click here to view Google Summer of Code group panorama

I spoke about our participation and the benefits for the VR community with Leslie Hawthorn, Program Manager - Open Source at Google; Pablo d'Angelo, Lead Engineer for hugin; and Ken Turkowski, VR scientist and member of our steering committee.

click here to view panorama with Yuval Levi, Pablo d'Angelo, and Ken Turkowski

Yuval: Hi Leslie! On behalf of the hugin/panotools team, thank you for having us in the GSoC.
Leslie: Hi Yuv, it's really our pleasure.

Can you tell briefly our readers about the GSoC?
GSoC is a Google initiative to support the Open Source community, and it's our third year doing it. We pay students to write code as a summer job. They can write whatever software they want. We only ask them to do so within a mentoring framework and to release the software under an Open Source license.

Sounds too good to be true. Where is the catch?
No catch. Google gets the same benefit as the rest of the world - more useful code. We think the program is an effective way to give something back to the community.

Why not simply donate the money to the projects?
We get better overall value this way. We give students in need an opportunity for a paycheck, but also for a rewarding experience. We give Open Source communities an opportunity to get things done, but also to attract fresh talent. The goal is for the students to join Open Source development permanently, the paycheck is just a little temporary incentive. We create synergies and we also learn about software development.

So how many projects do you work with?
This year we worked with 130 mentoring organizations, up from 100 last year. We try to fit in as many as we can, but we could not possibly fit all the 400+ that applied this year.

How do you choose which organizations to sponsor?
We don't sponsor organizations, we sponsor students. We select organizations based on their ability to give our students adequate mentoring and challenging projects to match the student's interests.

And how do you choose which students to sponsor?
We allocate a number of slots to each mentoring organization based on its popularity among student applicants. Mentoring organizations rank their applicants based on their merits. Combine those two processes, add a few sleepless nights of manual tweaking and mentors begging for more students and it yields a list of the lucky 900+ students accepted out of thousands of applicants. It would be impossible to review 6000+ applications ourselves in such a short time span.

How was this year's GSoC?
Terrific! More projects, more organizations, more code, more efficiency. Everything got better.

What was the best moment for you of GSoC so far?
There have been many highlights, though what I found particularly awesome, both this year and last were the face to face meetings here at the Mentor Summit.

And what was the worst moment for you?
Hearing the story from one of our mentors about his student who cheated. Long story, needless to say really disappointing. It's okay to fail. It's not okay to cheat.

How will you remember the hugin/panotools participation to GSoC 2007?
Hrm. As the project that got us the high resolution group picture we have been missing for the past two editions? Oh yes, and "sushi."

(sushi is a reference to a session I led at the Mentor Summit and dealing with cross cultural communication and management in an Open Source environment. The full title was “The Stick, The Carrot, The Sushi and The Cutlery” and the story is coming soon on Yuval’s blog

I know you can't make promises about next year, but should Google decide to run this initiative again in 2008, what are our chances to participate again?
We like to work with organizations who have a good track records. Mentoring organizations that performed well this year are likely to be admitted next year too.

Thank you very much, Leslie, for allocating five students to us this year and for this interview. I hope to see you next year, same time, same place.
Thank you and the hugin/panotools team for taking care of our students, Yuv.

When Pablo d'Angelo, initiator of hugin and FreePV, saw the invitation for mentoring organizations to apply for GSoC, he launched the idea in our community. I picked up the ball and we started rolling. While Pablo was still completing his PhD, I set up the organization. Together, the community of developers around hugin/panotools went on to obtain mentoring organization status, a privilege that we thought was reserved to much larger organizations with a longer track record, such as the Mozilla Foundation, maker of the Firefox web browser, or Drupal, make of the CMS that powers many websites worldwide including the IVRPA's website.

Hi Pablo, how do you like it to be at Google's headquarters for the Mentor Summit?
It is great to finally meet face to face people with shared interest. We are very thankful to Google for this unique opportunity.

So tell me, a little bit about hugin. How did you start it?
When I started hugin, PTStitcher, the core application behind panotools was not available in source code, so I started developed my own stitching engine, nona. It now includes some powerful features that are not available in PTStitcher.

