Our earlier experiments with motion prediction on soft bodies was really just a precursor to a larger goal: Tracking a curtain or sheet as it falls to the ground.
We loaded up the software with a camera and projector in our New Jersey shop and dropped a piece of white fabric. You can see the results here:
You’ll notice, especially in the slow motion clip, that the top edge of the video doesn’t exactly track to the top of the curtain. It follows the general position and doesn’t lag like some of the tracking in the earlier video, but it seems to bounce around within the sheet. Well, by manipulating different pieces of the prediction algorithm we can tweak the performance to get an effect that we like. In this case, we opt for a faster, more-responsive but bouncier effect. Conversely, we could opt for a smoother response by allowing some lag in the tracking when the sheet first begins to fall. We encounter these sorts of tradeoffs frequently when developing tracking systems, and have learned that some of the most important aspects of such development are patience and persistence!