Photosynth location browser

July 29, 2006 10:02 AM

Photosynth is a new image processing technology from Microsoft that takes a collection of images (say, of a famous location), analyzes them for similarities & rebuilds the location in virtual space for the user to fly though, zooming in on details, panning around like a 3D Hockney piece. Video of how it works here.
posted by lilbrudder (83 comments total)

Crashed FireFox.
posted by odinsdream at 10:17 AM on July 29, 2006

This looks like bullshit. It only works cos they went out and took all the photos themselves - they seem to be claiming they can line up photos taken by numerous people at numerous different times.
posted by Orange Goblin at 10:18 AM on July 29, 2006

That's actually really cool, if it works. It'd be hard to resist going "enhance... enhance..." while you're using it.
posted by interrobang at 10:20 AM on July 29, 2006

Didn't crash my Firefox.

Interesting attempt and experiment. We'll see where it goes.

Of course, it'll only really be cool after Apple invents it.
posted by juiceCake at 10:33 AM on July 29, 2006

This looks like bullshit.

I wouldn't dismiss this project out of hand. Sure the examples may be somewhat contrived, but remember, this is a rapid prototyping lab, not a production house. Your asking for a consumer-ready product from something that was never designed to be.

If people were not feee to design things like this, we would never know whether or not they were useful.
posted by WetherMan at 10:37 AM on July 29, 2006

Going off-topic, but, is it just me, or is live.com becoming a bigger rip-off of google.com every day? They have a "new" search engine, a customisable home page, a clever javascript maps/earth photos engine, and now even a "labs" subdomain.

Seems a bit blatant.
posted by Mwongozi at 10:40 AM on July 29, 2006

Photosynth whatever aside, has there been any talk on the blue about how Microsoft has been adopting this new, cleaner design look?
posted by jeremias at 10:40 AM on July 29, 2006

One day I am going to hook up some voice recognition to my computer for some software like this, or even Google Maps, so I can really do a "Computer! Zoom in! Enhance!"
posted by TheOnlyCoolTim at 10:57 AM on July 29, 2006

Mwongozi writes "Going off-topic, but, is it just me, or is live.com becoming a bigger rip-off of google.com every day? They have a 'new' search engine, a customisable home page, a clever javascript maps/earth photos engine, and now even a 'labs' subdomain.

"Seems a bit blatant."

You just noticing that MS are shameless followers?

I think not being first into a market is somewhere in their mission statement. They don't invent, they embrace and extend. Oh, and then they break it so it'll only work with windows.
posted by mullingitover at 11:22 AM on July 29, 2006

This looks like bullshit. It only works cos they went out and took all the photos themselves - they seem to be claiming they can line up photos taken by numerous people at numerous different times.

Hahah, as if that wouldn't be impressive? Even if they did take all the pictures themselves, it would be a huge breakthrough. And it dosn't seem like it would be very hard to move from a bunch of pictures they took to a bunch of pictures other people took...
posted by delmoi at 11:37 AM on July 29, 2006

Seems a bit odd to trust random people's photos like that. What if I decide to start taking pictures of landmarks and then photoshopping in a big gorilla?
posted by reklaw at 11:52 AM on July 29, 2006 [1 favorite]

This would work great with Flickr and a particular tag. They just need to hook it up to RSS and they're set.
posted by mullingitover at 11:53 AM on July 29, 2006

This looks like bullshit. It only works cos they went out and took all the photos themselves - they seem to be claiming they can line up photos taken by numerous people at numerous different times.

Though they've obviously skewed their data set to get the coolest possible results and no mismatches, what you're describing is as good as possible with the current state of the art algorithms. And even though Microsoft might be shameless followers when it comes to other things, their image processing research department is a force to be reckoned with.

Another application of this research, and the one which most other people focus on, is image based GPS. The scenario is simple: You take a picture with your camera phone, it gets uploaded and matched à la Photosynth to a large database of photos, and immediately you get your position (and whatever useful local information you need.)
posted by Runkst at 11:53 AM on July 29, 2006

"What if I decide to start taking pictures of landmarks and then photoshopping in a big gorilla?"

