Metric Regression Forests for Correspondence Estimation

Abstract

We present a new method for inferring dense data to model correspondences, focusing on the application of human pose estimation from depth images. Recent work proposed the use of regression forests to quickly predict correspondences between depth pixels and points on a 3D human mesh model. That work, however, used a proxy forest training objective based on the classification of depth pixels to body parts. In contrast, we introduce Metric Space Information Gain (MSIG), a new decision forest training objective designed to directly minimize the entropy of distributions in a metric space. When applied to a model surface, viewed as a metric space defined by geodesic distances, MSIG aims to minimize image-to-model correspondence uncertainty. A naïve implementation of MSIG would scale quadratically with the number of training examples. As this is intractable for large datasets, we propose a method to compute MSIG in linear time. Our method is a principled generalization of the proxy classification objective, and does not require an extrinsic isometric embedding of the model surface in Euclidean space. Our experiments demonstrate that this leads to correspondences that are considerably more accurate than state of the art, using far fewer training images.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10

Notes

  1. 1.

    Note that this is an extended version of Pons-Moll et al. (2013). Some portions of Taylor et al. (2012) have been included for clarity.

  2. 2.

    Distinct subscripts indicate whether \(p\) and \(l\) refer to vertices or spheres.

References

  1. Baak, A., Müller, M., Bharaj, G., Seidel, H., & Theobalt, C. (2011). A data-driven approach for real-time full body pose reconstruction from a depth camera. In: IEEE international conference on computer vision pp. 1092–1099.

  2. Balan, A., Sigal, L., Black, M., Davis, J., & Haussecker, H. (2007). Detailed human shape and pose from images. In: IEEE conference on computer vision and pattern recognition.

  3. Bentley, J. (1975). Multidimensional binary search trees used for associative searching. Communications of the ACM, 18(9), 509–517.

    MATH  MathSciNet  Article  Google Scholar 

  4. Besl, P., & McKay, N. (1992). A method for registration of 3d shapes. IEEE Transactions on Pattern Analysis and Machine Intelligence, 12, 239–256.

    Article  Google Scholar 

  5. Black, M., & Rangarajan, A. (1996). On the unification of line processes, outlier rejection, and robust statistics with applications in early vision. International Journal on Computer Vision, 19(1), 57–91.

    Article  Google Scholar 

  6. Bo, L., & Sminchisescu, C. (2010). Twin gaussian processes for structured prediction. International Journal on Computer Vision, 87, 28–52.

    Article  Google Scholar 

  7. Bregler, C., Malik, J., & Pullen, K. (2004). Twist based acquisition and tracking of animal and human kinematics. International Journal on Computer Vision, 56(3), 179–194.

    Article  Google Scholar 

  8. Breiman, L. (1999). Random forests. Berkeley: UC. (Technical Report TR567).

  9. Brubaker, M., Fleet, D., & Hertzmann, A. (2010). Physics-based person tracking using the anthropomorphic walker. In: International journal on computer vision.

  10. Buntine, W., & Niblett, T. (1992). A further comparison of splitting rules for decision-tree induction. Machine Learning, 8(1), 75–85.

    Google Scholar 

  11. Criminisi, A., & Shotton, J. (2013). Decision forests for computer vision and medical image analysis. London: Springer.

    Google Scholar 

  12. Deutscher, J., & Reid, I. (2005). Articulated body motion capture by stochastic search. International Journal on Computer Vision, 61(2), 185–205.

    Article  Google Scholar 

  13. Dijkstra, E. W. (1959). A note on two problems in connexion with graphs. Numerische Mathematik, 1(1), 269–271.

    MATH  MathSciNet  Article  Google Scholar 

  14. Gall, J., Rosenhahn, B., Brox, T., & Seidel, H. P. (2010). Optimization and filtering for human motion capture. International Journal on Computer Vision, 87, 75–92.

    Article  Google Scholar 

  15. Gall, J., Yao, A., Razavi, N., Van Gool, L., & Lempitsky, V. (2011). Hough forests for object detection, tracking, and action recognition. PAMI, 33(11), 2188–2202.

    Article  Google Scholar 

  16. Ganapathi, V., Plagemann, C., Koller, D., & Thrun, S. (2012). Real-time human pose tracking from range data. In: European conference on computer vision.

