#Motion Capture
19 messages in this thread
Which is best?
I have literature on Ascension's "Flock of Birds" and Polhemus's "Ultrtrak",
but I am unsure which system is superior in tracking full human movement.
Having a great need to capture some high speed athletic movement, fluidity and
accuracy is a must, thus my major concern is in data capture speed and
accuracy.
As far as using a service…I have talked with BIOVISION and I am confident
that they can provide me with motion data that I can run through Puppeteer, but
I would rather capture it myself.
I would appreciate any criticism or comparison of these two systems that anyone
can send me.
Roy…
Roy,
Last year I ran Flock O Birds with a Onyx Reality Engine set up with
Alias. My impression was fairly simple….Flock OB.s is very noisy and super
sensative to any metal object in the room like rebar in a concrete floor. We
had to build an all-wood 4 foot-high raised platform with minimum nails in a
large room and then suspend the FOB interface boxes overhead above the stage.
The big old magnetic transformer had to sit about six feet from the actor right
on the stage. A real pain!
When I imported the data, the model jumped and jerked and generally
caused a lot of effort wasted in getting stable moves. It seems to work fine
when the movements are real obvious, but it doesn't like it when you stand
still. My model would creep and eventually tip over if I left the system
running over lunch with the sensors on a maniquin.
I sure hope you have the space for this kind of project. Good luck.
As far as Polhemus goes, I havn't run it, but magnets are magnets and
both systems use them. Draw your conclusions from that.
Sanford
Sanford Kennedy Design
PMJI… Sanford, running a Reality Engine with Alias *PHEW* That must have been
interesting. Wonder what the monster ALPHAs with a string of chips will do at
30% the cost???
Jeffrey,
I think the Alpha would do much better. The Alias was actually quite
slow on the SGI.
Sanford
<<alias was slow on SGI>> Interesting! Seeing MAX become compatible on ALPHA
could DO IT… 🙂 Jeffrey
Jeffrey,
Alias is a very able program, but people are misinformed about its
speed. It isn't any faster than 3DS in many cases. It just has more
sophisiticated tools than 3DS4, better particle systems and extensive spline
and patch controls. But, after seeing MAX, I am very hopeful that we will soon
have splines native to the editor, true displacement mapping, and a lot better
control of our models, plus robust particle systems with volumetric and
directional controls. I guess I will have to wait until January to see
…..sigh….
Sanford
Sanford, MAX looks robust. With all the developers out there it should not take
long to become fully functional. I've been with 3-DS for so long I can hardly
imagine not seeing the familiar 4 windows on blue background for *MOST OF MY
WAKING HOURS 6 DAYS/WEEK*. Hell, I won't even complain about the cost of
ADD-ONS. In a year, if MAX does what it looks like it's gonna do, how much can
I sell my "Classic with Tons o' IPAS" for. <G> It's funny, once 4.0 was
released, I decided to dig down a lot deeper in my comprehension of some of the
more esoteric details regarding 3-DS. I even started venturing out into
scripting. With MAX in NT the allure of C++/dll is almost irresistable. I've
always felt that the artist/animator should confine their efforts to that, and
leave the code to the Codemeisters. But MAX makes it sooooooo attractive to
wander in that direction. (just some thoughts) Jeffrey
Jeffrey,
C++ is going to be manditory in the future. I have started studying,
but I am really not a programmer. I will probably keep my 3DS 4 machine intact
because I also have a ton of IPAS's that I don't want to throw away. Well,
that's progress.
Sanford
Sanford, <<ton of IPAS' that I don't want to throw away>>. Know what you mean.
Reasonably, it looks like we will be developing plenty of things in R4 (which
might get upgraded to R5 from what I hear). Also, the spline modeler going out
to all R4 owners in October will be very helpful. I've also heard a bit about
ways 3-DS R4 can be made to run on NT (that includes a special [whatever] so
that NT can read the lock. If R4 and Ani-Pro can run in NT I will be pleased.
This straddle of platforms between R4 and MAX is d-d-d-difficult enough. Your
saying that C++ is going to be manditory in the future sends chills up my spine
(not good ones either <VBG>). With the relative ease of developing routines
within NT, I'm afraid you are right. I've just avoided programming because it
takes me out of the Visual Art Groove and plunges me into the Visual C++ abyss
*admit that it is very seductive in there and hard to come back out*. Time to
start hording my bucks for a super-system (hoarding), once we know all
compatibility issues. Jeffrey
Jeffrey,
The super-system build up is next on my list. Dual 133mhz Pentiums
with 512 megs would be ideal. I checked out the Intergraph computer ….with 6
pentiums (so the hype says). Any idea if MAX will be able to take advantage of
a 6 processor setup, or is two the maximum that NT is optimized for? Would
that be faster than a 300 mhz Dec Alpha? I heard that Autodesk may continue a
DOS version of R4 past the release of MAX that will keep the R4 compatible
stuff alive. Maybe that is a pipe dream….
Sanford
>> Dual 133mhz Pentiums with 512 megs would be ideal. <<
I've heard that most motherboards will not support dual processors beyond
100Mhz. Could you pass on the makers of boards that go beyond this? I have
just started looking, but haven't found any yet.
