Cheap, easy Motion capture equipement
Author
Ok, I''m looking for cheap and easy to use Motion capture equipment for my milkshape meshes.
I''ve tried doing animations myself in Milkshape and other programs but find it a huge pain in the bum .. and nothing looks real.
So I thought maybe there''s some cheap motion capture equipment that I couple purchase and save alot of stress!
Any help?
it seems that today intersection(cheap stuff, motion capture equipment, easy to use stuff) = void
Even worse, any intersection of two of the above is void except cheap and easy to use.
You could try rotoscoping, that is filming stuff from one or more points of view, and using that as a model on which to align your animations.
btw, I work with pro animators, and it''s amazing what they can do without mocap!
Even worse, any intersection of two of the above is void except cheap and easy to use.
You could try rotoscoping, that is filming stuff from one or more points of view, and using that as a model on which to align your animations.
btw, I work with pro animators, and it''s amazing what they can do without mocap!
If you really have the money to pay for professional animators, just find good freelance animators, or a company that can provide outsourcing.
I (my company) payed to have graphics made by a company named "Game Consulting" based in Paris. Then again, it will cost you a few hundred euros/dollars per work day. I''m sure you can find such companies through resources like gamasutra. You could even try to find freelance workers on graphics/animation forums.
If you want to go cheaper (at least at first). Try rotoscoping.
I (my company) payed to have graphics made by a company named "Game Consulting" based in Paris. Then again, it will cost you a few hundred euros/dollars per work day. I''m sure you can find such companies through resources like gamasutra. You could even try to find freelance workers on graphics/animation forums.
If you want to go cheaper (at least at first). Try rotoscoping.
I'll let this thread remain in this forum, even though it has nothing or little to do with math/physics, since there really isn't a good forum for it.
I believe House of Moves provides motion capture services. Epic Games (who are located near me here in Raleigh, NC) have a motion capture studio that could possibly be hired. So does Redeye studio. As others have said, all these service providers are going to be expensive for custom data. You may be able to find pre-recorded libraries of animations for cheaper even from the major motion capture service providers.
There may be some cheap sensors that you can find that in theory could be used to build your own motion capture capability. But, to write the software to read the data, stuff it into the correct format, post process into something that could be attached to bones in Milkshape or whatever, clean up bad data points, smooth the data, etc., would be quite a chore. Still, I don't mean to discourage anyone. It may be possible.
Graham Rhodes
Senior Scientist
Applied Research Associates, Inc.
[edited by - grhodes_at_work on February 27, 2004 10:48:12 AM]
I believe House of Moves provides motion capture services. Epic Games (who are located near me here in Raleigh, NC) have a motion capture studio that could possibly be hired. So does Redeye studio. As others have said, all these service providers are going to be expensive for custom data. You may be able to find pre-recorded libraries of animations for cheaper even from the major motion capture service providers.
There may be some cheap sensors that you can find that in theory could be used to build your own motion capture capability. But, to write the software to read the data, stuff it into the correct format, post process into something that could be attached to bones in Milkshape or whatever, clean up bad data points, smooth the data, etc., would be quite a chore. Still, I don't mean to discourage anyone. It may be possible.
Graham Rhodes
Senior Scientist
Applied Research Associates, Inc.
[edited by - grhodes_at_work on February 27, 2004 10:48:12 AM]
OK, something that might drag this thread into being on-topic for this forum:
If you want cheap motion capture, try getting three cheap webcams, and doing some creative coding. Film yourself from three (known) points of view, against a black background, wearing a black catsuit with ping-pong balls glued at the joints. From there, you can take the three videos and process them, using the known camera locations, into a mocap''d stick figure. It''s a relatively straightforward process, but the mathematicians who inhabit this forum will be better qualified than I to give details - especially those who like linear algebra. See? Maths stuff!
