www.flickr.com/emaad

Andres

unread,
Dec 26, 2010, 4:08:22 PM12/26/10
to hugi...@googlegroups.com

Offtopic, but for historical correctness

> i dont think it can handle 300 image.
I once fed about 920 images (png, ~350 kB each, good overlap) to ICE. Both alignment and export took about 12 hours.
 

Regards,
Andres

Tom Sharpless

unread,
Dec 26, 2010, 4:30:05 PM12/26/10
to hugin and other free panoramic software
Thanks Andreas

On Dec 26, 1:16 pm, Andreas Metzler <ametz...@downhill.at.eu.org>
wrote:
> Tom Sharpless <tksharpl...@gmail.com> wrote:
> > On Dec 25, 1:41 pm, Yuval Levy <goo...@levy.ch> wrote:
> >> On December 25, 2010 12:38:39 pm Tom Sharpless wrote:
> >>> I keep suggesting that someone should make Hugin do CP finding
> >>> intelligently, when the arrangement of the source images is already
> >>> known.  But so far nobody has volunteered.
> >> when did you last check Hugin?  Thomas Modes has added the
> >> multi-row strategy a while ago...
> > I have 2010.2.0.d8ce0ba947cc open in front of me.  Like most other
> > recent releases it is unable to generate any control points, and
> > offers no help to understand why.   It does present under preferences/
> > control point detectors a promising-looking option called "Autopano-
> > SIFT-C (multrow/stacked)".  But of course that is not supported by the
> > files in the release.
>
> [...]
>
> Hello Tom,
>
> Upgrading does not automatically make the new control point detector
> presets available. You'd need to either add them manually or use the
> [reset to default] button.
>
> If you have autopano-sift-c available, choosing Autopano-SIFT-C
> (multirow/stacked) as cp-detector instead of regular Autopano-SIFT-C
> should give better (i.e. faster) results.
>
> The heuristics used is described onhttp://wiki.panotools.org/Hugin_Parameters_for_Control_Point_Detector...
>
> Afaict there is no way for the user to simply change the way it works
> (except manually for selecting a subset of images in which cps are to
> be searched). The heuristics assume that images are sorted this way:
>
>    1   2   3   4
>    5   6   7   8
>    ...
>
> If autopano-sift-c is not available to you the setting obvously will
> not work for you. Especially in that case you really should upgrade to
> 2010.4.0 rc1, which *includes* cpfind.
>

Well, I configured the "Autopano-SIFT-C (multirow/stacked)" option to
use the 2-step finder (from a previous build) and set it to work on a
20 x 2 array, shot with a 50mm lens. After 7 minutes 30 sec it
reported 11 unconnected groups. One of those contained the complete
upper row, several others were correctly aligned segments of the lower
row (which is mostly water). Then I tried the "all images at once"
option. It took about 13 minutes but produced a better result: 14
images in one connected group, plus six disconnected images. Under
the same option, the one-step apsc crashed (out of memory) in the
matching stage.

PTGui aligned this set in 2 min 30 sec, leaving 4 disconnected images
in the lower row.

I will repeat once more and then shut up: Hugin's CP finding is a
disgraceful mess, which prevents me -- and I expect many others --
from using this otherwise excellent stitcher. At the very least,
there must be a web page that spells out how to get and use the
various finders now available. I am sick of hearing "oh, that is
patented, so we can't talk about it." Yes You Can.

At best, someone will put together a distributable package that 'just
works'.

Best for 2011, Tom

Yuval Levy

unread,
Dec 26, 2010, 8:21:08 PM12/26/10
to hugi...@googlegroups.com
On December 26, 2010 04:30:05 pm Tom Sharpless wrote:
> Well, I configured the "Autopano-SIFT-C (multirow/stacked)" option to
> use the 2-step finder (from a previous build) and set it to work on a
> 20 x 2 array, shot with a 50mm lens.

and how does 2010.4 with cpfind (not Autopano-SIFT-C) perform on that same set
of 20x2 images?