More like photoshopping the Washington Monument into the middle of St. Peter's Square.
posted by Ethereal Bligh at 12:04 PM on July 29, 2006

Of course, it'll only really be cool after Apple invents it.

Hey, look! They did. It's called Quicktime VR. Woooo, fancy.
posted by loquacious at 12:16 PM on July 29, 2006

Hey, look! They did. It's called Quicktime VR. Woooo, fancy.

Uh... huh.
posted by kbanas at 12:23 PM on July 29, 2006

Well, yeah, it's basically QTVR stitching in real-time. So is that the huge big advancement from Microsoft?
posted by Blazecock Pileon at 12:36 PM on July 29, 2006

No, it's, in fact, a hell of a lot more than QTVR stitching in real-time.

Have you watched the videos on the site?

If I'm not mistaken, it takes a random assortment of photos, dynamically stitches them in real time and then uses that to render a 3-d model of the location.

If I'm not mistaken, doesn't QTVR simply allow you to pan and zoom around a fixed point in a psuedo 3D manner?
posted by kbanas at 12:39 PM on July 29, 2006

If I'm not mistaken, it takes a random assortment of photos, dynamically stitches them in real time and then uses that to render a 3-d model of the location.

That's pretty much QTVR, kbanas. The only difference here is that you can jump around to different nodes, but the underlying process is the same.
posted by Blazecock Pileon at 12:45 PM on July 29, 2006

Blazecock: WATCH THE VIDEOS!
posted by VulcanMike at 1:02 PM on July 29, 2006

I've watched the videos! It's not new stuff!
posted by Blazecock Pileon at 1:04 PM on July 29, 2006

That's pretty much QTVR, kbanas. The only difference here is that you can jump around to different nodes, but the underlying process is the same.

No. It's not.

Seriously, did you watch the video? Do you know anything about QTVR? Everyone knows it's possible to stitch still images into a 3d view. What Photosynth does -- supposedly, because we can't play with it yet -- is sift through a hugh-ass bin of images and algorithmically identify points of similarity to figure out which ones are different views of the same thing. Then, it stitches together the different individual images automatically.

It may be vapor, sure. All we can see is blog posts and a video clip right now. But saying this is "just like QTVR" is like saying that OCR software is "just like typing." The big deal isn't "viewing things in 3d," it's the automatic creation of that 3d view from many disparate stills not explicitly taken for the purpose of stitching
posted by verb at 1:07 PM on July 29, 2006

.
posted by verb at 1:08 PM on July 29, 2006

.
posted by Blazecock Pileon at 1:10 PM on July 29, 2006

The underlying process of this is a lot more complicated than QTVR. They're not just stitching together photos but also calculating 3D positions (from what I can tell in the video). Plus, the fact that the photos are taken from all over the place makes the matching/stitching a lot more difficult (I don't even know if QTVR even does matching).
posted by Runkst at 1:11 PM on July 29, 2006

Thanks for the post. This technology in general, whether this one from Microsoft, QTVR or others, is certainly fascinating and promising. But in the video showing off the software at the linked site, did it appear to anyone else that as the demonstrator navigated around what was captioned as, and initially in fact was, "St. Peter's Square" (in Rome, of course), the images shifted eventually to views of St. Mark's Basilica in Venice? I caught a glimpse of the bell tower in St. Mark's Square somewhere in there and thought perhaps I'd been mistaken, but then nearer the end, where he zoomed in on some arched mosaics on the facade's exterior, it was even more definitely obvious that the images had transitioned to the Venetian square and at least this segment wasn't depicting St. Peter's any longer. Assuming the software found "similar images" elsewhere on the web and compiled them into the VR simulation, it appears it is prone to make a few mistakes. Still impressive, but maybe not quite ready for prime time? (Myself, I'd rather look at Venice than the Pope's palace any day, but I thought it was worth noting that the demo designed to show off the software apparently had a slip-up they overlooked.)
posted by NetizenKen at 1:17 PM on July 29, 2006

It'd be hard to resist going "enhance... enhance..." while you're using it.

