Skip to main content

2016 | OriginalPaper | Buchkapitel

Efficient Continuous Relaxations for Dense CRF

verfasst von : Alban Desmaison, Rudy Bunel, Pushmeet Kohli, Philip H. S. Torr, M. Pawan Kumar

Erschienen in: Computer Vision – ECCV 2016

Verlag: Springer International Publishing

Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.

search-config
loading …

Abstract

Dense conditional random fields (CRF) with Gaussian pairwise potentials have emerged as a popular framework for several computer vision applications such as stereo correspondence and semantic segmentation. By modeling long-range interactions, dense CRFs provide a more detailed labelling compared to their sparse counterparts. Variational inference in these dense models is performed using a filtering-based mean-field algorithm in order to obtain a fully-factorized distribution minimising the Kullback-Leibler divergence to the true distribution. In contrast to the continuous relaxation-based energy minimisation algorithms used for sparse CRFs, the mean-field algorithm fails to provide strong theoretical guarantees on the quality of its solutions. To address this deficiency, we show that it is possible to use the same filtering approach to speed-up the optimisation of several continuous relaxations. Specifically, we solve a convex quadratic programming (QP) relaxation using the efficient Frank-Wolfe algorithm. This also allows us to solve difference-of-convex relaxations via the iterative concave-convex procedure where each iteration requires solving a convex QP. Finally, we develop a novel divide-and-conquer method to compute the subgradients of a linear programming relaxation that provides the best theoretical bounds for energy minimisation. We demonstrate the advantage of continuous relaxations over the widely used mean-field algorithm on publicly available datasets.

Sie haben noch keine Lizenz? Dann Informieren Sie sich jetzt über unsere Produkte:

Springer Professional "Wirtschaft+Technik"

Online-Abonnement

Mit Springer Professional "Wirtschaft+Technik" erhalten Sie Zugriff auf:

  • über 102.000 Bücher
  • über 537 Zeitschriften

aus folgenden Fachgebieten:

  • Automobil + Motoren
  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Elektrotechnik + Elektronik
  • Energie + Nachhaltigkeit
  • Finance + Banking
  • Management + Führung
  • Marketing + Vertrieb
  • Maschinenbau + Werkstoffe
  • Versicherung + Risiko

Jetzt Wissensvorsprung sichern!

Springer Professional "Technik"

Online-Abonnement

Mit Springer Professional "Technik" erhalten Sie Zugriff auf:

  • über 67.000 Bücher
  • über 390 Zeitschriften

aus folgenden Fachgebieten:

  • Automobil + Motoren
  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Elektrotechnik + Elektronik
  • Energie + Nachhaltigkeit
  • Maschinenbau + Werkstoffe




 

Jetzt Wissensvorsprung sichern!

Springer Professional "Wirtschaft"

Online-Abonnement

Mit Springer Professional "Wirtschaft" erhalten Sie Zugriff auf:

  • über 67.000 Bücher
  • über 340 Zeitschriften

aus folgenden Fachgebieten:

  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Finance + Banking
  • Management + Führung
  • Marketing + Vertrieb
  • Versicherung + Risiko




Jetzt Wissensvorsprung sichern!

