i seldomly post about nerd programming stuff here but tonight i feel like i want to. I just hit my first milestone on an Animation/Motioncapture tool i am working on for Moviesandbox. The basic idea is simple: You use two regular webcams to film yourself (or a wooden Mannequin in this case) from the front and side view and create keyframes on critical positions in the video.
You then use a very powerful algorithm to find the bones in the two simultaneously recorded videos - human reasoning.
Which basically means that you have to tell the software where the bones actually are in the picture - something that's really easy for the human eye/brain but rather hard for the machine. After finding the bones, the computer does the math (something it's really good at) and sends the rotations back to Unreal - and voila - the pose magically appears!
Note: In future versions, different colors will make it possible for the program to capture most bone rotations, but even then it will be necessary to adjust bones that are covered and not seen by the cameras.
Tuesday, April 10, 2007
You then use a very powerful algorithm to find the bones in the two simultaneously recorded videos - human reasoning.
Which basically means that you have to tell the software where the bones actually are in the picture - something that's really easy for the human eye/brain but rather hard for the machine. After finding the bones, the computer does the math (something it's really good at) and sends the rotations back to Unreal - and voila - the pose magically appears!
Note: In future versions, different colors will make it possible for the program to capture most bone rotations, but even then it will be necessary to adjust bones that are covered and not seen by the cameras.
Tuesday, April 10, 2007
