Some generic thoughs on Autopano (rant-ish)

[igor]

by igor » Fri Jan 12, 2007 10:36 am

I think the issue is that most people asking for new features are the people that are most likely very advanced in their field and see this software as a way to bridge the gap between their average users and more accomplished ones.
Let me take an example:
when the first geocam (no relation to us) took photos from the shuttle they got this
http://www.musc.edu/cando/geocam/atacama/catlarge.html
it is now painless to do two things:
- download very large :-) 300 K photos
- putting them together to get a larger region than the initial photos.

One can now see the difference between images like this one
http://www.musc.edu/cando/geocam/atacama/c03l.jpeg
and compare them to current resolution obtained from satellites
http://maps.google.com/maps?f=q&hl=en&ie=UTF8&om=1&z=10&ll=-23.12268,-70.135345&spn=0.568334,0.933838&t=h

I have done similar comparison with photos taken from the International Space Station with current SLRs (http://hasp-geocam.blogspot.com/2006/12/orders-of-magnitude-resolution.html )
But in the same way that satellites are expensive objects (and using pushbroom technology, not CMOS/CCD technology), a little like the shuttle was when the geocam shots were taken, use of UAV/planes/balloon is very likely to allow less expensive means to gather the same data without the need for capital investment like 200 M$.

I have mentioned it before but I see APP as a way to allow to build remote sensing maps.
http://blogs.zdnet.com/emergingtech/?p=446
Sure the panoramas are nice when doing home stuff, but remote sensing people see this tool as a way to produce a different product altogether.

I could think of at least two other reasons not related to remote sensing where I think APP, if it is developed following my requirements would allow some new fields of endeavors to be created.

I realize that APP would like to know more about the features being used, but I think it is not appropriate for three reasons:
- you get on the privacy issue, my advice : avoid this at any cost.
- switch to a server based model if you really want to know. I wrote something about it a while back: http://nuit-blanche.blogspot.com/2005/11/server-side-scientific-software-as-way.html
yes, i know people will want to upload giga files, but it looks like in 1993, 300 K were very large downloads :-). I have mentioned before that Google seemed to associate themselves with panorama endeavors
http://www.cs.cmu.edu/~globalconn/, which would mean to me that bandwidth will not be an issue eventually.
- the advanced users like me (in my field) are tinkerers, there is no way you can reverse engineer my thought on how I am using a feature. And most importantly, I have seen this when building a web based task manager, sometimes, a tiny feature makes a whole difference. Most statistical tools are not that sophisticated to figure these things out.

Igor.

[hankkarl]

by hankkarl » Fri Jan 12, 2007 1:11 pm

Igor's post caused me to think about APP. IMO, the casual user will use a "free" stitcher, like ULead 360, or photoshop or Canon's stitcher. Yes, PS is not free, but the stitcher is no extra cost. So people only buy APP when the free software does not do what they want it to.

I found that APP was the best in terms of automatic stitching. RealViz was good but it was $500! Many of the other programs just didn't do what APP did, and had complicated user interfaces.

So all the people asking for features are advanced users, but eventually the new users will become advanced users, and may need the same things.

One issue is that I (and probably others) will see a need and suggest a feature. I think the suggestion is not the only way to do things and that better ways of filling that need may exist (Like the control point editor!). But the need is there. The feature suggestions (in the ideal case) result in users having a good discussion, better describing the need, and coming up with a solution that's better than any single proposed solution.

As far as the sub-question "why so many images"? The most extreme cases I use are an attempt to break APP. I think that if APP gives an error message that says "go buy more memory" or "learn to take panos" its ok, but for APP to crash is not ok.

[igor]

by igor » Fri Jan 12, 2007 1:33 pm

some more thought after seeing what Hakkarl wrote.

Take some advice from Kathy:
http://headrush.typepad.com/creating_passionate_users/

In the app business you seem to be in the disruptive business whereby you start with a small niche market that allows you to fund bigger and better projects (expansion of the original idea). As you can see from some of my request, I eventually want you guys to implement some features on PS.

Besides the server side business idea, there is another one which is that of STK (http://www.stk.com/) from AGI. You get the basic product nearly for free (STK) and then specialized users buy "plug-ins" that espouse better their needs, and trust they are making a killing on the specialized plug-ins (they are expensive).

Igor.