Funny, because I'll be expecting to see this around about the time speech recognition fully matures.
posted by MetaMonkey at 1:19 PM on July 29, 2006

I am the director of Live Labs (I am the guy in the video with the short ratty hair). And before you shout "pepsi blue", please know that seeing our project linked at MeFi was a pleasant surprise for me. Anyhow, some details:

1. The photos are analyzed offline to determine how they spatially relate to one another. Absolutely no human labor is used to position them. Moreover, the accuracy rate is pretty stunning. We've used collections from flickr and other sources, so while St. Peter's looks great, this will work for your house as well and with any digital camera.

2. The online stitching and animation is greatly simplified because of the offline work. We are presently working on making this more intuitive and even slicker.

3. We are building this on top of our seadragon platform, which means that the photos can be stored remotely with close to no performance hit.

So, to make an analogy, step 1 is akin to building a web search index (a lot of heavy lifting in advance); step 2 is like doing a web search (using the index to make more sense of many objects); step 3 is like browsing the web but only through remote caches.

We are shooting to have all of this ready for people to play with in the fall. We released the movie and other details now because the academic work will be presented at SIGGRAPH on Monday. I want to also point out that this is a joint effort between MSR, the PIX and Virtual Earth teams, U. of Washington, and Live Labs. The project is for months old and it is not a product in the usual sense. It is a preview which we hope to inspire discussions like this because we are eager to see how this project can evolve, and believe that input from the larger Internet community will make it better for all.
posted by dr.flakenstein at 1:20 PM on July 29, 2006 [10 favorites]

Turns out the 3D modeling part was done by a group at Carnegie Mellon lead by Efros, outlined in this SIGGRAPH paper from 2005, and QTVR's image stitching algorithms have been around since at least 1995, possibly earlier. Interesting demonstration of extant work.
posted by Blazecock Pileon at 1:22 PM on July 29, 2006

NetizenKen: we have collections from different locations, and you can view all of them independently. When we shot the video, we were playing with multiple collections to show off the technology better. However, in editing, different demo sessions got spliced together. So, no, we didn't confuse two different locations - we just had to edit the video down.
posted by dr.flakenstein at 1:24 PM on July 29, 2006

Orange Goblin: that is exactly what we are claiming and exactly what we do. One of our contributors is getting a Ph.D. for the work that he's done. His work is being presented at SIGGRAPH next week, so the work has been thoroughly vetted through BS filters.
posted by dr.flakenstein at 1:26 PM on July 29, 2006

Here's how Photosynth differs from QTVR:

1. Photosynth uses regular images from multiple cameras, times, and sources.

2. Photosynth stitches everything together automatically without human intervention (but offlline).

3. Photosynth has a browse mode that does not require the data of these collections to be locally stored. We have individual collections that are many GB in size, and we can give the fluid experience shown in the videos even with only one copy of that data on one server location.
posted by dr.flakenstein at 1:29 PM on July 29, 2006

That's come crazy bladerunner shit right there.

Enhance 224 to 176. Enhance, stop. Move in, stop.
posted by angelplasma at 1:43 PM on July 29, 2006

One minor correction to my last comment.... Item 2 should say "spatially relates" instead of "stitches". The stitching is actually done in realtime.
posted by dr.flakenstein at 1:45 PM on July 29, 2006

...and two people beat me to the obvious.

in any case, what does this have to with QTVR, aside from the mac faithful (of which i'm one, really) looking at MS and going, "whoa, wait, they... wait, Microsoft is doing something cool? well, crap."

guess that's what happens when you got money to burn. it gets put toward cool projects now and then.
posted by angelplasma at 1:46 PM on July 29, 2006

dr.flakenstein, how successful is the software at spatially relating large, unstructured data sets? I would imagine it gets pretty tricky when you have thousands, or tens of thousands of photos of, say, european churches, which may look fairly similar. Or would you rely on tagging or somesuch in the final system to do the initial grouping?
posted by MetaMonkey at 1:51 PM on July 29, 2006

