GSoC2009
prokoudine
We are accepted :) Now we want students!
Here is the list of ideas we've come up with so far:
http://wiki.panotools.org/SoC_2009_idea
Feel free to tell us about your own ideas and don't hesitate to send
us patches to tell us you are serious about participating ;-) I'm
pretty sure that Yuv will write a separate mail with all the fine
details the we he always does as soon as he's back from his short
trip.
Let's rock again!
Alexandre
Yuval Levy
to add mentors to the group and recruit students.
This year we have some additions this year - in case there are students
interested in projecting on large 3D displays.
<http://panospace.wordpress.com/2009/03/19/bliss-and-a-lot-of-work-coming-up/>
prokoudine wrote:
> Feel free to tell us about your own ideas and don't hesitate to send
> us patches to tell us you are serious about participating ;-) I'm
> pretty sure that Yuv will write a separate mail with all the fine
> details the we he always does as soon as he's back from his short
> trip.
tonight I'm quite tired. Baby, wife, parents are all asleep, and it
won't be long until I follow.
> Let's rock again!
This will be another great summer!
Yuv
Leon Moctezuma
Congratulations that Panotools did it again and well I'm interested in taking part of GSoC 2009 and I would like to work again with FreePV, I want to make its integration with VLC this summer. Two months ago I was checking VLC code and I just compiled it from source code... not big deal... but I tried to understand the code and it will be a challenge for me... I think I should apply to both organizations, but probably in this case a mentor familiarized with VLC would be more helpful... anyway I hope to get a slot this summer.
Looking fordward to get coding this summer...
León Moctezuma
Lukáš Jirkovský
I'd really want to participate, but I've not yet decided what should
be my application. think about two things:
1. Making deghosting act more like as a library. This probably would
ease integration to enfuse. The second part would be adding algorithm
described in [1] or adding support for deghosting to enfuse. The
problem of this is that I think I'd not be able to do the math of the
algorithm without help (there are some things I've never seen before).
If I would decide to integrate it to enfuse instead of implementing
mentioned algorihtm the problem would be in fact that both khan and
this algorithm are intended for use with HDR.
Anyway, even if I would choose one of these goals I'd like to
implement also the second one but I think it would be quite
problematic to do the both during summer so the second one wouldn't
have so high priority.
2. Add support for selecting different projections for parts of
images. This would need to calculate some function which would make
transition between parts with different projections smooth. I thing
that it could be done by selecting some discrete points from functions
describing both projections and then describe them by using bezier
curves. The problems are that I don't have any idea how the
projections are described so this approach may be impossible and (it
may be a bigger problem) that I would have to create some GUI for it
(I don't know creating GUI using wxWidgets and overall I hate creating
any GUI).
[1] http://www.ee.oulu.fi/mvg/files/pdf/cpgr_paper-1.pdf
Lukáš Jirkovský
Serge Droz
you asked for it:
Bruno put up a wonderful Tutorial about masking
(http://hugin.sourceforge.net/tutorials/enblend-svg/en.shtml)
The super simplified version of that can be done in hugin by croping images,
i.e. "mask" parts of the borders away.
Getting a simple masking gui into hugin would be awesome.
In any case, thanks a lot for all the work.
It's most appreciated
Serge
Bruno Postle
>
>Getting a simple masking gui into hugin would be awesome.
Ok, I wrote something on the wiki:
http://wiki.panotools.org/SoC_2009_idea#Simple_mask_editing
Everyone else, now is a good time to mention all those good ideas
for hugin features.
--
Bruno
Oskar Sander
I.e. the two suggestions "3D extension of panotools library " or
"Implementing a new projection model for the creation of mosaics in panotools"
There are commercial tools like photoshop for stiching mosaics with
manual "puzzling" however the Hugin workflow for enfusing, emblending
and photometric parameters are so good now, that a photo mosaic
workflow in Hugin would easily surpass the rest. It seems like the
"only" issue beeing the geometric model.
Cheers
O
2009/3/19 Bruno Postle <br...@postle.net>:
/O
Dale Beams
Linear Panos ... ie click, step, click and capture a city street block
Ability to use outside program (kino or kdenlive?) to "capture" shots in a movie export and create linear panos with some sort of automation.
Philospheres (would really like this).
One click transform of panos into tiny worlds.