[[bo]]

by [bo] » Fri Jan 12, 2007 3:34 pm

you start with a small niche market that allows you to fund bigger and better projects (expansion of the original idea). As you can see from some of my request, I eventually want you guys to implement some features on PS.

Following that logic, everybody could want a different set of "some PS features". So in the end you kinda envision a single mega-software that does it all - from getting the photos off your camera, editing vector floorplans, doing 3D modeling SketchUp-style, mixing video and audio files, exporting to 250 formats and, oh, I almost forgot - doing panoramas.

I think that each software should perform a simple set of things, but to perform them really well. And it's not just me - Adobe seems to think likewise too. And Apple. And Microsoft. The list goes on. Read near the top of the topic where I write on this.

Once upon a time, before the Adobe Creative Suite appeared, there were speculations on one single mega application that retouches photos, does vector work and lays out pages. That PS, IL and ID are merged. Only Adobe knew better and now there is PS Elements and Lightroom and Bridge, so PS actually was split into several parts, in order to not loose focus.

So, if I was making APP, I would implement plugin architecture, put the most essential auto-stuff in the core and sell the high end specific manual things as plugins. And I would post my custom plugin creation fees, so Igor can contact me and I can tailor him a "Low Orbit Thingie Plugin" and I could do all Panorama Fanatic's requests as "The Gigantic Plugin for Fanatics" for a fitting, gigantic price, of course, etc. And publish an API, so other can develop and sell plugins too, which will drive more customers in my direction.

How's that for a world domination plan, huh?

[igor]

by igor » Fri Jan 12, 2007 3:59 pm

I think it is really a non debate to talk about world domination, what is known is that single products that found a good enough of a niche to survive and eventually develop to yield the mega application have done so through some organic growth. By that I mean that they found additonnal niches as the software grew. There is simply no business case to build this large application in the first place.

The reason I think of the server side is really because of the ability to not have to depend on the OS configurations of so many people and the fact that eventually bandwidth will not be an issue. Bug tracking could even be done real time.

I agree with Bo that an API where developer lambda can build a plug-in from a known API and sell it would be an attractive way of developing different communities. I think I mentioned to Alexandre before that if I were to be able to have additional capabilities, I would pay extra.

Igor.

[steve]

by steve » Sat Jan 13, 2007 7:56 am

To Steve. I just don't see why you should use hugin to fix something. We have a far better tool with our control point editor. I'm really curious to see some of your panorama as the only case we cannot handle yet is the full blue picture in the sky

Hi AlexandreJ! My photos are Nikon D80 + 10.5mm, through DeFish to 85x135 with no EXIF information... so I think you already know why they don't work properly.

I use Hugin to set the control points and to do optimisation, as Autopano cannot optimise or lens calculate for the 85x135 photos, even though they're DeFished to 0.00042 distortion.

My problems will go away and I will only need Autopano (no more Hugin or DeFish) when 1.4 is released with the fisheye lenses support. So I'll wait instead of using up your time.

If you want a beta tester that knows a lot about lens distortion and fisheye, and can compare against manual results, I'll be in on that one.

Last edited by steve on Sat Jan 13, 2007 8:00 am, edited 1 time in total.

[GURL]

by GURL » Sat Jan 13, 2007 10:40 am

Photoshop

Photoshop was designed, between 1987 and 1990, for "the graphic art industry", not for photographers.

"The program was intended from the start as a tool for manipulating images that were digitized by a scanner, which was a rare and expensive device in those days." ( http://creativebits.org/the_first_version_of_photoshop ). It was designed for "desktop publishing" and was then used by graphic designers long before it was actually used by photographers.


Stitchers

Stitchers are mostly seen as a panoramic images tool, but I don't fully agree.

Stitchers could be seen as a tool to be used when one want to produce a single final image from several source images:
- panoramas (in the usual meaning of large FOV photographs)
- high resolution mosaics (to avoid using very expensive or not "still available" cameras) where FOV is anything you need
- "time lapse" images where, for many possible reasons, all the sources images where not taken at the same moment (to remove peoples and cars, to add more peoples and cars, to show them at different positions on the same image, to combine different lightings onto the same subject, etc)
- high DOF images (à  la CombinZ or Helicon Filter)
- HDR (high dynamic issues are common in wide-angle panoramas. When a very large FOV and very high resolution are not involved, taking a single shot using an (not too expensive) HDR camera would be preferable.
- did I missed some ?

