Look into finding motion capture data sets, the cameras are set in multiple locations and their position and angle are needed to properly calculate the movements they capture.
The problem there is usually the cameras don't move. I'm actually buying a gopro and very happy this project convinced me to buy one. I probably should have years ago.
Though maybe the multiple positions is enough. It would be cool if I could successfully interpolate in-between positions. Something tells me the method I plan to use won't do that well from three of four positions.
(post is archived)