Presentation is loading. Please wait.

Presentation is loading. Please wait.

Rounding-based Moves for Metric Labeling M. Pawan Kumar École Centrale Paris INRIA Saclay, Île-de-France.

Published byOsborne Flowers Modified over 8 years ago

Similar presentations

Presentation on theme: "Rounding-based Moves for Metric Labeling M. Pawan Kumar École Centrale Paris INRIA Saclay, Île-de-France."— Presentation transcript:

1 Rounding-based Moves for Metric Labeling M. Pawan Kumar École Centrale Paris INRIA Saclay, Île-de-France

2 Metric Labeling Variables V = { V 1, V 2, …, V n }

3 Metric Labeling Variables V = { V 1, V 2, …, V n }

4 Metric Labeling VaVa VbVb Labels L = { l 1, l 2, …, l h } Variables V = { V 1, V 2, …, V n } Labeling f: { 1, 2, …, n}  {1, 2, …, h} E(f) = Σ a θ a (f(a)) + Σ (a,b) w ab d(f(a),f(b)) min f θ a (f(a)) θ b (f(b)) w ab d(f(a),f(b)) w ab ≥ 0 d is metric

5 Metric Labeling VaVa VbVb E(f) min f NP hard = Σ a θ a (f(a)) + Σ (a,b) w ab d(f(a),f(b)) Low-level vision applications

6 Outline Approximate Algorithms Comparison Rounding-based Moves

7

8 Boykov, Veksler and Zabih Kleinberg and Tardos Efficiency Accuracy Move-Making Algorithms Convex Relaxations

9 Kolmogorov and Boykov Move-Making Algorithms Convex Relaxations Chekuri, Khanna, Naor and Zosin Efficiency Accuracy

10 Outline Approximate Algorithms –Move-Making Algorithms –Linear Programming Relaxation Comparison Rounding-based Moves

11 Move-Making Algorithms Space of All Labelings f

12 Expansion Algorithm Variables take label l α or retain current label Slide courtesy Pushmeet Kohli Boykov, Veksler and Zabih, 2001

13 Expansion Algorithm Sky House Tree Ground Initialize with TreeStatus:Expand GroundExpand HouseExpand Sky Slide courtesy Pushmeet Kohli Variables take label l α or retain current label Boykov, Veksler and Zabih, 2001

14 Multiplicative Bounds f*: Optimal Labelingf: Estimated Labeling Σ a θ a (f(a)) + Σ (a,b) w ab d(f(a),f(b)) Σ a θ a (f*(a)) + Σ (a,b) w ab d(f*(a),f*(b)) ≥

15 Multiplicative Bounds f*: Optimal Labelingf: Estimated Labeling ≤ B Σ a θ a (f(a)) + Σ (a,b) w ab d(f(a),f(b)) Σ a θ a (f*(a)) + Σ (a,b) w ab d(f*(a),f*(b)) Ask me the obvious question

16 Outline Approximate Algorithms –Move-Making Algorithms –Linear Programming Relaxation Comparison Rounding-based Moves

17 Integer Linear Program Number of facets grows exponentially in problem size Minimize a linear function over a set of feasible solutions Indicator x a (i)  {0,1} for each variable V a and label l i Indicator x ab (i,k)  {0,1} for each neighbor (V a,V b ) and labels l i, l k

18 Linear Programming Relaxation Schlesinger, 1976; Chekuri et al., 2001; Wainwright et al., 2003 Indicator x a (i)  {0,1} for each variable V a and label l i Indicator x ab (i,k)  {0,1} for each neighbor (V a,V b ) and labels l i, l k

19 Linear Programming Relaxation Schlesinger, 1976; Chekuri et al., 2001; Wainwright et al., 2003 Indicator x a (i)  [0,1] for each variable V a and label l i Indicator x ab (i,k)  [0,1] for each neighbor (V a,V b ) and labels l i, l k

20 Approximation Factor x*: LP Optimal Solutionx: Estimated Integral Solution Σ a Σ i θ a (i)x a (i) + Σ (a,b) Σ (i,k) w ab d(i,k)x ab (i,k) ≥ Σ a Σ i θ a (i)x* a (i) + Σ (a,b) Σ (i,k) w ab d(i,k)x* ab (i,k)

21 Approximation Factor x*: LP Optimal Solutionx: Estimated Integral Solution Σ a Σ i θ a (i)x a (i) + Σ (a,b) Σ (i,k) w ab d(i,k)x ab (i,k) ≤ Σ a Σ i θ a (i)x* a (i) + Σ (a,b) Σ (i,k) w ab d(i,k)x* ab (i,k) F