That's very cool actually. Apart the (quite important) fact that the process seems to be largely automated, how different is it from the various photogrammetry /camera mapping software that's used for CG? And since a 3d model (or at least a point cloud) is reconstructed from the photographs, will it be possible to export it in a regular 3D format for later use? Or is the purpose completely different?
posted by elgilito at 1:56 PM on July 29, 2006

So, in the event that there's sufficient photo data to link the interior and exterior of a building - or even say a house - someone unwittingly posting photos of their cool stuff in their studio might unwittingly be setting themselves up as a victim for burglary, as here's a technology that finds the links to the information the thief needs that the photographer didn't even realise existed. If he's really lucky he might be able to case the joint's security from the comfort of his armchair.
posted by nthdegx at 3:20 PM on July 29, 2006

nthdegx: I really like your "turn cool technology into bad news" approach there. Keep up the good work!
posted by blindcarboncopy at 3:59 PM on July 29, 2006

Yeah, I'll admit I wasn't hesitant to post that, but can you tell me the scenario I outlined above is impossible? Don't get me wrong - I think this technology looks very cool. I'm particularly interested in the effects of throwing every pr0n photo on the web into the mix and then exploring all the weird places people have had sex, perhaps using Google Earth to chart correlation to lay lines, and generally I am an extreme technological optimist, but the thought occurred to me so I threw it out there.
posted by nthdegx at 4:09 PM on July 29, 2006

No, his point is just an example of a larger, legitimate worry about this and all similar technologies. The optimistic hopes about the application of such technologies—technologies that rely upon huge public databases of what would otherwise be private experience—completely ignore all the ways in which these things erode or eliminate effective privacy and whatnot and the bad things that can, and will, result.

But I'm quite close to someone who is effectively anti-privacy, so I don't much care, myself.
posted by Ethereal Bligh at 4:11 PM on July 29, 2006

How would it handle Levittown?
posted by StickyCarpet at 4:34 PM on July 29, 2006

This is definitely not bullshit. I've translated articles on fairly similar techniques.

elgilito: IIRC, they actually used the technique of constructing a 3D mesh from a bunch of 2D photos for Matrix Reloaded (specifically to build 3D images of Neo and Agent Smith, which were then computer animated...and you thought that was just Keanu's wooden acting). In that case, they were working in a more controlled environment (they new exactly the position and angle of the camera and subject), but clearly it could be extended to this kind of thing.
posted by adamrice at 5:08 PM on July 29, 2006

err new=knew
posted by adamrice at 5:11 PM on July 29, 2006

Warning — spoiler dialog:

"Chloe, open up a Photosynth socket, and fly me in there — stat!!"

posted by rob511 at 5:34 PM on July 29, 2006

I'm quite close to someone who is effectively anti-privacy

I'd love to see you try to defend that stance. Just not in this thread.
posted by stavrosthewonderchicken at 5:36 PM on July 29, 2006

nthdegx writes "So, in the event that there's sufficient photo data to link the interior and exterior of a building - or even say a house -"

Considering how many interior photos seem to include views out of windows I'd bet there are a lot of interior/exterior match ups out there.

I'd be more worried about stalkers than thieves. Unless you've got a Monet or something hanging on the wall thieves rarely travel much to get a specific item (profit magins are too low). Stalkers on the other hand seem to know no bounds.
posted by Mitheral at 6:03 PM on July 29, 2006

Man, I've been wanting something like this for Google/SketchUp. I should be able to take two photos 25' apart, and have software that figures out the relative heights and positions of everything in the image. With a dozen photographs I'd have my entire (steep) yard and house mapped into Sketchup, where I'd then be able to really start doing some kickass design work.
posted by five fresh fish at 6:20 PM on July 29, 2006

well the big problem i see is that its a small flaccid product.

those idiots can't seem to make a decent OS on their own even after nearly 20 years of trying, what makes anyone think they might succeed at something ground breaking is beyond me.

vaporware for sure and probably just a ploy to fool some bone head wall street analist (yes i do know how to spell analyst )
posted by coyote's bark at 7:09 PM on July 29, 2006 [1 favorite]

Ah yes, you can always depend on the HAHAa micro$oft $ux amirite brigade for their insightful analisis.