> I will repeat once more and then shut up: Hugin's CP finding is a
> disgraceful mess

No. You should not judge Hugin by the performance of third party tools. If
you want to complain about Hugin, please try cpfind first. Comments,
discussion, critique, feedback, improvements to cpfind are welcome. And if
you find cpfind to be a "disgraceful mess", we'll take notice.


> At the very least,
> there must be a web page that spells out how to get and use the
> various finders now available.

You (and anybody interested) are welcome to create such a page. The Panotools
wiki is a good place to host it. Hugin has nothing to do with it.


> I am sick of hearing "oh, that is
> patented, so we can't talk about it." Yes You Can.

Patents don't prevent talking. There is a lot of talk around, e.g. in this
very same thread with Panomatic and Microsoft ICE suggested to Emaad as
potential solutions to his CP finding needs. And if these tools do the work
for you, please use them...


> At best, someone will put together a distributable package that 'just
> works'.

... but don't ask others to contravene to SourceForge's terms and conditions
or to take legal risks to satisfy your wishes. Patents do prevent
distribution.

Although, distribution is mainly prevented by technical inferiority. If your
operating system had proper packet management, you would not be complaining
about "bad packages". You're barking up the wrong tree.

Yuv

signature.asc

Emad ud din Btt

unread,
Dec 27, 2010, 4:24:02 AM12/27/10
to hugi...@googlegroups.com
Hi,

Yuv, I have reported CPfind experience in my last email. 

"I have tried CPfind. It worked great for two projects. It added CPs for 217 images in 1008 seconds. I used default 10 cp settings. It even added its max 50 cps per image to these 217 images in very short time. But I dont know why its not working like that now for even smaller number of images. It goes on iterating again and again. Than keep displaying strategies and than back to iteration process. My previous experience with autopano sift c is somewhat same. It also keeps iterating even if you leave it for 24 hours. I have tested 


Can anyone, explain how CPfind works. I have allowed CPfind to work for hours but it keeps iterating.
"
It's good that we can discuss at least one CPgenerator here and i.e. CPfind. Please! Read my email and help me.

My second question is....Control point generation is most important part of panorama creation. So Why we cant work or discuss over it? 

When we talk about PTgui, Autopano or other commercial softwares, we assume its a full package and CP generation is part of it.I am not into comparisons here but you will admit CP generation is essential part of Panorama creation softwares. 

Regards,

Emaad
--


Emaad
www.flickr.com/emaad

kfj

unread,
Dec 27, 2010, 6:27:11 AM12/27/10
to hugin and other free panoramic software


On 27 Dez., 10:24, Emad ud din Btt <xyzt...@gmail.com> wrote:

> My second question is....Control point generation is most important part of
> panorama creation. So Why we cant work or discuss over it?

as you can see, we're discussing the matter at great length.

> When we talk about PTgui, Autopano or other commercial softwares, we assume
> its a full package and CP generation is part of it.I am not into comparisons
> here but you will admit CP generation is essential part of Panorama creation
> softwares.

CPs are - let's call it central - but you might even get away without
them in some projects. You can easily create them manually. But often
- not always - the input lends itself to automatic CP generation and
the CPGs do a good job. It's an area where there is a great wealth of
tools available. See, amongst others:

http://autopano.kolor.com/
http://aorlinsk2.free.fr/panomatic/
http://user.cs.tu-berlin.de/~nowozin/autopano-sift/

CPGs work, but there are strings attached:

- you may have to invest some effort in obtaining, installing and
interfacing to them

- you may have to learn how to use them - most are actually
documented, some of them even quite thoroughly, and especially on
Windows you have a very good selection, but they may not work for you
'out of the box' with standard parameters.

- you may not be allowed to use them commercially, due to software
patent issues

- finally, no CPG can work if the input isn't suitable, so sometimes
failing to get CPs from a CPG throws a light on flaws in your
photography.