> Date: Fri, 20 Mar 2009 18:15:57 +0100
> Subject: [hugin-ptx] Re: GSoC2009
> From: oskar....@gmail.com
> To: hugi...@googlegroups.com
dev g
I've been trying to use Hugin to stitch together image mosaics of a
planar object (a large painting) for some months now. Though my
images are heavily overlapped, because each image is taken from a
different location, I am dealing both with the problem of a shift in
the camera viewpoint and using Hugin's current projection model which
isn't designed for mosaics. Over the past two months, I have been
looking into Hugin's source code and finding my way around. I have
made some minor tweaks for my own use such as allowing the use of
optimized values of shear_x and shear_y, (a.k.a. g and t.) I have
compiled and built Hugin both on Windows and on Linux under Ubuntu.
I'd like to solve the problems I'm encountering as part of Summer of
Code. Specifically, I'd like to introduce a mosaic mode to Hugin.
This will require the development of the geometric framework to
project from the mosaic image viewpoints to that of an orthographic
view of the entire painting. I propose calculating the warp using a
homography, and applying the calculated mapping and projecting the
result to the orthographic viewpoint. A homography estimation
algorithm developed in C by Lourakis will provide the required
homography matrix calculations.
I'm a PhD candidate in Electrical Engineering and Computer Science at
Northwestern University, Evanston, Illinois working to develop a
software that can be used by art historians to study paintings under a
wide range of lighting conditions. I hope Hugin will become a vital
tool for pre-processing, warping, and stitching images for my project.
I plan to upload my full proposal to the wiki soon and look forward to
your feedback. I would especially be interested in hearing from
potential mentors. From the earlier messages, it's clear that there
is great demand for such a feature and I would enjoy working with the
Hugin community to make it better for all of us.
Best,
Dev
Dev Ghosh
I've been trying to use Hugin to stitch together image mosaics of a planar object (a painting) for some months now. My images are heavily overlapped, but because each image is taken from a different location, I am dealing both with the problem of a shift in the camera viewpoint and using Hugin's current projection model which isn't designed for mosaics. Over the past two months, I have been looking into Hugin's source code and finding my way around. I have made some minor tweaks for my own use such as allowing the use of optimized values of shear_x and shear_y, (a.k.a. g and t.) I have compiled and built Hugin both on Windows and on Linux under Ubuntu.
I'd like to solve the problems I'm encountering as part of Summer of Code. Specifically, I'd like to introduce a mosaic mode to Hugin. This will require the development of the geometric framework to project from the mosaic image viewpoints to that of an orthographic view of the entire painting. I propose calculating the warp using a homography, applying the calculated mapping, and projecting the result to the orthographic viewpoint. A homography estimation algorithm developed in C by Lourakis will provide the required homography matrix calculations.
I'm a PhD candidate in Electrical Engineering and Computer Science at Northwestern University, Evanston, Illinois working to develop software that can be used by art historians to study paintings photographed under a wide range of lighting conditions. I hope Hugin will become a vital tool for pre-processing, warping, and stitching images for my project.
I plan to upload my full proposal to the wiki soon and look forward to your feedback. I would especially be interested in hearing from potential mentors. From the earlier messages, it's clear that there is great demand for such a feature and I would enjoy working with the Hugin community to make it better for all of us.
Best,
Dev
dmg
> Specifically, I'd like to introduce a mosaic mode to Hugin. This will
> require the development of the geometric framework to project from the
> mosaic image viewpoints to that of an orthographic view of the entire
> painting. I propose calculating the warp using a homography, applying the
> calculated mapping, and projecting the result to the orthographic
> viewpoint. A homography estimation algorithm developed in C by Lourakis
> will provide the required homography matrix calculations.
I have added this to the list of projects. this should be done
primarily in panotools.
I think that it is highly related to allowing shift in the camera's location,
so those two projects should be combined.
I am probably the potential mentor for this (panotools particularly).
--
--dmg
---
Daniel M. German
http://turingmachine.org
Luca N Vascon
The first problem is that when somebody wants to install it, even on
windows, he steps on old unupdated versions, when luckyly find the
compiled exes.
Ubuntu debian self installing thing seems old.
It completely lacks clear instructions on how to install what.
Never seen the new updated interface. Does it exist?
How is integration of old things?
Please, the first act should be reorganization of knowledge and concepts.
It is 3 years I'm lost to hugin, I've no clue on how to do things out of
basic ones cause I can't access knowledge base on my own...