And yes, I do know how to spell analysis.
posted by stavrosthewonderchicken at 7:27 PM on July 29, 2006

wow, coyote's bark, your rant makes you sound like a 15 year old.
posted by jonson at 7:40 PM on July 29, 2006

so do you disagree with my assessment that ms has not produced a quality os or that they are incapable of innovation?

the track record of ms speaks for itself

they buy a good idea, and spend millions on screwing up..
posted by coyote's bark at 7:49 PM on July 29, 2006

It doesn't seem like the people involved in this lab/project have anything to do with the OS.

Do they somehow inherit failure from the name of the company? Or maybe the money they're paid with is covered in some sort of failure agent?
posted by ODiV at 8:01 PM on July 29, 2006

id suggest that they do inherit a failure mode from the company, and the track record of ms supports this..

the culture of the company tends to ruin good ideas in very short order.

can you point toward an example of a really good ms product?

ya i didnt think so.
posted by coyote's bark at 8:05 PM on July 29, 2006

I hate Microsoft as much as the next guy, but keep in mind that they spent many years sucking up all the smart people into their R+D development operations, much like Google is doing now, so something good had to come of all that.

MS products suck in general because the market does not value quality in software. It values feature lists. So they are simply adapting to what the market desires, creating buggy software that has a gazillion features that 99.9% of us will never use. But that doesn't mean they are incapable of producing very good and interesting software.

Thanks Dr. F for your comments!
posted by intermod at 8:16 PM on July 29, 2006

I like their Media Center. I haven't had any problems with their optical scroll mice. Windows 2k is a solid OS in my experience. I own an Xbox and I would say it's a really good entertainment console.

These are all just my opinions though, so I guess there's nothing stopping you from just saying "No, those all suck." (except without punctuation)
posted by ODiV at 8:21 PM on July 29, 2006

Rather, Microsoft may very well be good at producing excellent software (NT kernel for one, NTFS for another are both first-class technically) but the company as a whole doesn't have enough enlightened management or QA sophistication to get those technical wins to market effectively, instead relying on cramming in more features, buzzwords, embrace-and-extend, eyecandy, etc. (Windows XP, Vista, everything else)
posted by Skorgu at 8:23 PM on July 29, 2006

intermod-

good analysis, but i dont know how well ms is doing at hanging on to the smarty pants people they hire. you may be correct as to the feature list comments but i would be very curious to hear what you think they have produced that could be considered very good and or even interesting.

dr f seems to be a smart guy, smart enough to probably be thinking about how he can keep the juggernaught from killing his baby.
posted by coyote's bark at 8:23 PM on July 29, 2006

Grr. I'm regretting my punctuation dig. I didn't mean to get personal (and I should have said capital letters :P ).

You were just getting under my collar because you seem to dismiss the efforts of these people out of hand when it they seem very capable and committed to this.
posted by ODiV at 8:30 PM on July 29, 2006

i have no doubt that the people involved are both brilliant and commited, its the company behind them that manages to screw up their efforts, and that fact has been proven time and time again. put those same people in another company and their dreams might have a good chance of making a difference. as it stands, their work will probably be wasted.

as for the dig, i take that stuff as a compliment, because im silly that way
posted by coyote's bark at 8:40 PM on July 29, 2006

"so do you disagree with my assessment that ms has not produced a quality os"

Yeah. NT 3.1 is almost universally thought to have been a fine OS. Notwithstanding that Cutler brought a lot with him when he went to MS, there was a lot of good work in the OS. And it should be noted that much of what was good about NT became not-so-good in subsequent years, and that why and how that it is exemplify Microsoft's problems.

I have no particular interest in defending Microsoft. But simplistic teenboi-like disparagments like yours are annoying the same way teenboi-like your-band-sucks are annoying.

"I'd love to see you try to defend that stance. Just not in this thread."