Forums

Four or five years ago I decided to learn GIMP. I found a decent Windows version and a very good manual but was not able to find any GIMP photographers forum (though I found several used by graphic designers.) I was discouraged.

That APP forum could reflect an inexact image of its actual users is a possible issue.

[steve]

by steve » Sat Jan 13, 2007 10:47 am

That APP forum could reflect an inexact image of its actual users is a possible issue.

I like to think I'm a typical user, because I only want the panoramas to come out properly every time. It is true that I have a reason for wanting such a high degree of reliability, but this should be what Autopano achieves anyway.

The other things that I see Autopano being used for are definitely pushing the boundaries of traditional photography, but they're also a consequence of digital photography and computers put together.

I think it would be good if Autopano could be driven by command line and scripts, so that processing can be done in stages and invoked in various orders. Currently, the GUI limits the options somewhat.

If there's no button for it, then it can't happen. Command line would work around this problem, allowing Autopano to be embedded into various workflows.

[igor]

by igor » Sat Jan 13, 2007 5:08 pm

That APP forum could reflect an inexact image of its actual users is a possible issue.

I like to think I'm a typical user, because I only want the panoramas to come out properly every time. It is true that I have a reason for wanting such a high degree of reliability, but this should be what Autopano achieves anyway.

The other things that I see Autopano being used for are definitely pushing the boundaries of traditional photography, but they're also a consequence of digital photography and computers put together.

I think it would be good if Autopano could be driven by command line and scripts, so that processing can be done in stages and invoked in various orders. Currently, the GUI limits the options somewhat.

If there's no button for it, then it can't happen. Command line would work around this problem, allowing Autopano to be embedded into various workflows.

A command line would be great if an API cannot be developed. I for one, will eventually have to hack into the .pano file to find some of the information I want.

I am sure that the user base interacting in this forum eventually has many specialists of their own fields and indeed, do not reflect the average user. But I don't think this is a reason to not be responsive to that community on the pretense that it does not reflect the average user base of the software (not that I think that APP has ever been non responsive).


Igor.

[hankkarl]

by hankkarl » Sat Jan 13, 2007 11:44 pm

First, I think Alexandre has to tell us what APP is and where it is going. At first, I thought APP was only a stitcher, but it seems to do something with HDR.

Will APP be a great stitcher with some HDR capability, or will the HDR capability be increased to compete with PhotoMatrix? The development plan http://en.wiki.autopano.net/Autopano_Pro_Devplan doesn't make this clear.

Note that APP can generate images larger than PhotoMatrix can handle. :-)

[AlexandreJ]

by AlexandreJ » Sun Jan 14, 2007 10:22 am

Our final plan ?

To make the best framework for stitching and blending :
- Panoramic stitching
- 2D stitching orthographic
- HDR (yes, we are working on tone mapper to compete with photomatix 'without the r' )
- We will also do ghost remove on HDR
- New blending mode : to enhance DOF for example
- Our internal ghost remover which will be compatible with the cell rendering
I think that the best way to achieve that will for us to it in two steps. First step will end with 1.4 version (fisheye and orthographic mode).
The second step will be v2.0, I still don't know what will be really inside but I still have a huge list of ideas ... We may rethinking many part of the software to achieve an even better workflow.

There are two features that we really want to add but for the moment I don't think it can belong to the Autopano Pro design or goal. I see then rather in another Autopano (let's call it Autopano Framework) :
- a full plugin architecture : you will be able to write your own control point detector or validator, your blender, a new lens model or our own exporter.
Anything should be customizeable.
- scripting language. I think it will be the first feature available : we want a scripting language (like javascript for example) to control the software.
You will be able to control the software like nothing else. You will have access to the groups, you could create groups using your coded rules, parse folder, start rendering,
check stitching, change CP, anything should be scriptable. The aim is to have a command line like this : autopanoframework yourscript.panoscript : and it will the job of the script.
The target is fully automatic workflow but still customizable.
I don't think Autopano Pro has this goal : it's more a GUI tool. So this idea of Autopano Framework version is perhaps a good split to target two different audiences.

We're still not fixed here if this direction is right or not. We still have work for months in the devplan on autopano pro until the final 1.4 version.
BTW : nobody talked about the Autopano Light ... Why ?