I hope I'm not the only one, cause means my stupidity is growing with age.
dmg ha scritto:
--
IUAV università degli studi, Venezia
Dr. Luca Vascon
tel . (+39) 041.2571262, e-mail vas...@iuav.it
laboratorio multimedia Facoltà di Design e Arti, DADI
Magazzino 7 ex Ligabue, Dorsoduro 1827 30123 Venezia
Dale Beams
I build hugin svn regularly and post them to http://tatteredmoons.org/hugin_svn.html with the dependencies needed and instructions to install them. I've also began testing them before posting any new debs. Currently the debs listed are only for the 8.04.1 release, as it's the latest stable long term release. I do have 8.10 debs, but have not had a chance to seperate the web pages this week into two seperate versions. I only build i386 because of comaptibilty.
I also make an effort to look over the wiki (ubuntu) and make corrections when needed. I too have been considering restructing that part of the wiki to make it clearer to understand to the average user (me).
I've been considering writing a manual in scribus now that I know how to use scribus, and would love your help. Help can be questions that may seem "non smart", because I do understand a user's prespective and understand there are no "dumb" questions. You too can download and use scribus and I'd be willing to work with you as a collabartive effort if you choose.
I do believe that hugin needs to move foward, and am happy to see what's being listed for GSOC.
Drop me a note privately if you want help. I'm only a user, but I'm sure I can help you through it.
Dale
> Date: Sun, 22 Mar 2009 17:07:18 +0100
> From: vas...@iuav.it
> To: hugi...@googlegroups.com
> Subject: [hugin-ptx] Re: GSoC2009
RizThon
> Date: Sun, 22 Mar 2009 17:07:18 +0100
> From: vas...@iuav.it> Subject: [hugin-ptx] Re: GSoC2009
>
>> Please, I think the BASIC thing hugin needs is an updated website.
> The first problem is that when somebody wants to install it, even on
> windows, he steps on old unupdated versions, when luckyly find the
> compiled exes.
> Ubuntu debian self installing thing seems old.
> It completely lacks clear instructions on how to install what.
> Never seen the new updated interface. Does it exist?
> How is integration of old things?
>
> Please, the first act should be reorganization of knowledge and concepts.
> It is 3 years I'm lost to hugin, I've no clue on how to do things out of
> basic ones cause I can't access knowledge base on my own...
> I hope I'm not the only one, cause means my stupidity is growing with age.
If there are things you don't know or understand how to do, just ask! That's how we can make things even better, and there are probably other people who are just like you, a little bit lost!
Adrien
Hello Bruno,
When you say that "masking to put one object in-front of another is
never going to be done in hugin, this is a job for an image
editor" ... I'm thinking that this would actually be a great feature
to be included in hugin, because it would allow to do everything
(quickly) in pre-processing instead of post-processing, and still have
the benefit of the blending step.
Just to make sure we are talking about the same thing, I understand
this "mask editing" topic is all (and only) about what should be
included or not included in image seams.
Now maybe I'm missing something, but if the steps performed once you
press the "stitch now" button are:
- remapping images
- remapping masks
- applying masks to remapped images
- blending masked remapped images
.. then for the two cases you mentioned:
1. masking out objects that you don't want to appear in the scene.
=> the mask is applied to one image only (the one from which you want
to mask out the objects)
2. masking to put one object in-front of another.
=> the mask is applied to all images but one (the one from which you
want to keep the objects)
Cheers,
Adrien
Adrien
Hello Bruno,
When you say that "masking to put one object in-front of another is
never going to be done in hugin, this is a job for an image
editor" ... I'm thinking that this would actually be a great feature
to be included in hugin, because it would allow to do everything
(quickly) in pre-processing instead of post-processing, and still have
the benefit of the blending step.
Just to make sure we are talking about the same thing, I understand
this topic is all about what should be included or not included in
image seams.
Now maybe I'm missing something, but if the steps performed once you
press the "stitch now" button are :
- remapping images
- remapping masks
- applying masks to remapped images
- blending remapped images
... then for the two cases you mentioned :
1. masking out objects that you don't want to appear in the scene.
=> the mask is applied to one image only (the one from which you want
to mask out the objects)
2. masking to put one object in-front of another.
=> the mask is applied to all images but one (the one from which you
want to keep the objects)
Cheers,
Adrien
On Mar 19, 8:13 pm, Bruno Postle <br...@postle.net> wrote:
Irena
Hi guys!
I'm a Msc CS student in the inter disciplinary center in Herzliya,
Israel.
