🎉 Celebrating 25 Years of GameDev.net! 🎉

Not many can claim 25 years on the Internet! Join us in celebrating this milestone. Learn more about our history, and thank you for being a part of our community!

Affordable Motion Capture?

Started by
11 comments, last by SD2 22 years, 5 months ago
My question is about mocap equipment. Does anyone know of any possible motion capture solutions that are affordable? I have been searching the internet with no luck. It all seems very expensive and I wanted to know if there is any way to gain access to them for under 2000 dollars? I have asked various companies for their prices and none of them were affordable. Does anyone know of any possible options for under 2000 dollars? (second hand? Cheap Renting? etc.) Thanks alot
Advertisement
You can rent a studio with all that technology inside. I remember to got a leaflet with info about such an offer one year back. With 2000$ you are not that far away I guess.

My companies website: www.nielsbauergames.com

A couple of web cameras, a box of ping-pong balls, some glue and a monkey would also do the trick.

Write a bit of code to convert the balls to 3d points, then you get the monkey to go through every frame connecting the points up to form bone data.


Alternatively you could hire time at one of the studios mentioned above. Also for many common motions, some places sell ready made and properly cleaned mocap data on CD so you can do the same for even less.

--
Simon O''''Connor
Creative Asylum Ltd
www.creative-asylum.com

Simon O'Connor | Technical Director (Newcastle) Lockwood Publishing | LinkedIn | Personal site

Another way to do it cheaply would be
(read the bottom two paragraphs first)
a couple of cameras sensitive to infrared light,
a filter that only lets in a certain frequency of infrared
light,
an actor in a suit that absorbs infrared(all black should
do),
and the suit should also have powerful infrared LED''s that
shine at the frequency the filter lets through,
and of course the software to translate the bright points of
light into 3d points.

If you need to scan in a model there are the professional
laser scanners, but another method used to give robots
machine vision(which is how I know) is to set the camera as
above, set a beam to go almost parallel to the camera,
have the beam go through a vertical and horizontal
diffraction grating which will produce a grid of dots,
on a flat surface each dot should be spaced evenly apart.
In other words, the space between the dots determines the shape.
The mechanism for operating this however is more
complicated. The camera/beam/grating assembly should move
upward at a constant pace or the snapshots should be taken
at regular intervals in space. The model would need to
stand still(just like the professional laser scanner) and
the assembly would have to move up and down straight(also
like a professional).

If you use pingpong balls do make sure the actor is in all
black as well as the background, and I think you might need
a camera with better resolution than a web camera.

If you have a scanner, take polaroids and use equipment
like they use in stop motion photography(ie clay-mation).
If you do it yourself, you''ll still need the software to
translate it though.
quote: Original post by Anonymous Poster
Another way to do it cheaply would be
(read the bottom two paragraphs first)
a couple of cameras sensitive to infrared light,
a filter that only lets in a certain frequency of infrared
light,
an actor in a suit that absorbs infrared(all black should
do),
and the suit should also have powerful infrared LED''s that
shine at the frequency the filter lets through,
and of course the software to translate the bright points of
light into 3d points.

If you need to scan in a model there are the professional
laser scanners, but another method used to give robots
machine vision(which is how I know) is to set the camera as
above, set a beam to go almost parallel to the camera,
have the beam go through a vertical and horizontal
diffraction grating which will produce a grid of dots,
on a flat surface each dot should be spaced evenly apart.
In other words, the space between the dots determines the shape.
The mechanism for operating this however is more
complicated. The camera/beam/grating assembly should move
upward at a constant pace or the snapshots should be taken
at regular intervals in space. The model would need to
stand still(just like the professional laser scanner) and
the assembly would have to move up and down straight(also
like a professional).

If you use pingpong balls do make sure the actor is in all
black as well as the background, and I think you might need
a camera with better resolution than a web camera.

