Citation
Ben Sapp, Ben Taskar. "MODEC: Multimodal Decomposable Models for Human Pose Estimation". CVPR Computer Vision Conference, June, 2013.

Abstract
The PIs propose a multimodal, decomposable model for articulated human pose estimation in monocular images. A typical approach to this problem is to use a linear structured model, which struggles to capture the wide range of appearance present in realistic, unconstrained images. In this paper, the PIs instead propose a model of human pose that explicitly captures a variety of pose modes. Unlike other multimodal models, the approach includes both global and local pose cues and uses a convex objective and joint training for mode selection and pose estimation. It also employs a cascaded mode selection step which controls the trade-off between speed and accuracy, yielding a 5x speedup in inference and learning. The model outperforms state-of-the-art approaches across the accuracy-speed trade-off curve for several pose datasets. This includes a newly-collected dataset of people in movies, FLIC, which contains an order of magnitude more labeled data for training and testing than existing datasets. The new dataset and code are available online.

Electronic downloads

• http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6619315

• HTML

  Ben Sapp, Ben Taskar. <a
  href="http://www.terraswarm.org/pubs/65.html"
  >MODEC: Multimodal Decomposable Models for Human Pose
  Estimation</a>, CVPR Computer Vision Conference, June,
  2013.

• Plain text

  Ben Sapp, Ben Taskar. "MODEC: Multimodal Decomposable
  Models for Human Pose Estimation". CVPR Computer Vision
  Conference, June, 2013.

• BibTeX

  @inproceedings{SappTaskar13_MODECMultimodalDecomposableModelsForHumanPoseEstimation,
      author = {Ben Sapp and Ben Taskar},
      title = {MODEC: Multimodal Decomposable Models for Human
                Pose Estimation},
      booktitle = {CVPR Computer Vision Conference},
      month = {June},
      year = {2013},
      abstract = {The PIs propose a multimodal, decomposable model
                for articulated human pose estimation in monocular
                images. A typical approach to this problem is to
                use a linear structured model, which struggles to
                capture the wide range of appearance present in
                realistic, unconstrained images. In this paper,
                the PIs instead propose a model of human pose that
                explicitly captures a variety of pose modes.
                Unlike other multimodal models, the approach
                includes both global and local pose cues and uses
                a convex objective and joint training for mode
                selection and pose estimation. It also employs a
                cascaded mode selection step which controls the
                trade-off between speed and accuracy, yielding a
                5x speedup in inference and learning. The model
                outperforms state-of-the-art approaches across the
                accuracy-speed trade-off curve for several pose
                datasets. This includes a newly-collected dataset
                of people in movies, FLIC, which contains an order
                of magnitude more labeled data for training and
                testing than existing datasets. The new dataset
                and code are available online. },
      URL = {http://terraswarm.org/pubs/65.html}
  }
  

Posted by Mila MacBain on 13 May 2013.