Kay

Yuval Levy

unread,
Dec 28, 2010, 12:08:58 PM12/28/10
to hugi...@googlegroups.com
Hi Emaad,

On December 27, 2010 04:24:02 am Emad ud din Btt wrote:
> Yuv, I have reported CPfind experience in my last email.
>
> "I have tried CPfind. It worked great for two projects. It added CPs for
> 217 images in 1008 seconds. I used default 10 cp settings. It even added
> its max 50 cps per image to these 217 images in very short time. But I
> dont know why its not working like that now for even smaller number of
> images. It goes on iterating again and again. Than keep displaying
> strategies and than back to iteration process. My previous experience with
> autopano sift c is somewhat same. It also keeps iterating even if you
> leave it for 24 hours. I have tested

it is impossible to help you / come to a conclusion from what you describe
above. If I understand correctly: cpfind and autopano-sift-c work for the
same large set of images and don't work for a smaller set of images? can you
make the smaller set of images available?

if two cp generators fail the same images set, an issue with the input images
should be considered/tested. you could try other CP generators on the same,
failing set of images.

also: did you try cpfind and autopano-sift-c from the command line? just to
exclude an unrelated GUI freeze/problem, I'd run the tests from the command
line.


> Can anyone, explain how CPfind works.

http://wiki.panotools.org/Cpfind


> Please! Read my email and help me.

It's a two way street. Can't help you without your help.


> My second question is....Control point generation is most important part of
> panorama creation. So Why we cant work or discuss over it?

Is CP generation really the most important part of panorama creation? I don't
think so. Image alignment is certainly crucial, but there are other ways to
align images than by generating CPs (manual or generated) and optimizing CPs.


> When we talk about PTgui, Autopano or other commercial softwares, we assume
> its a full package and CP generation is part of it.I am not into
> comparisons here but you will admit CP generation is essential part of
> Panorama creation softwares.

Panorama creation is a _workflow_ not a _software_.

There is more than one workflow to create panoramas, and while CP generation
is 'essential' to _your_ workflow, it was not to the workflow underlying the
creation of [0]; and it is certainly not essential part of any software.

It is wrong to assume that CP generation _must_ be part of any software just
because there are software that you deem to be 'full packages'. What is the
definition of 'full package' anyway?

Following your logic, you will admit that blending is 'essential part of
Panorama reation softwares'. And yet Enblend is not part of Hugin. Those are
two separate distinct packages. Hugin must redistribute Enblend with the
Windows installer to palliate a major design fault in the Windows operating
system: it does not have a package manager. In systems with package manager,
all Hugin has to do is 'tell' the package manager that it needs Enblend.

Applying the same logic to the CP generation, it would make sense for the
Hugin Windows installer to distribute a CP generator because it can't simply
'tell' the package manager that it needs one. Unfortunately the
(re)distribution of most CP generators is encumbered by patents. Blame it on
the patent holders and on the legal system in which we operate (SourceForge is
located in the United States).

Even if one or many software packages do cover what you think is essential to
your workflow, it is wrong to expect any other software package to do the
same. Your workflow is your responsibility, not the responsibility of the
software package or its authors.

The performance and bugs of third-party tools is not the responsibility of
those developing, building, packaging or distributing Hugin. Feel free to
discuss them, but don't expect them to be part of Hugin, nor for Hugin
developers to take responsibility for them.

Yuv

[0] <http://www.seamlesscity.com/purchase.html>

signature.asc

Emad ud din Btt

unread,
Dec 30, 2010, 4:02:43 AM12/30/10
to hugi...@googlegroups.com
hi yuv,

it is impossible to help you / come to a conclusion from what you describe
above. If I understand correctly: cpfind and autopano-sift-c work for the
same large set of images and don't work for a smaller set of images? can you
make the smaller set of images available?


No, Problem is same for large or smaller set of images. I have attached a pdf file with screenshots of Autopano Sift-c (multirow stacked) cpgeneration process. This screenshot was taken after 16 hours approx. There are 290 images in set.Than I manually interrupted process. It howerver added CPs and thats good about autopano sift c. Cpfind does not single CP if manually stopped.  Autopano Sift-c (multirow stacked) added CPs to set of 217 images. This quite similar to CPfind process. It keeps iterating, strategies and repeating process again n again.   

 
also: did you try cpfind and autopano-sift-c from the command line? just to
exclude an unrelated GUI freeze/problem, I'd run the tests from the command
line


Dont know how to do it from command line. Is it going to folder, placing CPG there and than run it with parameter. Little assistance required so that baby do not loose path :) I am not an expert on command line or scripting. But ready to learn how to walk in this two way street :)

panorama creation is a _workflow_ not a _software_.


