Abstract:

Reconstructing an arbitrary configuration of 3D points from their projection in an image is an ill-posed problem. When the points hold semantic meaning, such as anatomical landmarks on a body, human observers can often infer a plausible 3D configuration, drawing on extensive visual memory. We present an activity-independent method to recover the 3D configuration of a human figure from 2D locations of anatomical landmarks in a single image, leveraging a large motion capture corpus as a proxy for visual memory. Our method solves for anthropometrically regular body pose and explicitly estimates the camera via a matching pursuit algorithm operating on the image projections.

Paper (ECCV ’12)    Code 

Figure 2.  Reconstruction of scenes with multiple people with limbs annotated. Consistent relative camera estimates enable a realistic 3D reconstruction of the scene.

Figure 1.  Given the 2D locations of anatomical landmarks in a single image we reconstruct the 3D human pose and relative camera location.

Varun Ramakrishna, Takeo Kanade, Yaser Sheikh

Carnegie Mellon University.

Acknowledgements:

This research was funded (in part) by the Intel Science and Technology Center on Embedded Computing, NSF CRI-0855163, and DARPA's Mind's Eye Program. We also thank Daniel Huber and Tomas Simon for providing valuable feedback on the manuscript.

Citation:

@article{ramakrishna2012reconstructing,

  title={{Reconstructing 3d Human Pose from 2d Image Landmarks}},

  author={Ramakrishna, V. and Kanade, T. and Sheikh, Y.},

  journal={Computer Vision--ECCV 2012},

  pages={573--586},

  year={2012},

  publisher={Springer}

}