No Bias Left behind: Covariate Shift Adaptation for Discriminative 3D Pose Estimation

  • Makoto Yamada
  • Leonid Sigal
  • Michalis Raptis
Part of the Lecture Notes in Computer Science book series (LNCS, volume 7575)


Discriminative, or (structured) prediction, methods have proved effective for variety of problems in computer vision; a notable example is 3D monocular pose estimation. All methods to date, however, relied on an assumption that training (source) and test (target) data come from the same underlying joint distribution. In many real cases, including standard datasets, this assumption is flawed. In presence of training set bias, the learning results in a biased model whose performance degrades on the (target) test set. Under the assumption of covariate shift we propose an unsupervised domain adaptation approach to address this problem. The approach takes the form of training instance re-weighting, where the weights are assigned based on the ratio of training and test marginals evaluated at the samples. Learning with the resulting weighted training samples, alleviates the bias in the learned models. We show the efficacy of our approach by proposing weighted variants of Kernel Regression (KR) and Twin Gaussian Processes (TGP). We show that our weighted variants outperform their un-weighted counterparts and improve on the state-of-the-art performance in the public (HumanEva) dataset.


Importance Weight Kernel Regression Gaussian Process Regression Covariate Shift Relative Importance Weight 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


  1. 1.
    Agarwal, A., Triggs, B.: Monocular human motion capture with a mixture of regressors. In: CVPR Workshop (2005)Google Scholar
  2. 2.
    Agarwal, A., Triggs, B.: Recovering 3D human pose from monocular images. IEEE Trans. on PAMI 28, 44–58 (2006)CrossRefGoogle Scholar
  3. 3.
    Bissacco, A., Yang, M., Soatto, S.: Fast human pose estimation using appearance and motion via multi-dimensional boosting regression. In: CVPR, pp. 1–8 (2007)Google Scholar
  4. 4.
    Bo, L., Sminchisescu, C., Kanaujia, A., Metaxas, D.: Fast algorithms for large scale conditional 3d prediction. In: CVPR (2008)Google Scholar
  5. 5.
    Bo, L., Sminchisescu, C.: Twin gaussian processes for structured prediction. Int. J. Comput. Vision 87, 28–52 (2010)CrossRefGoogle Scholar
  6. 6.
    Ek, C.H., Torr, P., Lawrence, N.D.: Gaussian Process Latent Variable Models for Human Pose Estimation. In: Popescu-Belis, A., Renals, S., Bourlard, H. (eds.) MLMI 2007. LNCS, vol. 4892, pp. 132–143. Springer, Heidelberg (2008)CrossRefGoogle Scholar
  7. 7.
    Ionescu, C., Bo, L., Sminchisescu, C.: Structural svm for visual localization and continuous state estimation. In: ICCV (2009)Google Scholar
  8. 8.
    Kanaujia, A., Sminchisescu, C., Metaxas, D.: Semi-supervised hierarchical models for 3d human pose reconstruction. In: CVPR (2007)Google Scholar
  9. 9.
    Navaratnam, R., Fitzgibbon, A., Cipolla, R.: The joint manifold model for semi-supervised multi-valued regression. In: ICCV (2007)Google Scholar
  10. 10.
    Rosales, R., Sclaroff, S.: Learning body pose via specialized maps. In: NIPS (2002)Google Scholar
  11. 11.
    Salzmann, M., Ek, C.H., Urtasun, R., Darrell, T.: Factorized orthogonal latent spaces. In: AISTATS (2010)Google Scholar
  12. 12.
    Sigal, L., Memisevic, R., Fleet, D.: Shared kernel information embedding for discriminative inference. In: CVPR (2009)Google Scholar
  13. 13.
    Shakhnarovich, G., Viola, P., Darrell, T.: Fast pose estimation with parameter-sensitive hashing. In: ICCV, vol. 2, pp. 750–757 (2003)Google Scholar
  14. 14.