22 Outline Approximate Algorithms Comparison Rounding-based Moves

23 Theoretical Guarantees ExpansionLP Uniform22 Metric2MO(log h) Truncated Linear 2M2 + √2 Truncated Quadratic 2MO(√M) M = ratio of maximum and minimum non-zero distance

24 Outline Approximate Algorithms Comparison Rounding-based Moves –Complete Rounding –Interval Rounding –Hierarchical Rounding

25 Complete Rounding Treat x a (i)  [0,1] as probability that f(a) = i Cumulative probability y a (i) = Σ j≤i x a (j) 0y a (1) y a (2) y a (h) = 1 y a (k) y a (i) Generate a random number r  (0,1] Assign the label next to r r

26 Example 0y a (1) y a (4) y a (3) y a (2) 0.25 0.5 0.75 1.0 0 y b (1) y b (4) y b (3) y b (2) 0.70.80.91.0 0 y c (1) y c (4) y c (3) y c (2) 0.1 0.20.3 1.0 r r r

27 Complete Move A move that mimics complete rounding Considers all random variables and labels Assigns labels in one iteration

28 Key Observation If d is submodular d(i,k) + d(i+1,k+1) ≤ d(i,k+1) + d(i+1,k), for all i, k Schlesinger and Flach, 2003 energy can be minimized via minimum cut

29 Complete Move VaVa VbVb θ ab (i,k) = w ab d(i,k)NP-hard

30 Complete Move VaVa VbVb θ ab (i,k) = w ab d’(i,k) d’(i,k) ≥ d(i,k) d’ is submodular

31 Complete Move VaVa VbVb θ ab (i,k) = w ab d’(i,k) d’(i,k) ≥ d(i,k) d’ is submodular

32 Complete Move New problem can be solved using minimum cut Same multiplicative bound as complete rounding Multiplicative bound is tight

33 Outline Approximate Algorithms Comparison Rounding-based Moves –Complete Rounding –Interval Rounding –Hierarchical Rounding

34 Interval Rounding Treat x a (i)  [0,1] as probability that f(a) = i Cumulative probability y a (i) = Σ j≤i x a (j) 0y a (1) y a (2) y a (h) = 1 y a (k) y a (i) Choose an interval of length h’

35 Interval Rounding Treat x a (i)  [0,1] as probability that f(a) = i Cumulative probability y a (i) = Σ j≤i x a (j) r Generate a random number r  (0,1] Assign the label next to r if it is within the interval y a (k)-y a (i) 0 Choose an interval of length h’ REPEAT

36 Example 0y a (1) y a (4) y a (3) y a (2) 0.25 0.5 0.75 1.0 0 y b (1) y b (4) y b (3) y b (2) 0.70.80.91.0 0 y c (1) y c (4) y c (3) y c (2) 0.1 0.20.3 1.0

37 Example 0y a (1) y a (2) 0.25 0.5 0 y b (1) y b (2) 0.70.8 0 y c (1) y c (2) 0.1 0.2 r r r

38 Example 0y a (1) y a (4) y a (3) y a (2) 0.25 0.5 0.75 1.0 0 y b (1) y b (4) y b (3) y b (2) 0.70.80.91.0 0 y c (1) y c (4) y c (3) y c (2) 0.1 0.20.3 1.0

39 Example 0 y c (1) y c (4) y c (3) y c (2) 0.1 0.20.3 1.0

40 Example 0 y c (3) y c (2) 0.10.2 r -y c (1)

41 Example 0y a (1) y a (4) y a (3) y a (2) 0.25 0.5 0.75 1.0 0 y b (1) y b (4) y b (3) y b (2) 0.70.80.91.0 0 y c (1) y c (4) y c (3) y c (2) 0.1 0.20.3 1.0

42 Interval Move A move that mimics interval rounding Considers all variables and an interval of labels Changes labeling iteratively

43 Key Observation If d is submodular d(i,k) + d(i+1,k+1) ≤ d(i,k+1) + d(i+1,k), for all i, k Schlesinger and Flach, 2003 energy can be minimized via minimum cut

44 Interval Move VaVa VbVb θ ab (i,k) = w ab d(i,k) Choose an interval of length h’

45 Interval Move VaVa VbVb θ ab (i,k) = w ab d(i,k) Choose an interval of length h’ Add the current labels

46 Interval Move VaVa VbVb θ ab (i,k) = w ab d’(i,k) Choose an interval of length h’ Add the current labels d’(i,k) ≥ d(i,k) d’ is submodular Solve to update labels Repeat until convergence

47 Interval Move Each problem can be solved using minimum cut Same multiplicative bound as interval rounding Multiplicative bound is tight