If you have a scanner, take polaroids and use equipment
like they use in stop motion photography(ie clay-mation).
If you do it yourself, you''ll still need the software to
translate it though.


Interesting. This type of thing really works? Is it really possible to get proper motion capture results like this? The ping pong ball idea seems to be the exact same thing as the 150 000 dollar optical motion capture they use (the same in theory). Would one actually be able to do these things and get proper results?

quote: Original post by SD2
Interesting. This type of thing really works? Is it really possible to get proper motion capture results like this? The ping pong ball idea seems to be the exact same thing as the 150 000 dollar optical motion capture they use (the same in theory). Would one actually be able to do these things and get proper results?


Same basic idea. The clever thing that the $150,000 system companies have done is to commercialize the process. They developed software to automate (or help) postprocess the raw data, including just visualizing the data as it comes in. They manufactured, or bought, scaffolding to hold the cameras in place so they don''t shake, designed/manufactured or bought high speed data collector devices (aka data acquisition systems), worked deals with camera, light, and ball manufacturers, etc., to bring the rest of the pieces together. They put the package together, do training, do hardware setup, support, etc. Some systems may use 20 or so high-res video cameras (for large rooms, tracking many pointers on multiple characters). If they use 20 miniDV cameras, you''re talking $20,000 for single-CCD cameras alone. They may use 3-CCD cameras (which produce higher-quality images), which would at least double that cost just for the cameras.

The better optical systems do not use ping pong balls. They use smaller balls that appear to have an anisotropic surface. They are "retro-reflective" markers, and they might be easier to identify from a video stream using DSP hardware/software for these commercial, automated systems.

I would guess that the do-it-yourself approach would work, but perhaps not quite as well and certainly not as fast as the commercial system (given that you would probably either write your own software or hire a monkey to postprocess the data). You''d probably need to expect lower-resolution data, with more noise and "bad" data samples.

Anyway, I only meant to describe the value added by commercial systems, not discourage you, . It would be very cool to see you succeed in implementing some level of el-cheapo, home-grown motion capture system. Good luck!


Graham Rhodes
Senior Scientist
Applied Research Associates, Inc.
Graham Rhodes Moderator, Math & Physics forum @ gamedev.net
Another approach you can also do is to make a deal with a company that is having the necessary equipment in your area. If you feel that your game is really gonna be good and that you really want to use MoCap, then you can always try to make a on royalties deal (It works but you have to carefully plan that deal, I did that so I know what Im talkin about).

Its like the company sort of rent you the equipment then you pay them once the game is out or as a co-production project. Note that for co-produced game you have to be ready to give a % of the intellectual property to the company and your game is better be good.

Anyway, I doubt this could help much but that was only my 2 cents here about another possibility of what you can do without too much money.



What is a better quote than this : "blank".
What is a better quote than this : "blank".
You could also buy a CD with mocap data if you don''t need anything too original.

Jack
It would seem your best bet would be to purchase ready-made mocap data from some place like Motek or Turbo Squid.

I work for a company that makes software for motion-capture and 3D animation, and it''s not as simple as it seems. One of your biggest problems with getting data out of such an ersatz system will be that the data will not be clean. It''s very sensitive to things like vibration, etc.

If you can''t find pre-canned mocap data, booking some time at a studio is another option, as others have pointed out. If you plan to do this, make sure you organize your shoot very well before hand. If I''m not mistaken, there is a very good article on Gamasutra that outlines how to do this so that you maximize your time there. Renting a studio with a good Vicon camera system and some decent technicians can become very costly.

R.
_________________________The Idea Foundry
For a cheap alternative just get a digicam and any actor, put your guy in marker clothes and just film him. Then use the footage and animate by hand.

Mocap data usually has to be refined and adapted pretty much, so you might actually be better off not doing mocap. And the most convincing results I´ve seen so far (facial animation) have been done not with mocap but with filmed actors as templates for hand-done animation.

This topic is closed to new replies.

Advertisement