    Sminchisescu, C., Kanaujia, A., Li, Z., Metaxas, D.: Discriminative density propagation for 3d human motion estimation. In: CVPR (2005)Google Scholar
  15. 15.
    Sminchisescu, C., Kanaujia, A., Metaxas, D.: Learning joint top-down and bottom-up processes for 3d visual inference. In: CVPR (2006)Google Scholar
  16. 16.
    Urtasun, R., Darrell, T.: Sparse probabilistic regression for activity-independent human pose inference. In: CVPR (2008)Google Scholar
  17. 17.
    Zhao, X., Ning, H., Liu, Y., Huang, T.S.: Discriminative estimation of 3d human pose using gaussian processes. In: CVPR (2008)Google Scholar
  18. 18.
    Aytar, Y., Zisserman, A.: Tabula rasa: Model transfer for object category detection. In: ICCV (2011)Google Scholar
  19. 19.
    Kulis, B., Saenko, K., Darrell, T.: What you saw is not what you get: Domain adaptation using asymmetric kernel transforms. In: CVPR (2011)Google Scholar
  20. 20.
    Saenko, K., Kulis, B., Fritz, M., Darrell, T.: Adapting Visual Category Models to New Domains. In: Daniilidis, K., Maragos, P., Paragios, N. (eds.) ECCV 2010, Part IV. LNCS, vol. 6314, pp. 213–226. Springer, Heidelberg (2010)CrossRefGoogle Scholar
  21. 21.
    Stark, M., Goesele, M., Schiele, B.: A shape-based object class model for knowledge transfer. In: ICCV (2009)Google Scholar
  22. 22.
    Torralba, A., Efros, A.: Ubiased look at dataset bias. In: CVPR (2011)Google Scholar
  23. 23.
    Sigal, L., Black, M.J.: Humaneva: Synchronized video and motion capture dataset for evaluation of articulated human motion. TR CS-06-08, Brown Univ. (2006)Google Scholar
  24. 24.
    Shimodaira, H.: Improving predictive inference under covariate shift by weighting the log-likelihood function. Journal of Statistical Planning and Inference 90 (2000)Google Scholar
  25. 25.
    Sigal, L., Balan, A., Black, M.: Combined discriminative and generative articulated pose and non-rigid shape estimation (2007)Google Scholar
  26. 26.
    de Campos, T., Murray, D.: Regression-based hand pose estimation from multiple cameras. In: CVPR, vol. 1, pp. 782–789 (2006)Google Scholar
  27. 27.
    Rosales, R., Athitsos, V., Sigal, L., Scarloff, S.: 3D hand pose reconstruction using specialized mappings. In: ICCV, vol. 1, pp. 378–385 (2001)Google Scholar
  28. 28.
    Fanelli, G., Gall, J., Gool, L.V.: Real time head pose estimation with random regression forests. In: CVPR (2011)Google Scholar
  29. 29.
    Huang, J., Smola, A.J., Gretton, A., Borgwardt, K.M., Schölkopf, B.: Correcting sample selection bias by unlabeled data. In: NIPS (2007)Google Scholar
  30. 30.
    Sugiyama, M., Nakajima, S., Kashima, H., Buenau, P.V., Kawanabe, M.: Direct importance estimation with model selection and its application to covariate shift adaptation. In: NIPS (2008)Google Scholar
  31. 31.
    Kanamori, T., Hido, S., Sugiyama, M.: A least-squares approach to direct importance estimation. JMLR 10, 1391–1445 (2009)MathSciNetzbMATHGoogle Scholar
  32. 32.
    Yamada, M., Suzuki, T., Kanamori, T., Hachiya, H., Sugiyama, M.: Relative density-ratio estimation for robust distribution comparison. In: NIPS (2011)Google Scholar
  33. 33.
    Cortes, C., Mansour, Y., Mohri, M.: Learning bounds for importance weighting. In: NIPS (2010)Google Scholar
  34. 34.
    Schölkopf, B., Smola, A.J.: Learning with Kernels. MIT Press (2002)Google Scholar
  35. 35.
    Sugiyama, M., Krauledat, M., Müller, K.-R.: Covariate shift adaptation by importance weighted cross validation. JMLR 8, 985–1005 (2007)zbMATHGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2012

Authors and Affiliations

  • Makoto Yamada
    • 1
  • Leonid Sigal
    • 2
  • Michalis Raptis
    • 2
  1. 1.NTT Communication Science LaboratoriesJapan
  2. 2.Disney ResearchPittsburghUSA

Personalised recommendations