There is more than one workflow to create panoramas, and while CP generation
is 'essential' to _your_ workflow, it was not to the workflow underlying the
creation of [0]; and it is certainly not essential part of any software.


Yuv, I here for solution and advice. I am ready to follow your advice. Not here for wasting time on arguments. Can you please tell me what workflow is best for doing it right.

The performance and bugs of third-party tools is not the responsibility of
those developing, building, packaging or distributing Hugin. Feel free to
discuss them, but don't expect them to be part of Hugin, nor for Hugin
developers to take responsibility for them.


Ok thanks for allowing to discuss it. 

I have attached two files. One file is CP generation screenshots. Other picture contains fast preview of one my project. Autopano sift c mutlirow stacked added (30 CPs per image set) and than optimized it with YPRVB parameters. Allignment is not right.
--


Emaad
www.flickr.com/emaad

Autopano Sift-c (multirow stacked) test.pdf
hugin alling.JPG

kfj

unread,
Dec 30, 2010, 11:14:43 AM12/30/10
to hugin and other free panoramic software


On 30 Dez., 10:02, Emad ud din Btt <xyzt...@gmail.com> wrote:

>  hugin alling.JPG

With images like that, you're asking for trouble. Your images are
narrow-angle, and the content is often barely more than blank walls or
featureless surfaces. You can't reasonably expect a CP generator to
find CPs in this situation. If you force it to produce it's best shot
guess at what CPs might be there, it will likely come up with plenty
of false CPs. Then you can't optimize the thing. I recommend you to
buy a fisheye lens if you want to make indoor panoramas like this -
with a wider angle lens you at least have a chance to have enough
overlapping content in successive images.

Kay

Yuval Levy

unread,
Dec 31, 2010, 9:55:28 AM12/31/10
to hugi...@googlegroups.com
Hi Emaad,

We had a major change of plan here because of the weather, so I can only
answer you short.

You were complaining about cpfind and autopano-sift-c hanging indefinitely.
But the screenshot you sent me tells me that the problem lies somewhere else.

Also, I see in your taskbar that there are three Hugin windows open. One is
the fast preview (screenshot). One is the main window. And the third one?

The screenshot you send shows an optimization issue, not a CP generation
issue. Optimization is a garbage-in garbage-out process, and bad CPs may well
be the 'garbage-in'.

Let's try to go at this one step at a time, together. OK?

I want you to start an empty instance of Hugin.

Go into the Images Tab and hit on the "Add individual images" tab.

Add all of your images.

Save this project as step1.pto.

Start a command line in the same folder where you saved step1.pto and issue
the command:

cpfind -o step2.pto step1.pto

Post step1.pto and step2.pto to this mailing list.

Thanks
Yuv

signature.asc

Emad ud din Btt

unread,
Jan 1, 2011, 12:17:42 AM1/1/11
to hugi...@googlegroups.com
Hi Yuv,

We had a major change of plan here because of the weather, so I can only
answer you short.


I can understand. Your short answer is enough for first step. Thanks.

You were complaining about cpfind and autopano-sift-c hanging indefinitely.
But the screenshot you sent me tells me that the problem lies somewhere else.


Yes, They hang or you can say keep working indefinitely. Please! View pdf file I also attached as a reference. I manually stopped APSC process and it however added lots of CPs. Than I optimized it with YPRVB parameters and sent screenshot. So there are two issues indefinite CP generator working and allignment. 

I have started Command line CP generation with CPfind. It's  01-01-2011 and 10:06 A.M. now.

I normally take bracketed images for very bright areas. I have followed your previous advice to put pre Enfused images. Manually fused images with enfuse allign droplet. Enfuse GUI is a good option but it does not have allign feature so I preferred droplet.

I will post files as soon as they are finished.