Can you tell us the origin of the name hugin?
It's a raven in Norvegian mythology. http://en.wikipedia.org/wiki/Hugin_and_Munin

What makes hugin unique?
First of all, it is OpenSource, so everybody can help improve it, or adapt it to his needs. It has been designed with multiple platforms in mind and runs on Linux, Mac and Windows. Then it has an unbeatable price tag. Humor aside: its one of the first tools to introduce a powerful and flexible method to deal with color and exposure correction.

Can you elaborate?
Most stitching tools correct scientifically the perspective for lens distortion but consider blending as simply “smoothing the differences at the seams so that they are no longer noticeable”. Hugin takes a different approach. It uses a physically plausible model for the image formation, basically it knows how the color values in the image are created.

What does this mean for the average user?
Vignetting is corrected automatically, blue skys (especially in panos with assembled from many images) will be more uniform as they should be. Additionally hugin can creates very nice panos from images shot with different white balance and exposure, and it can be used to calculate and remove some forms of chromatic aberrations.

And what about high dynamic range?
HDR Panorama shooting has never been easier. Hugin can directly generate HDR panoramas directly from the source images and there is no need to bracket each shot, as long as the input pictures capture the relevant dynamic range, hugin will seamlessly fit it in the overall stitched image.

So what impact did GSoCTM have on hugin?
A very positive impact! We had three sponsored students working on different parts of the code. GSoC participation also generated visibility and buzz. I was invited to present hugin at Libre Graphics Meeting and gave a presentation at the European panotools user community meeting. We made contacts with very interesting people and expanded collaboration with the Open Desktop community, for example on a forthcoming lens correction database that might be an interesting project for a future GSoC edition.

In what areas where the three students active?
Ippei Ukai has completely refactored the hugin code base into a new GUI framework. Jing Jin has written a deghosting tool for the processing of HDR stacks and Zoran Mesec has written a new control points generator.

Did hugin really need a new control points generator?
The SIFT algorithm used by the current control point generators is subject to a US patent that might restrict use in some jurisdiction. As a consequence it is not distributed with Fedora Linux. For some type of images, notably fish eye lenses, the matching results could be better. We wanted a clean, free Open Source solution. There is always room for improvement and it was an interesting challenge for the student concerned. A few weeks ago another person came up in our mailing list with another solution to the problem, there is room for variety.

And what does the deghosting tool do?
In HDR photography subjects in the bracketed exposures are seldom 100% identical. Subject movement produce an effect called ghosting. The deghosting tool isolate this effect by determine which of the stacked bracketed exposure is most relevant and discarding the information from the rest.

And what is the new GUI framework about?
We moved the functionality to Qt, a more flexible GUI toolkit, and cleaned up the code. The refactored code will make it easier for other contributors to add both functionality and interactivity.

Many people in the VR community are waiting for the next hugin release?
The reality of hugin is that of many Open Source projects: we have limited resources and time. I can't promise a release date. If people want the latest and best hugin, they can check out the code from the SVN repository and build it.

Our audience are photographers, I am afraid it is not so easy for them as it is for regular Open Source contributors...
We have a friendly mailing list at http://groups.google.com/group/hugin-ptx - if somebody hits a road block when building they can get help there. There are fairly detailed instructions for building on Linux, Mac and even for Windows. We are currently stepping up efforts to set up a build system to make regular snapshots available to test users on all platforms. But I must emphasize that we are looking for users willing to give something back to the community.

Most hugin users are no coder. How can they give back to the community?
No need to be a coder to give back to the community. There are plenty of ways in which they can help us. Testing is one area. We are also looking for people who can help to extend the manual, provide ideas for further design, especially related to the GUI, artists to draw better looking toolbar icons. Supporting new user in their effort of adopting hugin.

What was the best moment in this GSoCTM participation for you?
When I found out that we were accepted as a mentoring organization.

and the worst moment?
When we had to fail the one student in our team that was not perform it, despite an incredible round the clock effort of five mentors.

and how did the other students do?
They all did very well. Jing's deghosting tool is material for a scientific paper, though I do not know if we will write it. She presented her results at the IVRPA conference in Berkeley. Ippei was with me at the Panotools meeting in Lucerne and is maintaining the OSX port of hugin, although he has little time now that university started again. Zoran still has some performance issues with his code, but he keeps working on it and is hanging on the community mailing list, so overall I think we did very well with these students.