My main interests are image processing, 2nd graphics and computational
photography.
from the short description in the GSoC site about this project it
seems that I will be able to contribute to it, and would actually be
happy to do so :)
unfortunately I cannot see the wiki of the project since it has been
down for almost all day long, and cannot see what are the suggested
projects Ideas.
are the projects that you are looking for are mainly environment (like
phyton bindings etc,) and gui, or are you interested in students to
implement all kinds of advanced computational algorithms?
Thank you!
Irena
Bruno Postle
>
>Please, I think the BASIC thing hugin needs is an updated website.
>The first problem is that when somebody wants to install it, even on
>windows, he steps on old unupdated versions, when luckyly find the
>compiled exes.
>Ubuntu debian self installing thing seems old.
The ubuntu instructions are really for people who _want_ to compile
their own. The way it should work for everyone else is that the
distribution provides the binaries - i.e. hugin is part of ubuntu
and you don't need to download anything from the hugin website.
This situation isn't how it should be, we need more regular releases
of hugin and related tools, and Linux distributions should be
prompted to fix the stuff they do supply.
--
Bruno
Bruno Postle
>
>When you say that "masking to put one object in-front of another is
>never going to be done in hugin, this is a job for an image
>editor" ... I'm thinking that this would actually be a great feature
>to be included in hugin, because it would allow to do everything
>(quickly) in pre-processing instead of post-processing, and still have
>the benefit of the blending step.
>
>Just to make sure we are talking about the same thing, I understand
>this "mask editing" topic is all (and only) about what should be
>included or not included in image seams.
>
>Now maybe I'm missing something, but if the steps performed once you
>press the "stitch now" button are:
>- remapping images
>- remapping masks
>- applying masks to remapped images
>- blending masked remapped images
Currently masks are applied in three possible ways:
1. the Crop tab lets you exclude edge areas, these pixels are simply
not sampled when remapping.
2. alpha channels in source photos are remapped in exactly the same
way as RGB colour channels, which leads to:
3. 100% transparent alpha channels in remapped images are treated as
hard image boundaries by enblend.
Note that enblend merges images by taking the entire overlap area
and feathering it (in a clever multiresolution way). This only
works when the overlap has a significant width and the two images
are similar.
>.. then for the two cases you mentioned:
>1. masking out objects that you don't want to appear in the scene.
>=> the mask is applied to one image only (the one from which you want
>to mask out the objects)
>2. masking to put one object in-front of another.
>=> the mask is applied to all images but one (the one from which you
>want to keep the objects)
Yes this works for forcing parts of images into the output, see the
'process-masks' tool in Panotools::Script for a way of doing this
with the current enblend and three different levels of alpha channel.
The problem with putting one object in front of another is that you
need an abrupt transition, the 'feather' between pictures is only a
pixel or two wide, not something enblend can do.
--
Bruno
Pablo d'Angelo
Lukáš Jirkovský wrote:
> Hello,
> I'd really want to participate, but I've not yet decided what should
> be my application. think about two things:
>
> 1. Making deghosting act more like as a library. This probably would
> ease integration to enfuse. The second part would be adding algorithm
> described in [1] or adding support for deghosting to enfuse. The
> problem of this is that I think I'd not be able to do the math of the
> algorithm without help (there are some things I've never seen before).
While I won't be able to be a full time mentor this year, I'd be more
than happy to answer your questions about the algorithm in [1]. Don't
let the formulas scare you. From a quick look it seems to work quite
similar to the khan deghosting (whose paper also sounds more complicated
than it really is), but contains additional steps to further improve the
results.
> If I would decide to integrate it to enfuse instead of implementing
> mentioned algorihtm the problem would be in fact that both khan and
> this algorithm are intended for use with HDR.
Actually, both algorithms produce a weight values for each pixel in each
exposure. This weight value could be combined with the weight enfuse
computes for each pixel.
> Anyway, even if I would choose one of these goals I'd like to
> implement also the second one but I think it would be quite
> problematic to do the both during summer so the second one wouldn't
> have so high priority.
I don't think doing both is possible in a summer, especially when time
for learning how the algorithms really work is also needed.
ciao
Pablo
Pablo d'Angelo
Dev Ghosh schrieb:
> Hello,
>
> I'd like to solve the problems I'm encountering as part of Summer of
> Code. Specifically, I'd like to introduce a mosaic mode to Hugin. This
> will require the development of the geometric framework to project from
> the mosaic image viewpoints to that of an orthographic view of the
> entire painting. I propose calculating the warp using a homography,
> applying the calculated mapping, and projecting the result to the
> orthographic viewpoint. A homography estimation algorithm developed in
> C by Lourakis will provide the required homography matrix calculations.