Anhänge
Nur mit Berechtigung zugänglich
Literatur
1.
Zurück zum Zitat Ravikumar, P., Lafferty, J.: Quadratic programming relaxations for metric labeling and Markov random field MAP estimation. In: ICML (2006) Ravikumar, P., Lafferty, J.: Quadratic programming relaxations for metric labeling and Markov random field MAP estimation. In: ICML (2006)
2.
Zurück zum Zitat Kleinberg, J., Tardos, E.: Approximation algorithms for classification problems with pairwise relationships: metric labeling and Markov random fields. JACM 49, 616–639 (2002)MathSciNetCrossRefMATH Kleinberg, J., Tardos, E.: Approximation algorithms for classification problems with pairwise relationships: metric labeling and Markov random fields. JACM 49, 616–639 (2002)MathSciNetCrossRefMATH
3.
Zurück zum Zitat Kumar, P., Kolmogorov, V., Torr, P.: An analysis of convex relaxations for MAP estimation. In: NIPS (2008) Kumar, P., Kolmogorov, V., Torr, P.: An analysis of convex relaxations for MAP estimation. In: NIPS (2008)
4.
Zurück zum Zitat Chekuri, C., Khanna, S., Naor, J., Zosin, L.: Approximation algorithms for the metric labeling problem via a new linear programming formulation. In: SODA (2001) Chekuri, C., Khanna, S., Naor, J., Zosin, L.: Approximation algorithms for the metric labeling problem via a new linear programming formulation. In: SODA (2001)
5.
Zurück zum Zitat Krähenbühl, P., Koltun, V.: Efficient inference in fully connected CRFs with Gaussian edge potentials. In: NIPS (2011) Krähenbühl, P., Koltun, V.: Efficient inference in fully connected CRFs with Gaussian edge potentials. In: NIPS (2011)
6.
Zurück zum Zitat Tappen, M., Liu, C., Adelson, E., Freeman, W.: Learning Gaussian conditional random fields for low-level vision. In: CVPR (2007) Tappen, M., Liu, C., Adelson, E., Freeman, W.: Learning Gaussian conditional random fields for low-level vision. In: CVPR (2007)
7.
Zurück zum Zitat Koller, D., Friedman, N.: Probabilistic Graphical Models: Principles and Techniques. MIT Press, Cambridge (2009)MATH Koller, D., Friedman, N.: Probabilistic Graphical Models: Principles and Techniques. MIT Press, Cambridge (2009)MATH
8.
Zurück zum Zitat Adams, A., Baek, J., Abraham, M.: Fast high-dimensional filtering using the permutohedral lattice. In: Eurographics (2010) Adams, A., Baek, J., Abraham, M.: Fast high-dimensional filtering using the permutohedral lattice. In: Eurographics (2010)
10.
Zurück zum Zitat Krähenbühl, P., Koltun, V.: Parameter learning and convergent inference for dense random fields. In: ICML (2013) Krähenbühl, P., Koltun, V.: Parameter learning and convergent inference for dense random fields. In: ICML (2013)
11.
Zurück zum Zitat Baqué, P., Bagautdinov, T., Fleuret, F., Fua, P.: Principled parallel mean-field inference for discrete random fields. In: CVPR (2016) Baqué, P., Bagautdinov, T., Fleuret, F., Fua, P.: Principled parallel mean-field inference for discrete random fields. In: CVPR (2016)
12.
Zurück zum Zitat Vineet, V., Warrell, J., Torr, P.: Filter-based mean-field inference for random fields with higher-order terms and product label-spaces. IJCV 110, 290–307 (2014)MathSciNetCrossRefMATH Vineet, V., Warrell, J., Torr, P.: Filter-based mean-field inference for random fields with higher-order terms and product label-spaces. IJCV 110, 290–307 (2014)MathSciNetCrossRefMATH
13.
Zurück zum Zitat Ladicky, L., Russell, C., Kohli, P., Torr, P.H.S.: Graph cut based inference with co-occurrence statistics. In: Daniilidis, K., Maragos, P., Paragios, N. (eds.) ECCV 2010, Part V. LNCS, vol. 6315, pp. 239–253. Springer, Heidelberg (2010)CrossRef Ladicky, L., Russell, C., Kohli, P., Torr, P.H.S.: Graph cut based inference with co-occurrence statistics. In: Daniilidis, K., Maragos, P., Paragios, N. (eds.) ECCV 2010, Part V. LNCS, vol. 6315, pp. 239–253. Springer, Heidelberg (2010)CrossRef
14.
Zurück zum Zitat Kohli, P., Kumar, P., Torr, P.: P3 & beyond: solving energies with higher order cliques. In: CVPR (2007) Kohli, P., Kumar, P., Torr, P.: P3 & beyond: solving energies with higher order cliques. In: CVPR (2007)
15.
Zurück zum Zitat Long, J., Shelhamer, E., Darrell, T.: Fully convolutional networks for semantic segmentation. In: CVPR (2015) Long, J., Shelhamer, E., Darrell, T.: Fully convolutional networks for semantic segmentation. In: CVPR (2015)