  17. Ganapathi, V., Plagemann, C., Thrun, S., & Koller, D. (2010). Real time motion capture using a time-of-flight camera. In: Conference in computer vision and pattern recognition.

  18. Girshick, R., Shotton, J., Kohli, P., Criminisi, A., & Fitzgibbon, A. (2011). Efficient regression of general-activity human poses from depth images. In: IEEE international conference on computer vision, pp. 415–422.

  19. Kabsch, W. (1976). A solution for the best rotation to relate two sets of vectors. Acta Crystallographica, 32(5), 922–923.

    Article  Google Scholar 

  20. Lee, C., & Elgammal, A. (2010). Coupled visual and kinematic manifold models for tracking. International Journal on Computer Vision, 87, 118–139.

  21. Liu, W., & White, A. (1994). The importance of attribute selection measures in decision tree induction. Machine Learning, 15(1), 25–41.

    Google Scholar 

  22. Memisevic, R., Sigal, L., & Fleet, D. J. (2012). Shared kernel information embedding for discriminative inference. PAMI, 34(4), 778–790.

    Article  Google Scholar 

  23. Nowozin, S. (2012). Improved information gain estimates for decision tree induction. In: ICML.

  24. Parzen, E. (1962). On estimation of a probability density function and mode. The Aannals of Mathematical Statistics, 33(3), 1065–1076.

    MATH  MathSciNet  Article  Google Scholar 

  25. Pons-Moll, G., Baak, A., Gall, J., Leal-Taixe, L., Mueller, M., Seidel, H., & Rosenhahn, B. (2011). Outdoor human motion capture using inverse kinematics and von mises-fisher sampling. In: International conference on computer vision.

  26. Pons-Moll, G., Leal-Taixé, L., Truong, T., & Rosenhahn, B. (2011). Efficient and robust shape matching for model based human motion capture. In: DAGM.

  27. Pons-Moll, G., & Rosenhahn, B. (2011). Model-based pose estimation. In Visual analysis of humans (pp. 139–170). London: Springer.

  28. Pons-Moll, G., Taylor, J., Shotton, J., Hertzmann, A., & Fitzgibbon, A. (2013). Metric regression forests for human pose estimation. In: British machine vision conference (BMVC).

  29. Shotton, J., Fitzgibbon, A., Cook, M., Sharp, T., Finocchio, M., Moore, R., Kipman, A., & Blake, A. (2011). Real-time human pose recognition in parts from single depth images. In: IEEE conference in computer vision and pattern recognition, pp. 1297–1304.

  30. Shotton, J., Glocker, B., Zach, C., Izadi, S., Criminisi, A., & Fitzgibbon, A. (2013). Scene coordinate regression forests for camera relocalization in RGB-D images. In: Conference in computer vision and pattern recognition.

  31. Silverman, B. (1986). Density estimation for statistics and data analysis (Vol. 26). London: CRC press.

  32. Sminchisescu, C., Bo, L., Ionescu, C., & Kanaujia, A. (2011). Feature-based pose estimation. In Visual analysis of humans (pp. 225–251). London: Springer.

  33. Stoll, C., Hasler, N., Gall, J., Seidel, H., & Theobalt, C. (2011) Fast articulated motion tracking using a sums of gaussians body model. In: IEEE international conference on computer vision, pp. 951–958.

  34. Taylor, J., Shotton, J., Sharp, T., & Fitzgibbon, A. (2012). The Vitruvian manifold: Inferring dense correspondences for one-shot human pose estimation. In: Conference in computer vision and pattern recognition.

  35. Urtasun, R., & Darrell, T. (2008). Sparse probabilistic regression for activity-independent human pose inference. In: IEEE conference in computer vision and pattern recognition, pp. 1–8.

Download references

Author information

Affiliations

Authors

Corresponding author

Correspondence to Gerard Pons-Moll.

Additional information

Communicated by Tilo Burghardt, Majid Mirmehdi, Walterio Mayol-Cuevas, and Dima Damen.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Pons-Moll, G., Taylor, J., Shotton, J. et al. Metric Regression Forests for Correspondence Estimation. Int J Comput Vis 113, 163–175 (2015). https://doi.org/10.1007/s11263-015-0818-9

Download citation

Keywords

  • Human pose estimation
  • Model based pose estimation
  • Correspondence estimation
  • Depth images
  • Metric regression forests