I chose my words carefully, though perhaps not carefully enough. A better characterization would be something similar to Ellison's there isn't any, get over it, mildly justified with pointing out that modern urban notions of privacy are atypical in human history and obviously have social costs of their own.

But mostly I don't care much. Hopefully for the integrity of the thread, that wasn't an actual defense.
posted by Ethereal Bligh at 11:42 PM on July 29, 2006

what makes anyone think they might succeed at something ground breaking is beyond me

Well, for me at least, the long high-res demo videos of a functioning ground-breaking technology were my first clue.
posted by chrismear at 11:54 PM on July 29, 2006

Does Photosynth have anything to do with Autostitch?
posted by redteam at 1:08 AM on July 30, 2006

I once read an article by someone who said that the 3d CGI graphics in toy story were "No different" then the puppet animation in the 1970s. The people saying, "this is just like QTVR" are like the guy who wrote that article. They have no idea what they're talking about.
posted by delmoi at 1:18 AM on July 30, 2006

A better characterization would be something similar to Ellison's there isn't any, get over it

That I could go along with, depending on where and how one lives.
posted by stavrosthewonderchicken at 1:39 AM on July 30, 2006

Existing software like autopano-sift can automatically find correspondances between many images. One problem is that the amount of work increases as the square of the number of images, if any pair of images might contain correspondances--when you double the images and you'll have to wait four times as long. It looks like the SIFT method was published in 2004. It's not hard to believe that two more years of research on this very interesting topic has produced better results. But maybe efficiency isn't yet very good, the Labs poster above referred to this step being done "offline" (i.e., beforehand and slowly).

The blending step, as done by software like enblend, is not too tough either. It's easy to imagine a fast OpenGL (Or directx, if it's coming from Microsoft) implementation where edge-falloff and an overall correction value have been precomputed. Then the rendering step becomes a matter of drawing some textured polygons, lightning fast on today's new PCs.

I'm not trying to say it's not neat, but it is incremental. Like almost all research is.
posted by jepler at 6:15 AM on July 30, 2006 [1 favorite]

"... simplistic teenboi-like disparagments like yours are annoying the same way teenboi-like your-band-sucks are annoying."

interesting... i find that name calling is usually derived from the lack of a ability to form a coherent argument.

what exactly does teenboi-like mean anyway?

"And it should be noted that much of what was good about NT became not-so-good in subsequent years, and that why and how that it is exemplify Microsoft's problems."

an inelegant way of making my point .. thanks
posted by coyote's bark at 6:52 AM on July 30, 2006

So, this is a slight de-rail, and a bit of back-patting for all of us, but I think one of the coolest things about mefi is that we find ourselves discussing something I've never heard of and then someone who was either at the event or had a hand in whatever we're discussing chimes in with their take on things. I think that's pretty damn coo.
posted by craven_morhead at 8:15 AM on July 30, 2006

"interesting... i find that name calling is usually derived from the lack of a ability to form a coherent argument."

Perhaps you'll note that "teenboi" applies to the style of disparagement, not you. What does it mean? I meant it to imply that male adolescent sort of disdain that's soaked in notions of identity built around trivial technogeek personal associations. I've been in IT for years and years. My first Linux install was Slackware in 1994. It's a cliche, a stereotype, but it's also simply true that a large number of people regularly bash Microsoft using simplistic arguments they've heard from their peers to the point that it's more about social identity than any actual argument. Maybe you're not one of those people. Your comments sure as hell sound like theirs.

"an inelegant way of making my point .. thanks"

No, it was an elegant way of making your point—that Microsoft is and always has been badly flawed—compared to your very inelegant method, which was making an assertion that was untrue (that they've never produced a decent OS). I'd suggest you adopt the convention of supporting your arguments with true statements, along with the appropriate use of the shift key.
posted by Ethereal Bligh at 12:09 PM on July 30, 2006 [1 favorite]

dr.flakenstein, What an interesting and exciting project to be working on! I wish you and your team success. I look forward to seeing it in action.

It might also be useful as an interface between personal and collective memory, something like collected visions.

