Slashdot Mirror
Motion Capturing in Three Dimensions?
brokenbeaker asks: "I friend of mine is a modern dancer, and she is working with a 3D modelling program to record dance performances. Apparently, the system is very primitive - if you make a dancer jump, she doesn't fall back down to earth. My question is, does anyone know of a sophisticated (physics based, I guess) modelling program that can be used to record dance? How about motion capture in 3D - would this be feasible? Anyone looking for a programming challenge?"
The things you talk about seem to have more to do with composing dance than recording it...
The formal composition of dance is a field that is hardly developed, most of dance is mixtures of repartorie learned one on one through generations with no formal theory or notation behind it. That means you're not gonna have very good luck finding any computerized tools for dance composition, AFAICT.
...and this lie crawls out of its mouth: 'I, the state, am the people.'
I guess as this is the movie industry, such stuff would not be cheap and probably require a beowolf cluster or two to run it.
What software is she using?
I know that Maya has a pretty complex modeler and it has solvers that allow for all sorts of stuff.
You can do IK solvers for a body and you can also add influences to states, such as gravity and wind.
I created a simple marble toy using the system. It would drop a marble and launch another one off of a see-saw and then start rolling down a track.
If she really wanted realism, she should look into motion capture, as other have suggested. It doesn't really take that complicated of mocap setup. Just patience and time at the end to tweak the animation keys.
-Tim
-I just work here... how am I supposed to know?
It is difficult to appreciate just how difficult this problem is. In terms of Artificial Intelligence, we don't even really have the technology for stereoscopic vision yet, much less three dimensional modeling from visual input. For video games and movies, I believe the state of the art is to strap a bunch of specially-colored objects all over the person's body and put them in extremely bright light, then try to reconstruct the body's kinematics from the movements seen in various cameras around the room.
Otherwise, I would do a search for "visual hulls". You'll get some inforfmation about trying to define surfaces given multiple pictures of the same object. Still a difficult problem - getting correlated points is the hard part.
In any case, good luck - you'll need it.
Nonperiodic Central Trajectory
For hand-editing, the character animation software in Maya may be a good place to start. Or, look at any of the "bipedal toolkits" for other 3D animation packages. There are packages designed specifically for facilitating hand-animation of humanoid figures. Still, you're talking several grand and a relatively beefy computer for this.
I believe that AnimationMaster has basic support for this stuff as well, and it can be had as cheaply as $299. Possibly worth some extra investigation there.
Says the RIAA: When you EQ, you're stealing bass!
Although I cannot find a specific-cool link at their site, the Human Performance Lab in the Department of Kinesthetics has been doing some mightly cool motion capture and analysis of atheletes, normal people and people with physical limitations for years. Very cool Sun based motion tracking system. My climbing partner in Uni used to be the technician for the capture / analysis systems. They were SPARC systems at the time, from sun.
As other posters have mentioned, motion capture and motion synthesis tend to be very different problems, although in an end product (if it is a rendering) may contain elements of both.
Alias' tools still tend to be some of the better products out there for synthesis. If you are serious about capture, I know that the U of C department would at least be a starting place for software sources. I know that it wasn't an inexpensive setup.
You can check it out here. Found an article off of google about them now giving you a free 30 day full trial version - so at least you can check it out and see.
As I recall there were a bunch of pre built dance moves in there for choreographers - much easier to start with something then to make everything from scratch. I could keep going on about this but it's just much better to go to there site and see all the features.
Hope this will be helpful. Unless of course this was the program that she was already using.
That is your ass, and this over here is your elbow, and NO they ARE NOT the same thing.
Another similar tool is the actual real stick figures which are about a foot high and feed into a modelling program. More of a frame by frame too, these have been used for a bunch of 3d fighting games.
Also if the only goal is to make people return to earth, there is a slashdotter's company Animats.
I got to see a professional motion capture system in person. The reflective dots they used were to reflect infrared light. The dots were placed at a strategic locations, like at joints and whatnot. Surrounding the room were the infrared video cameras mounted with circular infrared LED arrays to flood the room with infrared light. Basically only the reflective pads would reflect the a signifigant amount of it, meaning that the cameras could easily tracked three dimesionally. This doesn't require that the dancer wear any special kind of black suit or background like in other motion tracking systems which relied on white light reflected from the pads.
Of course, this system is just as expensive as the others (on the order of several hundred thousand dollars), unless you created your own from scratch. The LED arrays are cheap, the cameras though I'm guessing would be expensive (probably where most of the cost comes from). With a basic knowledge of DSP I'd assume it wouldn't take too long to write your own program to track the movement of the infrared dots.
I saw a film from the 70s of a really old ghetto-style motion capture system that anyone could use. Basically the sports trainer in the film (he was using it to capture the performance of an athlete) would go through the video frame-by-frame and mark the joints with a mouse (the frame was projected onto the mousepad iirc). After this, the guy got a high-tech wireframe rendering of the athlete throwing a discus (hey, 70s computers weren't that fast). This is tedious, and time consuming, but it works. This is probably the most feasible option, as it's mostly done in software.
I also saw another ghetto style method, but it only would work for two dimensional motion in the way it was setup (which might still be useful). They strapped light bulbs to the joints of a person with a black suit. A camera with a slow capture speed was used to capture the person walking. Because of the slow capture, there were streaks on the photograph showing the motion as it progessed. I suppose with multiple slow shutter cameras setup it could work three dimensionally. This wasn't actually sent to a computer, but I suppose you could write software to analyze the image.
Of cource it's possible to track in 3D. Games use motion capture all the time. They use a special suit equipped with infrared lights on them. They have cameras setup in special places and the computer records the moements and them converts it into any format -- including what 3DSMax can use.
Look into that.
Good news: Your friend doesn't really need them.
The key insight is the realization that the issue is not one of technology per se but of conceptualization. In other words, you don't need a Beowulf cluster, you need a notation system for representing dance movement, and tools for scoring/editing/viewing sets of such data.
Fortunately, there is such a system: LabanWriter. It's even free as in beer. Here's a bit from the project web site:
As far as I know, the source is closed and the binaries are Mac only, but that may not be so terrible. Given the academic background of the project, source might be available for a someone looking to do a port.FWIW, there are several other similar projects linked off of the main page. I'd be surprised if somewhere in all of that there weren't an answer for you.
There was a package developed by Don Herbison-Evans at the Univesity of Sydney (Australia) some time ago called NUDES that did exactly this. Have a look at Dance and the Computer: A Potential for Graphic Synergy.
Try a Google search for "nudes" "dance" "software" "university".
Graham
Graham