16.
Zurück zum Zitat Chen, L., Papandreou, G., Kokkinos, I., Murphy, K., Yuille, A.: Semantic image segmentation with deep convolutional nets and fully connected CRFs. In: ICLR (2015) Chen, L., Papandreou, G., Kokkinos, I., Murphy, K., Yuille, A.: Semantic image segmentation with deep convolutional nets and fully connected CRFs. In: ICLR (2015)
17.
Zurück zum Zitat Schwing, A., Urtasun, R.: Fully connected deep structured networks. CoRR (2015) Schwing, A., Urtasun, R.: Fully connected deep structured networks. CoRR (2015)
18.
Zurück zum Zitat Zheng, S., Jayasumana, S., Romera-Paredes, B., Vineet, V., Su, Z., Du, D., Huang, C., Torr, P.: Conditional random fields as recurrent neural networks. In: ICCV (2015) Zheng, S., Jayasumana, S., Romera-Paredes, B., Vineet, V., Su, Z., Du, D., Huang, C., Torr, P.: Conditional random fields as recurrent neural networks. In: ICCV (2015)
19.
Zurück zum Zitat Zhang, Y., Chen, T.: Efficient inference for fully-connected CRFs with stationarity. In: CVPR (2012) Zhang, Y., Chen, T.: Efficient inference for fully-connected CRFs with stationarity. In: CVPR (2012)
20.
Zurück zum Zitat Wang, P., Shen, C., van den Hengel, A.: Efficient SDP inference for fully-connected CRFs based on low-rank decomposition. In: CVPR (2015) Wang, P., Shen, C., van den Hengel, A.: Efficient SDP inference for fully-connected CRFs based on low-rank decomposition. In: CVPR (2015)
21.
Zurück zum Zitat Goemans, M., Williamson, D.: Improved approximation algorithms for maximum cut and satisfiability problems using semidefinite programming. JACM 42, 1115–1145 (1995)MathSciNetCrossRefMATH Goemans, M., Williamson, D.: Improved approximation algorithms for maximum cut and satisfiability problems using semidefinite programming. JACM 42, 1115–1145 (1995)MathSciNetCrossRefMATH
22.
Zurück zum Zitat Lacoste-Julien, S., Jaggi, M., Schmidt, M., Pletscher, P.: Block-coordinate Frank-Wolfe optimization for structural SVMs. In: ICML (2013) Lacoste-Julien, S., Jaggi, M., Schmidt, M., Pletscher, P.: Block-coordinate Frank-Wolfe optimization for structural SVMs. In: ICML (2013)
23.
Zurück zum Zitat Yuille, A., Rangarajan, A.: The concave-convex procedure (CCCP). In: NIPS (2002) Yuille, A., Rangarajan, A.: The concave-convex procedure (CCCP). In: NIPS (2002)
24.
Zurück zum Zitat Sriperumbudur, B., Lanckriet, G.: On the convergence of the concave-convex procedure. In: NIPS (2009) Sriperumbudur, B., Lanckriet, G.: On the convergence of the concave-convex procedure. In: NIPS (2009)
25.
Zurück zum Zitat Kumar, P., Koller, D.: MAP estimation of semi-metric MRFs via hierarchical graph cuts. In: UAI (2009) Kumar, P., Koller, D.: MAP estimation of semi-metric MRFs via hierarchical graph cuts. In: UAI (2009)
27.
Zurück zum Zitat Scharstein, D., Szeliski, R.: A taxonomy and evaluation of dense two-frame stereo correspondence algorithms. IJCV 47, 7–42 (2002)CrossRefMATH Scharstein, D., Szeliski, R.: A taxonomy and evaluation of dense two-frame stereo correspondence algorithms. IJCV 47, 7–42 (2002)CrossRefMATH
28.
Zurück zum Zitat Everingham, M., Van Gool, L., Williams, C., Winn, J., Zisserman, A.: The PASCAL visual object classes challenge. In: VOC 2010 Results (2010) Everingham, M., Van Gool, L., Williams, C., Winn, J., Zisserman, A.: The PASCAL visual object classes challenge. In: VOC 2010 Results (2010)
29.
Zurück zum Zitat Snoek, J., Larochelle, H., Adams, R.: Practical bayesian optimization of machine learning algorithms. In: NIPS (2012) Snoek, J., Larochelle, H., Adams, R.: Practical bayesian optimization of machine learning algorithms. In: NIPS (2012)
30.
Zurück zum Zitat Boykov, Y., Veksler, O., Zabih, R.: Fast approximate energy minimization via graph cuts. PAMI 23, 1222–1239 (2001)CrossRef Boykov, Y., Veksler, O., Zabih, R.: Fast approximate energy minimization via graph cuts. PAMI 23, 1222–1239 (2001)CrossRef
Metadaten
Titel
Efficient Continuous Relaxations for Dense CRF
verfasst von
Alban Desmaison
Rudy Bunel
Pushmeet Kohli
Philip H. S. Torr
M. Pawan Kumar
Copyright-Jahr
2016
DOI
https://doi.org/10.1007/978-3-319-46475-6_50