[GURL]

by GURL » Sun Jan 14, 2007 1:29 pm

AP Lite

- it is becoming more and more obvious we need a AP Lite and AP Pro separate versions.

What I don't agree with is the separation of a lite and pro version. This makes things complicated and versions might drift apart.

[..] Adobe Photoshop Elements against Adobe Photoshop "full version". As it's more expensive for Adobe to produce and market different versions of the same software, the real question is: how many users of Elements bough it because they believed it was easiest to use than the full PS CS version?

I didn't really wanted this to be a Pro/Lite discussion... and I'm not sure whether this is a good road to follow. Maybe a "wizard"-kinda thing, I don't really know.

people ask for more features they can manage, for everything
buy a Porsche Cayenne 4WD but never get offroad
buy MSWorld just to type memos
you own a powerful system, your supposed to handle it in others mind, and that's enough for your happiness

on 80% of the projects I get acceptable results without opening the CP Editor. My thinking is: "let's try to reduce those 20% that need manual intervention via smart application logic"; as opposed to the much requested "give me more tools and options, so I can work on those 20% and the rest 80% myself".

(Note: I quote this from BO because I see the whole Link Editor as the most meaningful candidate for removal.)

Perhaps mak[e] the GUI have 2 settings, Normal (like we see now) and experienced (which allows people to have extra settings).

I agree that in the community here a great part is composed of high end users It's not representative of customers, globally.

If you want to see a totally automatic panorama tool, have a look at Calico. That's a drag'n'pray tool for those who can say "oh well, that one sucks, let's try something else". Autopano can aspire to that, or it can handle reality and provide the tools to correct for situations it cannot handle.

IMO, the casual user will use a "free" stitcher, like ULead 360, or photoshop or Canon's stitcher. [..] So people only buy APP when the free software does not do what they want it to.
I found that APP was the best in terms of automatic stitching. RealViz was good but it was $500! Many of the other programs just didn't do what APP did, and had complicated user interfaces.

Some questions:
- price ? (the main point - you must decide - we can't !)
- name ? (is "Pro" to be removed from the name?)
- limitations ? (fisheye, HDR, Layer Editor and PSD layers, export to PanoTools and the link editor are obvious candidates but you could decide of some "arbitrary limitations" like final size, source images number, RAW, 16 bits, batch rendering, TIFF, PSB)
- specific documentation ? (or would you remove any "non self-explanatory" feature:) )

Interestingly enough and to my great surprise, $49.95 (55,75€) MS Works still exists: http://www.microsoft.com/products/works/ProductDetails.aspx?pid=002&active_tab=Compare

[hankkarl]

by hankkarl » Sun Jan 14, 2007 3:50 pm

GURL,

I would like to see two levels of interface: Basic and Advanced. Ulead Photoimpact has this, and you can switch back and forth. I would probably leave things in advanced because I know it (as I do in photoimpact). An alternative would be to have a wizard that guides new users into building a panorama.

The Autopano Framwork (APF? F looks to close to P to my bleary eyes!) would be the very advanced tool that some need. I don't want to have to learn a new language to use it (I know C++ but not Java/Javascript) and besides, I want to make panoramas, not program.
"

[GURL]

by GURL » Sun Jan 14, 2007 9:01 pm

I have always wished for my computer to be as easy to use as my telephone; my wish has come true because I can no longer figure out how to use my telephone.

Somewhat out of topic, but I could not resist !

[[bo]]

by [bo] » Mon Jan 15, 2007 9:06 am

Well, it looks like all has been said now... Now it's time to just sit back and relax and do some panos!

[[bo]]

by [bo] » Mon Feb 12, 2007 1:36 pm

Oh, just posted: http://wisdump.com/business/everyone-wants-the-world/

One more post on the topic of "every user wants every feature, but you should not give it to them just because they want it".

[phototrek]

by phototrek » Mon Feb 12, 2007 7:13 pm

']What do you do with those huge projects you stitch?

I haven't done many silly big stitches, relatively speaking (for my print size of 60x40"). The two instances where I stitched absurdly large images:

- 154x 16MP: Just to see if it could do it, since I was bored on location I started snapping. It could do it - my Quad G5 took 8 hours.

- 12x 39MP: This was intended to be used with Zoomify, where a large scene was to be shown, with the option of reading license plates (as an example; the actual sought after detail was different).

But I agree, typically less is more.