Thanks,

Emaad
--


Emaad
www.flickr.com/emaad

Emad ud din Btt

unread,
Jan 1, 2011, 2:09:25 AM1/1/11
to hugi...@googlegroups.com
Hi Yuv,

CPfind has added cps to images now. It took 5997.08 seconds = 1.67 hours from command line. I am sending pto files with this email. Please! Check. 

Thanks,

Emaad
--


Emaad
www.flickr.com/emaad

step1.pto.mk
step2.pto
step1.pto

Yuval Levy

unread,
Jan 1, 2011, 2:32:17 PM1/1/11
to hugi...@googlegroups.com
First: I updated the ptoimggen.py script to deal with Windows paths. I love
Python. Attached (for Kay).

Now to Emaad:

On January 1, 2011 02:09:25 am Emad ud din Btt wrote:
> Hi Yuv,
>
> CPfind has added cps to images now. It took 5997.08 seconds = 1.67 hours
> from command line. I am sending pto files with this email. Please! Check.

I started checking. Then I realized that this is a waste of time.

How can you expect such a messy input to result in anything else but garbage?

You need to clean up your files; be more precise in your statements; and learn
to quote properly; before you can expect any help.

FILES:

There are 173 images in the projects you attached. These images belong to at
least two categories, if not three:
* 76 are JPG images that seem to be out of camera JPGs
* 97 are TIFF images that from their file names seem to be the result of some
enfusion
* of the 97 TIFF, some seem to be duplicate (e.g. PC180070_enfused_1.tif and
PC180070_enfused) and some seems to be enfused multiple times (e.g.
PC180139_enfused_enfused.tif)

clean up! either load only one set of enfused images; or one set of JPG
images. If the input is incomplete or garbage, you can't expect anything but
garbage in return. Don't blame it on Hugin, on cpfind, on autopano-SIFT-C, or
on anything that is not between the back of your chair and your keyboard.


CPFIND:

cpfind has "successfully" connected the 173 images with 26007 points. I don't
have the actual images and can't judge the quality of those points, but there
are definitely too many. It is garbage-in garbage out. Before doing anything
you need to clean up the input images mess.


STATEMENTS:

You stated in [0] that "It also keeps iterating even if you leave it for 24
hours.". You stated this in relationship with CPfind and with Autopano-SIFT-
C.

You need to understand that once the CPs are generated, it is no longer CPfind
or Autopano-SIFT-C. It is the Optimizer. And given the garbage fed into it,
I don't blame the Optimizer for not coming up with anything meaningful.

Even if those were 26K real CPs, you would be competing for the largest
gigapixel pano ever (assuming a recent full frame camera and 20% image
overlap). No wonder your computer churns for hours and hours.

This is not the first time that you are looking for help on this mailing list
and make unclear statements about what went wrong [1]. This is not helpful,
nor is it respectful of the time of those trying to help.

Moreover, your quoting is poor and makes it difficult to follow/understand a
thread with you. Please next time quote properly.

I have now wasted about half hour of my time to try to understand you.

If you want a reply from me in the future, you need to improve massively on
your end. You need to clean up your files; be more precise in your
statements; and learn to quote properly; before you can expect any help.

Until then, I will ignore you / your posts.
Yuv


[0] http://groups.google.com/group/hugin-ptx/msg/9fc8f88d2ab211f7
[1] http://groups.google.com/group/hugin-ptx/t/c1f8d1d51e9f455a

ptoimggen.py
signature.asc

kfj

unread,
Jan 2, 2011, 2:33:31 PM1/2/11
to hugin and other free panoramic software


On 1 Jan., 20:32, Yuval Levy <goo...@levy.ch> wrote:

> First:  I updated the ptoimggen.py script to deal with Windows paths.  I love
> Python.  Attached (for Kay).

Yuval, I suppose you want to make sure that, when using the script on
a Unix machine, ptos coming from Windows users will be properly
treated. But it makes the script non-portable. How about a Windows
user wanting to process a pto from a Unix machine?

name = os.path.basename ( line.n.value.replace ('\\', '/') ) +
'_dummy.jpg'

will only work one-way.
I think the best solution might be to use

name = re.sub ( '.*[/\\]' , '' , line.n.value )

which discards everything up to the last slash or backslash. What do
you say?