We did something similar in Computer Vision year before last (grey mists of time closing in already) but we basically were told ''there is a robust way to do this (2D->3D reconstruction) but since we''ve got a week to cover it we''ll just stick these vectors in a big matrix and invert it, which is numerically unstable but works for our cooked example''. Sadly, I can''t remember the robust way. Something about iterating on a solution.
If you want cheap motion capture, try getting three cheap webcams, and doing some creative coding. Film yourself from three (known) points of view, against a black background, wearing a black catsuit with ping-pong balls glued at the joints. From there, you can take the three videos and process them, using the known camera locations, into a mocap''d stick figure. It''s a relatively straightforward process, but the mathematicians who inhabit this forum will be better qualified than I to give details - especially those who like linear algebra. See? Maths stuff!
We did something similar in Computer Vision year before last (grey mists of time closing in already) but we basically were told ''there is a robust way to do this (2D->3D reconstruction) but since we''ve got a week to cover it we''ll just stick these vectors in a big matrix and invert it, which is numerically unstable but works for our cooked example''. Sadly, I can''t remember the robust way. Something about iterating on a solution.
Good call fractoid! If webcams provide sufficient resolution, then that''d be, say, $50-$100 per camera. Ping-pong balls are essentially free, and since a workable tripod can be had for, say, $30, then perhaps one could assemble an inexpensive, 3-camera home-grown motion capture studio for far less than $500. You may be talking around $200, though you may want to go with a 640x480 resolution webcam and fairly sturdy tripods, which would raise the price a bit. There may even be some freeware photogrammetry software out there, say from a University, that can help locate the points in space. You surely could use Intel''s OpenCV software to locate the ping-pong balls in corresponding frames of each of the videos, then do some quite familiar screen-space to world space transformations based on the known camera positions, orientations, and field-of-view (frustum).
Thinking along those lines, this could be a quite fun little project that could produce satisfactory results.
Things that will be difficult:
- tracking a specific ping-pong ball from frame-to-frame without having to manually mark each ball for each frame.
- dealing with occlusion, e.g., when a ball can only be seen from one camera, or when a ball cannot be seen----for a period of time----from any camera.
- smoothing the data. Because it surely will be noisy.
- converting the data into a standard motion capture file format. I think these are fairly well documented (there is even at least one recent book dedicated to the subject. Maybe a couple. I know Charles River Media has a book that probably covers the subject, and Morgan-Kauffman or Springer---someone else---also has a book that may be better than the CRM one.)
- dealing with the case where two or more balls are both nearly or actually touching in camera space. If the balls are color coded, the camera may see the collection of close proximity balls as a completely different color than the colors it is looking for. Plus, the occlusion issue also comes into play.
But, despite the hard problems, for motions which are clean, I think you could get very nice results without much coding and using very cheap hardware. Very cool idea!
Graham Rhodes
Senior Scientist
Applied Research Associates, Inc.
Thinking along those lines, this could be a quite fun little project that could produce satisfactory results.
Things that will be difficult:
- tracking a specific ping-pong ball from frame-to-frame without having to manually mark each ball for each frame.
- dealing with occlusion, e.g., when a ball can only be seen from one camera, or when a ball cannot be seen----for a period of time----from any camera.
- smoothing the data. Because it surely will be noisy.
- converting the data into a standard motion capture file format. I think these are fairly well documented (there is even at least one recent book dedicated to the subject. Maybe a couple. I know Charles River Media has a book that probably covers the subject, and Morgan-Kauffman or Springer---someone else---also has a book that may be better than the CRM one.)
- dealing with the case where two or more balls are both nearly or actually touching in camera space. If the balls are color coded, the camera may see the collection of close proximity balls as a completely different color than the colors it is looking for. Plus, the occlusion issue also comes into play.
But, despite the hard problems, for motions which are clean, I think you could get very nice results without much coding and using very cheap hardware. Very cool idea!
Graham Rhodes
Senior Scientist
Applied Research Associates, Inc.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