So when can users try Jing's deghosting in hugin?
We have planned two hugin releases: one with the new functionalities and the old GUI and one with the new GUI. The old GUI will incorporate most of the improvements done in GSoC, while most of the new GUI still needs to be created. The one with the new GUI will have all the functionality under the hood, but not all of it will be available through the user interface, to reduce complexity.

Can you give us an estimate when this will happen?
Not really. I'm going through a lot of change: we moved to another city, I started a new job and our family is growing. Not much time is left for hugin besides that, I am afraid.

So what is next?
We take it one step at the time. We have decided to increase build frequency and the community is currently looking into setting up the infrastructure so that I can release snapshot builds more frequently. With this we hope to attract tester and power user, and maybe even people with coding skills to help getting hugin out. The current release is getting really old, people sometime ask me if it is true that hugin can handle HDR, because they still have an older version that can not.

The last GSoCTM project in our portfolio was an extension of FreePV, the first QTVR viewer for Linux. The project was blessed with mentoring of no less than Aldo Hoeben, the maker of SPi-V, and Ken Turkowski, the engineer/mathematician who developed the viewer for Apple's QuickTime VR ten years ago.

Hi Ken, thanks for joining Pablo and me in representing hugin/panotools at the Mentor Summit.
My pleasure. I'm living close by and I actually work here on the Google campus during the week.

What are you doing at Google?
I do image processing and computer vision for Google Maps, Earth, and Street View.

Pablo, you also initiated FreePV? why?
Yes. Again, there was no QTVR player for Linux and I wanted to see on my Linux computer the QTVR that were produced using hugin. So I had to write a player. Earlier Thomas Rauscher and Fulvio Senore started the FreePV project, but didn't have time to do much work on it. So I picked it up and developed it into a hardware accelerated panorama viewer for QTVR panoramas as well as spherical and cylindrical panoramas.

Ken, how do you feel about seeing the format you developed being played on an Open Source viewer?
Ken I think that it is great. A lot of panoramas have been produced in the QuickTime VR format, and I like having them viewable on a wide variety of platforms. FreePV has better media integration and can achieve substantially higher frame rates at higher resolution than Apple's QuickTime VR.
Pablo That's an unfair comparison. QuickTimeVR is all CPU while FreePV uses OpenGL hardware acceleration.
Ken OpenGL was neither as ubiquitous nor powerful ten years ago as it is now.
Pablo If Apple wanted they could add hardware accelerated QTVR playback to QuickTime. In fact, FreePV is published under the LGPL license and there is no legal obstacle for Apple to just copy the code and integrate it into QuickTime.
Ken Apple's QuickTime group has had higher priority projects in recent years, and hasn't had the resources to enhance QuickTime VR. Maybe their priorities will shift, given the rising popularity of panoramas in Google Street View and panorama production software.

So what was the goal for this fourth GSoC project?
Pablo Leon Moctezuma added support for SPi-V panoramas, cleaned up the QTVR code and started to add hotspots.

And how was it?
Pablo With so many expert mentors around it could not have been anything but successful.
Ken I participated in a few of the online meetings and found them very productive. I only had to mention quaternions as being a solution to a problem and Leon went on the net, searched and learned about quaternions and came back the week after with the code.

And what is the current status of FreePV?
Pablo There are still two or three known and annoying bugs before we can issue the 0.2 release. Leon has been quite busy. After the GSoC he came to Germany for an exchange year. He spent the first two weeks at our place. He is now slowly but surely getting back onto the code again, but the pace of development will be much slower because of other obligations.

Gentlemen, it was a pleasure being here with you. Thanks for the interview.
Ken Thank you.
Pablo Thank you, and hopefully we'll do better next year.

Yuval Levy, born in Israel, raised in Switzerland, currently living in Canada, Yuval's life is an exciting journey through multiple cultures. Educated as an MBA, trained as a CFA he never fitted into any one mold. 20% schooled and 80% self-taught, Yuval came to VR in 2003 and is fully into it, artistically and technically. In 2007 he was the administrator of the hugin/panotools participation to Google’s Summer of Code

Links:
Summer of code
Hugin homepage
Freepv homepage
Hugin at panotools wiki with help and support for users
Panoramas

Previous article about Yuval Levy:
PEACEFUL SCENES OF HAIFA AND THE NORTHERN ISRAELI COAST