Crowdsourcing 3D Motion Reconstruction
Reconstructing 3D motions from 2D video recordings is a useful yet difficult task. In this paper, we present a crowdsourcing system for motion reconstruction. Our insight is that we may outsource motion reconstruction to indefinite users online by allowing the users to contribute their motions with popular motion sensors shipped with game consoles (e.g., Microsoft’s Kinect). The system we prototyped, Motionary, provides functions for requesters to post videos of human motions on our website, and for interested contributors to help reconstruct these motions by performing these motions to our system. The system then records the 3D motions with Kinect during their performance. Our experience shows that the design has the potential to obtain quality motion data without great cost.
KeywordsCrowdsourcing human computation motion reconstruction motion capture
Unable to display preview. Download preview PDF.
- 1.DeFranco, D.E., Cham, T.-J., Rehg, J.M.: Reconstruction of 3-D Figure Motion from 2-D Correspondences. In: Computer Vision and Pattern Recognition, CVPR (2001)Google Scholar
- 3.Gargallo, P., Sturm, P.: Bayesian 3D Modeling from Images Using Multiple Depth Maps. In: IEEE Workshop on Motion and Video Computing, pp. 885–891 (2005)Google Scholar
- 4.Wei, X., Chai, J.: VideoMocap: Modeling Physically Realistic Human Motion from Monocular Video Sequences. In: Proc. of SIGGRAPH (2010)Google Scholar