Kay

Yuval Levy

unread,
Jan 2, 2011, 4:10:40 PM1/2/11
to hugi...@googlegroups.com
On January 2, 2011 02:33:31 pm kfj wrote:
> name = re.sub ( '.*[/\\]' , '' , line.n.value )
>
> which discards everything up to the last slash or backslash. What do
> you say?

that you're the Python King and I have been away from Windows for too long to
care ;-)

well done, Kay.
Yuv

signature.asc

Bruno Postle

unread,
Jan 2, 2011, 6:34:41 PM1/2/11
to hugin and other free panoramic software
On Thu 30-Dec-2010 at 08:14 -0800, kfj wrote:
>
> With images like that, you're asking for trouble. Your images are
> narrow-angle, and the content is often barely more than blank
> walls or featureless surfaces. You can't reasonably expect a CP
> generator to find CPs in this situation.

One way of working around this is to mark the blank areas of walls
temporarily with post-it notes or similar. These help aligning the
shots and can easily be painted out after stitching.

--
Bruno

Emad ud din Btt

unread,
Jan 3, 2011, 3:43:41 AM1/3/11
to hugi...@googlegroups.com, Bruno Postle
Hi Yuv,

Here is my reply

FILES:

There are 173 images in the projects you attached.  These images belong to at
least two categories, if not three:
* 76 are JPG images that seem to be out of camera JPGs
* 97 are TIFF images that from their file names seem to be the result of some
enfusion

Yes, There is set of Tiff images in this set. But they are in sequence of shots and represent bright areas. So enfused them before putting it in Hugin. CPs are accurately added and you can see it in screenshot. By the way it was also your previous advice to clean up mess, manually enfuse image than put into hugin. I dont know why it's creating mess now. 

* of the 97 TIFF, some seem to be duplicate (e.g. PC180070_enfused_1.tif and
PC180070_enfused) and some seems to be enfused multiple times (e.g.
PC180139_enfused_enfused.tif)

Sorry these images were repeated. But this only 4 images. How can this create mess? If these images are identical and all CPs are accurately placed, how can it create mess. Technical answer required. 

clean up! either load only one set of enfused images; or one set of JPG
images. If the input is incomplete or garbage, you can't expect anything but
garbage in return. Don't blame it on Hugin, on cpfind, on autopano-SIFT-C, or
on anything that is not between the back of your chair and your keyboard.


Screenshots are attached and explain me where is garbage. 

CPFIND:

cpfind has "successfully" connected the 173 images with 26007 points. I don't
have the actual images and can't judge the quality of those points, but there
are definitely too many. It is garbage-in garbage out. Before doing anything
you need to clean up the input images mess.


Screenshots are a proof. CPfind did an excellent job. Placing CPs all over images and accurately. No garbage.

STATEMENTS:

You stated in [0] that "It also keeps iterating even if you leave it for 24
hours.". You stated this in relationship with CPfind and with Autopano-SIFT-
C.


I have followed your post and generated CPs from command prompt. I still say that when I tried to create CPs with GUI it kept iterating for 20+ hours. I have replied that CPfind created all these CPs within 1.67 hours. So It must be clear.

You need to understand that once the CPs are generated, it is no longer CPfind
or Autopano-SIFT-C. It is the Optimizer. And given the garbage fed into it,
I don't blame the Optimizer for not coming up with anything meaningful.


Yes, When you use Allign button on Hugin GUI, it adds cps and than optimization process starts. But I created only CPs. By going into images tab. 

Even if those were 26K real CPs, you would be competing for the largest
gigapixel pano ever (assuming a recent full frame camera and 20% image
overlap). No wonder your computer churns for hours and hours.


Command line CP generation has proved that there can be issue using CPFind with GUI. All these CPs are real and accurate. I dont think why 16 cores+16 GB ram is not enough even if it is gigapixel.

This is not the first time that you are looking for help on this mailing list
and make unclear statements about what went wrong [1]. This is not helpful,
nor is it respectful of the time of those trying to help.

Moreover, your quoting is poor and makes it difficult to follow/understand a
thread with you. Please next time quote properly.

