The video shows results from a project using machine learning to recover 3D articulated skeletons from tracked 2D features points.
This work is detailed in the paper "Unsupervised learning of skeletons from motion" by David Ross, Daniel Tarlow, and Richard Zemel, in the proceedings of the 10th European Conference on Computer Vision (ECCV 2008), 2008.
For more information visit the website: http://www.cs.toronto.edu/~dross/articulated/
Impressive!
omerjacobus 1 year ago