It reminds me in part of some of the applications of visualcomplexity, sort of a combination of pattern recognition and knowledge networks. Visual affinities.
posted by nickyskye at 12:37 PM on July 30, 2006

NT 3.1 is almost universally thought to have been a fine OS

Its VMS core did not originate within Microsoft, though, so it therefore seems hardly fair to ascribe Microsoft much praise for its stability.

Indeed, Digital had successfully sued Microsoft some years prior to NT 3.1's official release for stealing code from VMS, which was found in the Mica project that led up to NT.

Ironically, the mention of NT in this thread only underscores the nature of how Microsoft "innovates" to this very day, even with Photosynth: Copy, copy, copy, and if you get caught, settle out of court at pennies on the dollar.
posted by Blazecock Pileon at 9:44 PM on July 30, 2006

"Its VMS core did not originate within Microsoft..."

Yeah, I thought I explicitly acknowledged that or alluded to it in my comment. May have been lost in revision.

But for the purposes of the argument, that really isn't relevant. The claim was made that MS had never produced a decent OS in support of the argument that MS is crappy in general. I happen to agree with the conclusion, but the argument is faulty because that premise is untrue. If everying MS ever produced was second-hand technology but excellent it would be contradictory to acknowledge this and then assert that what MS produces is crappy. It has no bearing on that conclusion.

At any rate, MS is deeply to its core not an innovator, though they have a relatively very few things which are exceptions to the rule. In keeping with the previous paragraph, as a consumer that they're not really doesn't matter to me. I like innovators, but I like companies that produce solid, quality products even more. That they don't is where MS should be criticized—and Apple is similarly overpraised for its innovativeness. Someone has to be innovative. But what really matters is if the damn thing works or not.

I don't want to insult or be rude to dr. flakenstein, as I'm pleased with his contributions to this thread—which weren't required and he was sticking his neck out in posting them. But when I watched the two videos I was struck by how he, himself, nicely exemplified MS's problems in this context. The architect impressed me and seemed exactly like the sort of person you'd find in a technology lab like that. His explanation of the phtosynth and its applications were all of the same character and were very harmonious with the notion of such labs. DF, though, seemed to me to be obviously pulled in two seperate directions—describing and praising innovative tech lab stuff like this while also hitting all the talking points about application in very Wall Street and marketing inflected language. I don't see any need for such a speculative and immature technology to be forced to genuflect in the direction of the bottom-line. It's so typically Microsoft.

Worse, without having looked at DF's credentials and such, being that he's a "fellow", I assume he's someone with an academic and research background. But he's in the process of being borgified and spouting marketing-speak right before our very eyes.

I think the demise of the high-tech corporate pure research lab in the 80's is a very sad story. I understand the histocrical progression that led to it, and I well understand how much of a crap-shoot it really can be and that many were nothing more than money pits. But some clearly weren't. And, anyway, some companies simply are able to afford such money pits. Microsoft is certainly one of them. More to the point, so lacking in credibility, I think it'd be worth it to them in very indirect but real ways to throw money at their research lab(s) and never worry for one second about direct returns from applications.

Incidentally, I worked at Schlumberger for a short while at the Austin research campus. It had been built during the heyday of the research era, and boy howdy is it a lovely place to work. But it was also a little melancholy, as the portion of the campus which still involved pure research had dimished to a single floor in a single building. Maybe I'm too romantic, but I'm very sorry I missed the era of PARC.
posted by Ethereal Bligh at 11:33 PM on July 30, 2006

If everying MS ever produced was second-hand technology but excellent it would be contradictory to acknowledge this and then assert that what MS produces is crappy

If we were to agree everything produced is "second-hand technology", it would seem contradictory to assert that MS truly produces, on some meaningful level. In my opinion, they are excellent marketers first and (generally mediocre at best, thieving at worst) technologists second.

As with all things, it would be unfair to generalize all their people with this brush, but on the whole, a fair sampling of the history of Microsoft's interactions with technology companies should not be kind to this corporation.
posted by Blazecock Pileon at 11:44 PM on July 30, 2006

"...it would seem contradictory to assert that MS truly produces, on some meaningful level."

