GSoC - official call for students

[prokoudine]

Hi,

Since we are participating at Google Summer of Code 2008, this is
official call for students. Amount of slots we will be given very much
depends on amount of applications submitted by students.

Yuv has already written an introduction and I don't feel like I want
to duplicate it :)

http://panospace.wordpress.com/2008/03/17/google-summer-of-code-2008/

Please note that before submitting your proposal you need to talk to
us to find a potential mentor and better formulate project
description. We have a little less than two weeks for this.

Alexandre,
GSoC2008 hugin/panotools/enblend primary administrator

[prokoudine]

Here is a quick followup.

We will keep proposals here: http://wiki.panotools.org/SoC_2008_student_proposals

Alexandre

[Michael [Plouj] Ploujnikov]

On Tue, Mar 18, 2008 at 6:50 AM, prokoudine
<alexandre....@gmail.com> wrote:
>
> Hi,
>
> Since we are participating at Google Summer of Code 2008, this is
> official call for students. Amount of slots we will be given very much
> depends on amount of applications submitted by students.

Because I'm currently busy with some course work I'll start with a
vague statement. I am interested in working on freepv to polish it and
make it ready for widespread public use. Over the next few days I'll
be looking at various issues raised in this thread:

http://groups.google.com/group/hugin-ptx/browse_thread/thread/9109462d313a0421/22045f43cc489e94

and making a list of things I think should be done as a GSoC project.

Meanwhile, I would like to know if there are any mentors willing to
work with me on freepv specifically. From the SVN logs I'm guessing
that Bruno Postle is the only currently active developer on freepv and
thus is a good candidate

--
Libre Software:
http://www.gnu.org/philosophy/free-sw.html

[Bruno Postle]

On Tue 18-Mar-2008 at 10:14 -0400, Michael [Plouj] Ploujnikov wrote:
>
>Meanwhile, I would like to know if there are any mentors willing to
>work with me on freepv specifically. From the SVN logs I'm guessing
>that Bruno Postle is the only currently active developer on freepv and
>thus is a good candidate

I'm not responsible for any of the freepv code, I've just been
trying to pull it together into a releaseable state. Though I know
my way around somewhat and could be backup mentor.

FreePV is a good candidate for SoC in terms of bugfixes and
performance enhancements, there is also a VLC player summer of code
idea that would use a fixed-up freepv library:

http://wiki.videolan.org/SoC_2008#QuickTimeVR_Playing

..though I'm not sure two overlapping freepv projects would be
chosen.

--
Bruno

[Pablo d'Angelo]

Bruno Postle wrote:

> On Tue 18-Mar-2008 at 10:14 -0400, Michael [Plouj] Ploujnikov wrote:
>> Meanwhile, I would like to know if there are any mentors willing to
>> work with me on freepv specifically. From the SVN logs I'm guessing
>> that Bruno Postle is the only currently active developer on freepv and
>> thus is a good candidate
>
> I'm not responsible for any of the freepv code, I've just been
> trying to pull it together into a releaseable state. Though I know
> my way around somewhat and could be backup mentor.

While I'd like to be a mentor for this project, I can only mentor a single
project due to time constraints.

I will likely mentor a project working on hugin, so I likely won't be
available as a mentor for a FreePV project.

> FreePV is a good candidate for SoC in terms of bugfixes and
> performance enhancements, there is also a VLC player summer of code
> idea that would use a fixed-up freepv library:
>
> http://wiki.videolan.org/SoC_2008#QuickTimeVR_Playing
>
> ..though I'm not sure two overlapping freepv projects would be
> chosen.

Probably there should be some collaboration, so that at least the
requirements for such an integration are known and code can be written with
this in mind.

ciao
Pablo

[Yuval Levy]

Apology for top-posting.

What would it take to turn FreePV into a library of widgets that can be
used from all different applications?

VLC would be one of them. Hugin, specifically the panorama preview
window, could be another one.

The library would provide widgets to control and display VR for
wxWindows and Qt, with bindings for Python and other scripting languages.

With expert community help, I'd be willing to be primary mentor on this
project and coordinate with VLC. I can put five hours per week into
this, with a few bursts here and there.

Yuv

[Pablo d'Angelo]

Hi Yuv,

Yuval Levy schrieb:

> Apology for top-posting.
>
> What would it take to turn FreePV into a library of widgets that can be
> used from all different applications?
>
> VLC would be one of them. Hugin, specifically the panorama preview
> window, could be another one.

There are some differences between the preview window and a panorama viewer.
The panorama viewer does not have to take into account lens distortion and
the photometric transformations. These complicate matters.

It would be nice to implement this on the GPU, but especially the
photometric transformations will require writing custom shaders, which might
not work on all machines (especially on linux). It could be computed
beforehand, though. Distortion and fisheye projections could be approximated
by a mesh (might be less accurate) or done with shaders, too.

This will basically require rewriting the entire coordinate transform stack
of panotools to produce suitable meshes or shader programs. Actually, parts
of the code could probably be reused, but the student should not be afraid
of math and have a good grip on geometry.

It would be tempting to reuse this for the actual remapping as well, but
then scalability to projects with many images will probably require an
sophisticated approach to handling the limited GPU memory.

> The library would provide widgets to control and display VR for
> wxWindows and Qt, with bindings for Python and other scripting languages.
>
> With expert community help, I'd be willing to be primary mentor on this
> project and coordinate with VLC. I can put five hours per week into
> this, with a few bursts here and there.

I don't want to destroy hopes here, but if the student does not have very
good knowledge of OpenGL and knows a lot about how 3D rendering and panorama
creation is actually coded, at least the primary mentor needs to have this
experience.
Otherwise the student will waste his time with many low level details which
are hard to understand and implement and the success of the project is
questionable.

ciao
Pablo

[Yuval Levy]

Hi Pablo,

Pablo d'Angelo wrote:
> There are some differences between the preview window and a panorama viewer.
> The panorama viewer does not have to take into account lens distortion and
> the photometric transformations. These complicate matters.

the question is: what is a preview and how precise must it be.

You will forgive me the comparison with PTgui - I loaded the same 270+
images of my first gigapixel pano that I stitched with Hugin. The
panorama editor is fast and responsive. I can click and drag the
individual image.

The images are resized - I would estimate to 25% - which is good enough
for a preview IMO.

> It would be nice to implement this on the GPU, but especially the
> photometric transformations will require writing custom shaders, which might
> not work on all machines (especially on linux). It could be computed
> beforehand, though. Distortion and fisheye projections could be approximated
> by a mesh (might be less accurate) or done with shaders, too.

approximation is the key. IMO the primary function of the panorama
preview is to indicate if the last operation yielded an improvement or
if I better click undo and try something else.

In photometric terms, I don't need the seams to be invisible. I only
need a rough idea of the overall color balance that the panorama will have.

speed is more important than precision - it should not be a stitching
process migrated to the GPU.

Yuv