48 Boykov, Veksler and Zabih Kleinberg and Tardos Length of interval = 1 Move-Making Algorithms Convex Relaxations

49 Boykov, Veksler and Zabih Chekuri, Khanna, Naor and Zosin Length of interval = 1 Optimal interval length Move-Making Algorithms Convex Relaxations

50 Theoretical Guarantees MovesLP Uniform22 Metric2MO(log h) Truncated Linear 2 + √2 Truncated Quadratic O(√M) M = ratio of maximum and minimum non-zero distance

51 Outline Approximate Algorithms Comparison Rounding-based Moves –Complete Rounding –Interval Rounding –Hierarchical Rounding

52 Hierarchical Rounding L1L1 L2L2 l1l1 l2l2 l3l3 l4l4 l5l5 l6l6 l7l7 l8l8 l9l9 L3L3 Hierarchical clustering of labels (e.g. r-HST metrics)

53 Hierarchical Rounding L1L1 L2L2 l1l1 l2l2 l3l3 l4l4 l5l5 l6l6 l7l7 l8l8 l9l9 L3L3 Assign variables to labels L 1, L 2 or L 3 Move down the hierarchy until the leaf level

54 Hierarchical Rounding L1L1 L2L2 l1l1 l2l2 l3l3 l4l4 l5l5 l6l6 l7l7 l8l8 l9l9 L3L3 Assign variables to labels l 1, l 2 or l 3

55 Hierarchical Rounding L1L1 L2L2 l1l1 l2l2 l3l3 l4l4 l5l5 l6l6 l7l7 l8l8 l9l9 L3L3 Assign variables to labels l 4, l 5 or l 6

56 Hierarchical Rounding L1L1 L2L2 l1l1 l2l2 l3l3 l4l4 l5l5 l6l6 l7l7 l8l8 l9l9 L3L3 Assign variables to labels l 7, l 8 or l 9

57 Hierarchical Move L1L1 L2L2 l1l1 l2l2 l3l3 l4l4 l5l5 l6l6 l7l7 l8l8 l9l9 L3L3 Hierarchical clustering of labels (e.g. r-HST metrics)

58 Hierarchical Move L1L1 L2L2 l1l1 l2l2 l3l3 l4l4 l5l5 l6l6 l7l7 l8l8 l9l9 L3L3 Obtain labeling f 1 restricted to labels {l 1,l 2,l 3 }

59 Hierarchical Move L1L1 L2L2 l1l1 l2l2 l3l3 l4l4 l5l5 l6l6 l7l7 l8l8 l9l9 L3L3 Obtain labeling f 2 restricted to labels {l 4,l 5,l 6 }

60 Hierarchical Move L1L1 L2L2 l1l1 l2l2 l3l3 l4l4 l5l5 l6l6 l7l7 l8l8 l9l9 L3L3 Obtain labeling f 3 restricted to labels {l 7,l 8,l 9 }

61 Hierarchical Move L1L1 L2L2 L3L3 VaVa VbVb f 1 (a) f 2 (a) f 3 (a) Move up the hierarchy until we reach the root f 1 (b) f 2 (b) f 3 (b)

62 Hierarchical Move Each problem can be solved using minimum cut Same multiplicative bound as hierarchical rounding Multiplicative bound is tight

63 Boykov, Veksler and Zabih Kleinberg and Tardos Flat hierarchy r-HST hierarchy Move-Making Algorithms Convex Relaxations

64 Theoretical Guarantees MovesLP Uniform22 MetricO(log h) Truncated Linear 2 + √2 Truncated Quadratic O(√M) M = ratio of maximum and minimum non-zero distance

65 Questions? http://cvn.ecp.fr/personnel/pawan pawan.kumar@ecp.fr

Download ppt "Rounding-based Moves for Metric Labeling M. Pawan Kumar École Centrale Paris INRIA Saclay, Île-de-France."

Similar presentations

MAP Estimation Algorithms in M. Pawan Kumar, University of Oxford Pushmeet Kohli, Microsoft Research Computer Vision - Part I.

MAP Estimation Algorithms in M. Pawan Kumar, University of Oxford Pushmeet Kohli, Microsoft Research Computer Vision - Part I.
MAP Estimation Algorithms in

MAP Estimation Algorithms in
Algorithms for MAP estimation in Markov Random Fields Vladimir Kolmogorov University College London Tutorial at GDR (Optimisation Discrète, Graph Cuts.

Algorithms for MAP estimation in Markov Random Fields Vladimir Kolmogorov University College London Tutorial at GDR (Optimisation Discrète, Graph Cuts.
1 LP, extended maxflow, TRW OR: How to understand Vladimirs most recent work Ramin Zabih Cornell University.