While the homest package by Lourakis is a nice piece of work, I think
that it would be better to stick with the physical model that panotools
uses, as that also handle lens distortions properly and would allow
using fisheye and cylindrical cameras, too. This would mean that one
"just" needs to implement the right transformation functions, and reuse
the exiting optimisation algorithms.
The only problem is that all current panotools transformations have not
been build with moving viewpoints in mind, so the panotools projections
code would probably need some reorganisation/rewrite to allow this cleanly.
Having a mode for properly mosaicing planar scenes would be an extremly
welcome addition to hugin.
ciao
Pablo
Yuval Levy
Irena wrote:
> unfortunately I cannot see the wiki of the project since it has been
> down for almost all day long, and cannot see what are the suggested
> projects Ideas.
it is currently quite slow here too, but it loads
<http://wiki.panotools.org/SoC_2009_idea>
> are the projects that you are looking for are mainly environment (like
> phyton bindings etc,) and gui, or are you interested in students to
> implement all kinds of advanced computational algorithms?
both. it depends on your interests and skill.
For a starter, may I suggest that you go about building and hacking
hugin and related tools?
<http://wiki.panotools.org/Development_of_Open_Source_tools>
Yuv
Yuv
Dev Ghosh wrote:
> Specifically, I'd like to introduce a mosaic mode to Hugin.
Hugin?
> I plan to upload my full proposal to the wiki soon and look forward to your
> feedback.
codebase from you.
Yuv
Yuv
Lukáš Jirkovský wrote:
> I'd really want to participate
very valid patches and fixes and I look forward for you to work full
time on Hugin for a while.
> 1. Making deghosting act more like as a library.
you have talent and motivation to pick up this project. Don't worry
about picking up a difficult project, the choice will be accounted for
in the evaluation.
Next step for you would be to add yourself to the wiki: <http://
wiki.panotools.org/SoC_2009_student_proposals> and to apply in the
GSoC web based tool.
<http://socghop.appspot.com/>
The application is new so I have no clue how it looks from the student
side. I'd be grateful for feedback. I think you need to create
yourself a profile at
<http://socghop.appspot.com/user/edit_profile>
and then apply for hugin/panotools, but I don't know how you would
apply. If you do it, please share with us here your experience, it
will help the other students who are less advanced in the process than
you.
good luck
Yuv
Oskar Sander
testing this.
/O
2009/3/20 dmg <d...@uvic.ca>:
/O
Irena Damsky
I saw something on the mailing lists that is kinda interesting to me...
in the thread "adding missing sky" it is asked if you can give a simple solution to completing missing data to the regions which are not actually there from the input pictures.
As I've said on that thread, there are some scientific articles that deal with that problem i.e. there are some articles that offer to do that by using huge amount of graphical DB to complete the missing data from other images. OR there is the article that uses Poisson image editing to manipulate the data in such manner that almost no information can create huge amounts of info that will make the image look good.
I actually know that implementing the Poisson completion can be fun and it would be great to see it as part of something that people actually use...
I wonder, does something like that - image "completion" could work for a GSoC project with you guys?
Thanks,
Irena
Irena Damsky
irena....@gmail.com
057-8165528
052-3294417
you can also find me on MSN: ira_d...@hotmail.com
Smile! and the world will smile with you!
Cry! and you'll cry alone...
Irena Damsky
just saw that you replied on this in the other thread, and just wanted to say sorry that I've doubled posted...
(hmmm.. now I'm just spamming again ha? BUT I really wanted to say I'm sorry, please ignore me and I'll go sit alone in the dark and cry)
Irena
Yuval Levy
> (hmmm.. now I'm just spamming again ha? BUT I really wanted to say I'm
> sorry, please ignore me and I'll go sit alone in the dark and cry)
no need to be apologetic. We do not ignore potential contributors. We
also don't harass them. I won't grab you out of the dark corner you're
putting yourself in. I'll wait patiently for you to come out of there
and will welcome you to our community if you want to join.
There are plenty of interesting projects and capable mentors waiting for
you. But the application deadline is April 3. And you need a working set
of tools, as described in the wiki.
Don't make your life hard. Grab a simple Ubuntu or Windows, set yourself
up with the tools and let the fun begin.
Yuv