Active Deformable Part Models Inference
Menglong Zhu, Nikolay A. Atanasov, George Pappas, Kostas Daniilidis

Citation
Menglong Zhu, Nikolay A. Atanasov, George Pappas, Kostas Daniilidis. "Active Deformable Part Models Inference". European Conference on Computer Vision (ECCV), Vol. 8695, p.281-296, 6, September, 2014.

Abstract
This paper presents an active approach for part-based object detection, which optimizes the order of part filter evaluations and the time at which to stop and make a prediction. Statistics, describing the part responses, are learned from training data and are used to formalize the part scheduling problem as an offline optimization. Dynamic programming is applied to obtain a policy, which balances the number of part evaluations with the classication accuracy. During inference, the policy is used as a look-up table to choose the part order and the stopping time based on the observed lter responses. The method is faster than cascade detection with deformable part models (which does not optimize the part order) with negligible loss in accuracy when evaluated on the PASCAL VOC 2007 and 2010 datasets.

Electronic downloads

Internal. This publication has been marked by the author for TerraSwarm-only distribution, so electronic downloads are not available without logging in.

Citation formats

• HTML

  Menglong Zhu, Nikolay A. Atanasov, George Pappas, Kostas
  Daniilidis. <a
  href="http://www.terraswarm.org/pubs/487.html"
  >Active Deformable Part Models Inference</a>,
  European Conference on Computer Vision (ECCV), Vol. 8695,
  p.281-296, 6, September, 2014.

• Plain text

  Menglong Zhu, Nikolay A. Atanasov, George Pappas, Kostas
  Daniilidis. "Active Deformable Part Models
  Inference". European Conference on Computer Vision
  (ECCV), Vol. 8695, p.281-296, 6, September, 2014.

• BibTeX

  @inproceedings{ZhuAtanasovPappasDaniilidis14_ActiveDeformablePartModelsInference,
      author = {Menglong Zhu and Nikolay A. Atanasov and George
                Pappas and Kostas Daniilidis},
      title = {Active Deformable Part Models Inference},
      booktitle = {European Conference on Computer Vision (ECCV)},
      pages = {Vol. 8695, p.281-296},
      day = {6},
      month = {September},
      year = {2014},
      abstract = {This paper presents an active approach for
                part-based object detection, which optimizes the
                order of part filter evaluations and the time at
                which to stop and make a prediction. Statistics,
                describing the part responses, are learned from
                training data and are used to formalize the part
                scheduling problem as an offline optimization.
                Dynamic programming is applied to obtain a policy,
                which balances the number of part evaluations with
                the classication accuracy. During inference, the
                policy is used as a look-up table to choose the
                part order and the stopping time based on the
                observed lter responses. The method is faster
                than cascade detection with deformable part models
                (which does not optimize the part order) with
                negligible loss in accuracy when evaluated on the
                PASCAL VOC 2007 and 2010 datasets.},
      URL = {http://terraswarm.org/pubs/487.html}
  }
  

Posted by Barb Hoversten on 5 Feb 2015.
Groups: services