I have now wasted about half hour of my time to try to understand you.

If you want a reply from me in the future, you need to improve massively on
your end. You need to clean up your files; be more precise in your
statements; and learn to quote properly; before you can expect any help.

Until then, I will ignore you / your posts.
Yuv


Emotions are everywhere one must be gentle. I have explained myself very well in a fair tone. But language you used is very harsh and insulting. You can have difference of opinion but that does not give you right to use such language on such forums. This is also not first time I have to face such bad language from you. But I said nothing. If you don't like something or do not want to do something than its your right to do so.

You reply to my post or ignore me or you do whatever. But be GENTLE. 

Thank you very much,

Emaad
--


Emaad
www.flickr.com/emaad

1.JPG
2.JPG

kfj

unread,
Jan 3, 2011, 2:00:29 PM1/3/11
to hugin and other free panoramic software


On 2 Jan., 22:10, Yuval Levy <goo...@levy.ch> wrote:

> > name = re.sub ( '.*[/\\]' , '' , line.n.value )
> > What do you say?
>
> that you're the Python King and I have been away from Windows for too long to
> care ;-)

Too much honour. I sympathize with your sentiment about Windows, but
I've used it long enough to feel for it's users.

I've made the change to ptoimggen.py and beautified the script, also
added an option to name a specific output. See

http://bazaar.launchpad.net/%7Ekfj/%2Bjunk/script/annotate/head%3A/main/ptoimggen.py

Kay

Yuval Levy

unread,
Jan 9, 2011, 12:04:48 AM1/9/11
to hugi...@googlegroups.com
Hi Emaad,

On January 3, 2011 03:43:41 am Emad ud din Btt wrote:
> Here is my reply

http://groups.google.com/group/hugin-ptx/msg/6551ccc0aca0f842

can you distinguish between quoted text and reply? I can't. please start by
quoting properly.


> Until then, I will ignore you / your posts.
> Yuv
>
> Emotions are everywhere one must be gentle. I have explained myself very
> well in a fair tone. But language you used is very harsh and insulting. You
> can have difference of opinion but that does not give you right to use such
> language on such forums. This is also not first time I have to face such
> bad language from you. But I said nothing. If you don't like something or
> do not want to do something than its your right to do so.
>
> You reply to my post or ignore me or you do whatever. But be GENTLE.

no intention to offend you, but I also have no time to embellish text. please
don't read emotions where there are no. electronic communication leaves a lot
of void between the lines and readers fill them... with their own imagination
and emotions. a writer can trigger emotions in his readers, but that does not
mean that he feels those emotions. some writers are really good at
deliberately controlling the emotion they trigger. others (like me) are not.
so i have triggered a bad emotion in you, sorry. was not my intention.

afaik nobody has yet invented the emotion-reading email client.

this is not the first time I am trying to give you advice, and it is not the
first time that I get the impression that the advice does not lead to the
expected results, which i believe are the same for you and for me: helping
you make the best out of Hugin. As I stated repeatedly, it is not possible to
help you based on wrong / incomplete statements; and it is not possible to
help you without proper communication / feedback channel.

try again
Yuv

signature.asc

Emad ud din Btt

unread,
Jan 9, 2011, 2:39:19 AM1/9/11
to hugi...@googlegroups.com
Hi Yuv,

Its ok. I also did not want to continue this war of words. 

This group and people working here are very important for me. I have always appreciated your efforts and time for this group. 

I was following your advice. first step was command line CP generation with CPfind and it finished successfully. CPs are accurately placed. But than all this happened. There is mixture of jpgs and tiff files. Jpegs are SOOC
--


Emaad
www.flickr.com/emaad

Dale Beams

unread,
Jan 9, 2011, 8:53:11 AM1/9/11
to Hugin Group
Emaad,
 
E-Mail me privately, or chat with me on IRC and I'll see if I can help you.
 
Dale





 

Date: Sun, 9 Jan 2011 12:39:19 +0500
Subject: Re: [hugin-ptx] Re: control point generator help needed
From: xyz...@gmail.com
To: hugi...@googlegroups.com
Reply all
Reply to author
Forward
0 new messages