– Dave
Sanford,
Gary has said that more than 2 processors definitely seems uneconomical,
since the gains diminish rapidly. The 2nd processor gives about a 60% boost,
but things drop off after that. Best to have the flexibility of multiple
2-processor design stations and use network rendering.
Kevin Krell – Computer Support Associates
Sanford, NT can handle the chips, Alphas go up to 12. We have to wait until the
final word regarding MAX and compatibility with which chips and how it
particularly performs with multi-chips. Don't forget, we are very close to the
P6 generation. My prime interest would be to see MAX compatible with ALPHA.
Also curious to see if it will be compatible and fully optimized for the P6.
Clearly the crucial issue when running multi-chips is sufficient RAM. So my
next system is going to cost some serious $$. Would feel most comfortable
seeing those $$ go to an ALPHA, particularly when you consider how well the
ALPHA can be expanded/added on to. But you will also notice in this thread
(Kevin's post) that there is a question of how well MAX performs after the 2nd
chip. Interesting technical issue we will have to hear more about. Regarding
your mentioning R4: I keep hearing about R5 and think that we will be jumping
back and forth for a while. I'm just edgy about putting more money into IPAS'.
This whole IPAS Plug-in thing is *an issue*. Considering how enthusiastically
R4 brought us fully into the Plug-in landscape, we responded in-kind by
sticking thousands of dollars into our IPAS Libraries. Now the light in the MAX
hallway has been flicked on. OKAY – that's progress. But something tells me
there could be a closer relationship between IPAS' that function in R4 and
Routines that will be written for MAX. The result of that 'relationship' would
allow special paths of upgrading. Seems as if a routine that runs in DOS would
need certain re-writing and what not– but isn't there room there for some
help. For instance, if I purchase MeshPaint, and then MeshPaint is released for
MAX — do I have to buy MeshPaint again at full price? Not that I'm singling
out MeshPaint… just a good example since that is exactly the routine I want
next. Actually, based on opinions of other readers, this is a good point to
start seeking some kind of cross-platform policy. I know that GYost said that
they would do everything possible to make the migration of IPAS' reasonable.
Jeffrey
>> Your saying that C++ is going to be manditory in the future sends chills up
my spine <<
I don't think you need to worry about it being "required". I suspect though
that many will develop a strong interest in learning V-C++ because they can
fine-tune MAX and write thier own routines. Similarly, many people using
Renderman write their own custom shaders, and those using Alias write their own
custom routines.
The difference between R4 and MAX is that you won't need to own a
super-high-end- really-expensive-and- hard-to-use compiler to write your own
routines. Also, it sounds like the architecture has really opened up to make
add-ons more closely tied to the base program.
– Dave
Dave, I certainly agree with all you are saying about increased ease of
programming in NT. My problem is **how thinly can one be spread before the head
that supports many hats explodes?** Not enough time, energy, concentration,
brains and <GASP> talent to do too many things well. Besides, the anti-social
misenthrope in me would thrive in V-C++, and just the thought of spending the
next [x]years in a *horizontal to teh floor*feet stretched out*head
inclined*classic programmers official body position*concentrating on code
////somebody stop me \\\\\\ Jeffrey
>>Flock OB.s is very noisy and super sensative to any metal object in the room
like rebar in a concrete floor.<<
Well, I guess I will have to finish that Adobe out back after all. <g>
>> My model would creep and eventually tip over if I left the system running
over lunch with the sensors on a maniquin.<<
I think I saw that on Twilight Zone one night. You just cant trust a woody now
days.
Well…thanks for the info on Flock of Birds. Is this magnetic feild isolation
problem very common with FOB? Can FOB handle high speed movement such as
running in place and jumping back and forth?
Roy Baker….
Roy,
Flock O.B. seems to work great with rapid motion like fighting and
dancing, jumping, etc. The instability in those cases in not noticed.
Sanford
Metrolight was using the Superflou (sp?) system, out if Italy, which uses
infrared camera data, we found that the capturing technique was good, but the
main problem was the software that processed the motion data. It became a
really slow and tedious process. For some reason, the software had problems
compiling all the nodes attached to a person at one time (partial node
compiling worked). When we finally had some data, it was very good. In fact
it was used in the Herby Hancock video as well as a proof of concept project we
just completed. Unfortunately, time is a factor, and the Superflou system did
not meet our time constraints. We are currently looking at other motion
capture technology. If we find one that works well thoughout the process, I
will let you know.
I have been working on motion capture for a project and would be more than
happy to share what I have found. I also have a list on everyone who deals
with motion capture technology in a text file if anyone would like it.
Christopher Ryan
Metrolight Studios
>>I have been working on motion capture for a project and would be more than
happy to share what I have found. I also have a list on everyone who deals
with motion capture technology in a text file if anyone would like it.<<
Please feel free to send me your findings and your text file, It looks like I
will either have to buy a motion capture system or pay someone else for the
capture.
BTW do you think Metrolight Studios would be interested in capturing some
sports moves for me, I would like to get some competitive quotes to compare
against Biovision.
Roy Baker…