I'm very inclined to disagree because I think being innovative is not the larger part of what it takes to "produce". I don't just mean all the other things necessary for sufficient business success that is required for a viable product, but also just all the ways in which the quality of a product is inherently much more greatly determined by things other than innovativeness. For example, in many, many cases the utility of something is much more found in its reliability than in its innovation. Something that Microsoft still doesn't seem to understand, sadly.
posted by Ethereal Bligh at 11:50 PM on July 30, 2006

hey! i fucken' invented this idea! blogged about it ages ago. that gives me some sort of moral copyright, on the (*ahem*, rather tenuous) assumption that nobody in history has ever thought of automatically stitching random strangers' photos together. if this thing ever gets off the ground & starts making money, i'm gonna have to get serious & see if i can remember my livejournal username & password, then give my lawyers a call...

oh, crap. it's gonna be a patent issue, not a copyright one, right...?
posted by UbuRoivas at 3:41 AM on July 31, 2006

EB, What's "borgified" mean? ah.
posted by nickyskye at 11:22 AM on July 31, 2006

I can't believe you guys took a cool thread about an amazing idea and turned it into a overly polite debate arguing the degree to which Microsoft sucks and the type of sucking they do.
posted by Megafly at 5:16 PM on July 31, 2006

It may or may not be amazing, but it certainly isn't anything groundbreaking.
posted by Blazecock Pileon at 11:20 PM on July 31, 2006

What's interesting about this idea isn't the image stitching. As people have pointed out it's been done before - it's called image based modelling and companies like Realviz have been selling products which do this kind of thing for years. There's even an open source package that I used once that does a pretty good job. It's still amazing but not particularly new.
The interesting bit that Microsoft have added here is the idea of taking everyones photos - presumably they would make a website something like Flickr - and automatically attempting to index them spatially. I suspect that the video is being somewhat disingenuous by not mentioning that the images would almost certainly have to be tagged with a decent location marker (such as a GPS coordinate) beforehand. Having a reasonable starting location for an image would narrow the search to manageable size - it then easily within the bounds of existing technology to take all the images within a small area and try to match and stitch them. In a few years cameras with built in GPS will be very common anyway.
The innovation here is more a "mashup" of existing technologies, to produce something quite interesting. But that's fair enough. The "seadragon" software seems to be the main MS innovation - though they apparently bought it in from outside. It seems to be a streaming level-of-detail based image compression system. There are a few of these about too, but presumably Seadragon is more clever in some way.
Of course, after thinking about how cool it looks, one has to start to wonder about how useful a database which spatially indexed everyone's images could be to those whose profit center is the "war on terror". You wouldn't really need to do any of the Photosynth stuff, but simply do face recognition on the millions of images tagged with time and location that you now had in your database. If you'd like to "follow" someone, start with a known time and place where that person was and search the database outwards in time and space till you found him or her again. COOL.
posted by silence at 2:31 AM on August 1, 2006

Why "seadragon"?

Sounds too much like "seamonkey" to me.

Other than that, this looks cool - here's hoping MS won't make it into a product that refuses to work well with other non-MS programs like Sketchup, etc.

...and coyote's bark, you might want to hang around for a bit and get to know some folks before you start going after people for things like an inability to generate a good argument. You kind of picked the wrong person for that. You want to try debating grammar with languagehat next?
posted by caution live frogs at 8:56 AM on August 1, 2006

can you point toward an example of a really good ms product?

Decathlon!
posted by kindall at 10:20 PM on August 15, 2006

Also, to be honest, Excel is really a very good piece of software. You rarely hear people slagging on it the way they do Word and PowerPoint.
posted by kindall at 10:35 PM on August 16, 2006

Whenever I use Excel I think of a poor old man who should have retired years ago, but the company he worked for has denied him his pension and now he's a greeter at a large grocery store chain barely making ends meet and his feet knees hurt all the time now. New technologies like this are like Christmas for this old man: what is a joyous time for everyone else is, for him, a time of shame, pain and burning resentment.
posted by wobh at 4:54 PM on August 19, 2006