Perceiving Systems, Computer Vision

Dyna/4Cap -- SIGGRAPH 2015 -- Supplemental material

15 May 2015

12:01

This video shows all the data used to train our Dyna model of human soft-tissue motion -- all 40,000 meshes. It is not the most thrilling video but there is something remarkable about watching how human soft tissue really moves. We capture raw 3D scans at 60 frames per second using a custom 3dMD active stereo scanner. We then precisely register a common template mesh to all the scans in a process we call 4Cap - for 4D motion capture. This is like having really dense mocap that captures the subtle details of how the body moves. The video shows these registered meshes. Once all the meshes are in correspondence like this, we can model soft-tissue motions in our Dyna model. The data is available for research purposes here http://dyna.is.tue.mpg.de/ The full paper is: Dyna: A Model of Dynamic Human Shape in Motion Pons-Moll, G., Romero, J., Mahmood, N. and Black, M.J. ACM Transactions on Graphics, (Proc. SIGGRAPH), 2015. http://files.is.tue.mpg.de/black/papers/dyna.pdf

=