(care to post the math used to turn the 4 corner coordinates into 3D coords for the perspective projection? I've been fighting with that for a year and a half now...)
Wow thats a great demonstration, captures the imagination.
On a vaguely related note I tuned into CNN a few times during US election season and saw them mucking around with graphics that were attached to physical objects in the studio. It wasnt without its hitches and I dont think they had an actual purpose for doing it, but it made me chuckle to see someone desperately waving a bit of card around trying to get a piechart to sprout from it. Your demo was better :)
Yes...but I think I know the answer... which is, yes, if you want it to track well.
And I admit, I am sitting here messing with the CV and the MSA patch and still scratching my head a little in how you are hooking them up... but then again, I am only 5 minutes into it :o)
Really impressive. I saw that clip of the interactive reality stuff from that University that was making the rounds this last week, and immediately thought of QC and Kineme. It is awesome that you have it working somehow in this form.
The motion tracking algorithm currently used tracks brightness, so the marks on the corners helps the algorithm find where the corners move (more or less -- towards the end, the bottom right corner starts to slide a bit) -- There are alternative algorithms that could possibly get this to work without needing the marks. For example, it would track the card corners compared against the background (possibly with chroma data). OpenCV might or might not be able to do this, I'm not sure. But there's a ton of stuff in the CV field that I'm just beginning to figure out myself.
(As an adaptation on this method, I've wondered if having a tight cluster of points at each corner, and having them self-correct by keeping points within a certain radius, would keep the sliding problem at bay... lots of work for little gain though.)
the most bothering issue is the fact that KnM CV can't toss out points, so if you ever loose your tracking point, you're screwed (try to rotate your card too fast ...)
No, that is AR, a slightly different technology/concept. And it already exists for QC (if a bit lacking in full feature set). Maybe I'm wrong in definition on that though... I thought dynamic surface projection denotes "2D", maybe not!
Do a search, I've posted files and a compiled plugin.
Daniele, that's amazing! :)
(care to post the math used to turn the 4 corner coordinates into 3D coords for the perspective projection? I've been fighting with that for a year and a half now...)
i assume the corner pinning comes from memo ... and then it seems to be cleverly linked to KnM CV tracker... awesome
I have insanely modified the memo patch "msa Quad Warp"...
Wow thats a great demonstration, captures the imagination.
On a vaguely related note I tuned into CNN a few times during US election season and saw them mucking around with graphics that were attached to physical objects in the studio. It wasnt without its hitches and I dont think they had an actual purpose for doing it, but it made me chuckle to see someone desperately waving a bit of card around trying to get a piechart to sprout from it. Your demo was better :)
That's amazing!
Augmented reality seems to be becoming more and more commonplace. I just watched a similar video of it used in an architectural presentation.
... Minority Report?
a|x
Do you find that the crop marks that you drew on the paper are a must?
you mean if they are necessary?
Yes...but I think I know the answer... which is, yes, if you want it to track well.
And I admit, I am sitting here messing with the CV and the MSA patch and still scratching my head a little in how you are hooking them up... but then again, I am only 5 minutes into it :o)
Really impressive. I saw that clip of the interactive reality stuff from that University that was making the rounds this last week, and immediately thought of QC and Kineme. It is awesome that you have it working somehow in this form.
The motion tracking algorithm currently used tracks brightness, so the marks on the corners helps the algorithm find where the corners move (more or less -- towards the end, the bottom right corner starts to slide a bit) -- There are alternative algorithms that could possibly get this to work without needing the marks. For example, it would track the card corners compared against the background (possibly with chroma data). OpenCV might or might not be able to do this, I'm not sure. But there's a ton of stuff in the CV field that I'm just beginning to figure out myself.
(As an adaptation on this method, I've wondered if having a tight cluster of points at each corner, and having them self-correct by keeping points within a certain radius, would keep the sliding problem at bay... lots of work for little gain though.)
the most bothering issue is the fact that KnM CV can't toss out points, so if you ever loose your tracking point, you're screwed (try to rotate your card too fast ...)
source here... http://www.thefelicelab.com/ed/index.php?option=com_content&task=view&id...
ciao ciao
.... Minority Report!!!!
http://vimeo.com/2229299
I wonder which came first: the film of the product?
a|x
That's so cool looking...
Wasn't there some kind of Nintendo or Sega glove like this, back in the late 80's or early 90's?
John Underkoffler, the guy behind this, was the science and tech adviser for steven spielberg on the movie!
A flash implementation. http://ge.ecomagination.com/smartgrid/#/augmented_reality
No, that is AR, a slightly different technology/concept. And it already exists for QC (if a bit lacking in full feature set). Maybe I'm wrong in definition on that though... I thought dynamic surface projection denotes "2D", maybe not!
Do a search, I've posted files and a compiled plugin.
Ugh, I am so disgruntled after looking at that :o) I'm going to have to go back to working with that more.
I suppose you are right, I never really thought about it but the image isn't being rotated in space, only warped...
Bravo, that's awesome.