1 LP, extended maxflow, TRW OR: How to understand Vladimirs most recent work Ramin Zabih Cornell University.
Solving Markov Random Fields using Second Order Cone Programming Relaxations M. Pawan Kumar Philip Torr Andrew Zisserman.

Solving Markov Random Fields using Second Order Cone Programming Relaxations M. Pawan Kumar Philip Torr Andrew Zisserman.
Solving IPs – Cutting Plane Algorithm General Idea: Begin by solving the LP relaxation of the IP problem. If the LP relaxation results in an integer solution,

Solving IPs – Cutting Plane Algorithm General Idea: Begin by solving the LP relaxation of the IP problem. If the LP relaxation results in an integer solution,
Unsupervised Learning Clustering K-Means. Recall: Key Components of Intelligent Agents Representation Language: Graph, Bayes Nets, Linear functions Inference.

Unsupervised Learning Clustering K-Means. Recall: Key Components of Intelligent Agents Representation Language: Graph, Bayes Nets, Linear functions Inference.
Graph Cut Algorithms for Computer Vision & Medical Imaging Ramin Zabih Computer Science & Radiology Cornell University Joint work with Y. Boykov, V. Kolmogorov,

Graph Cut Algorithms for Computer Vision & Medical Imaging Ramin Zabih Computer Science & Radiology Cornell University Joint work with Y. Boykov, V. Kolmogorov,
ICCV 2007 tutorial Part III Message-passing algorithms for energy minimization Vladimir Kolmogorov University College London.

ICCV 2007 tutorial Part III Message-passing algorithms for energy minimization Vladimir Kolmogorov University College London.
Dependent Randomized Rounding in Matroid Polytopes (& Related Results) Chandra Chekuri Jan VondrakRico Zenklusen Univ. of Illinois IBM ResearchMIT.

Dependent Randomized Rounding in Matroid Polytopes (& Related Results) Chandra Chekuri Jan VondrakRico Zenklusen Univ. of Illinois IBM ResearchMIT.
The University of Ontario CS 4487/9687 Algorithms for Image Analysis Multi-Label Image Analysis Problems.

The University of Ontario CS 4487/9687 Algorithms for Image Analysis Multi-Label Image Analysis Problems.
An Analysis of Convex Relaxations (PART I) Minimizing Higher Order Energy Functions (PART 2) Philip Torr Work in collaboration with: Pushmeet Kohli, Srikumar.

An Analysis of Convex Relaxations (PART I) Minimizing Higher Order Energy Functions (PART 2) Philip Torr Work in collaboration with: Pushmeet Kohli, Srikumar.
Discrete Optimization for Vision and Learning. Who? How? M. Pawan Kumar Associate Professor Ecole Centrale Paris Nikos Komodakis Associate Professor Ecole.

Discrete Optimization for Vision and Learning. Who? How? M. Pawan Kumar Associate Professor Ecole Centrale Paris Nikos Komodakis Associate Professor Ecole.
Learning with Inference for Discrete Graphical Models Nikos Komodakis Pawan Kumar Nikos Paragios Ramin Zabih (presenter)

Learning with Inference for Discrete Graphical Models Nikos Komodakis Pawan Kumar Nikos Paragios Ramin Zabih (presenter)
Approximation Algoirthms: Semidefinite Programming Lecture 19: Mar 22.

Approximation Algoirthms: Semidefinite Programming Lecture 19: Mar 22.
Robust Higher Order Potentials For Enforcing Label Consistency

Robust Higher Order Potentials For Enforcing Label Consistency
Design of Optimal Multiple Spaced Seeds for Homology Search Jinbo Xu School of Computer Science, University of Waterloo Joint work with D. Brown, M. Li.

Design of Optimal Multiple Spaced Seeds for Homology Search Jinbo Xu School of Computer Science, University of Waterloo Joint work with D. Brown, M. Li.
P 3 & Beyond Solving Energies with Higher Order Cliques Pushmeet Kohli Pawan Kumar Philip H. S. Torr Oxford Brookes University CVPR 2007.

P 3 & Beyond Solving Energies with Higher Order Cliques Pushmeet Kohli Pawan Kumar Philip H. S. Torr Oxford Brookes University CVPR 2007.
Improved Moves for Truncated Convex Models M. Pawan Kumar Philip Torr.

Improved Moves for Truncated Convex Models M. Pawan Kumar Philip Torr.
2010/5/171 Overview of graph cuts. 2010/5/172 Outline Introduction S-t Graph cuts Extension to multi-label problems Compare simulated annealing and alpha-

2010/5/171 Overview of graph cuts. 2010/5/172 Outline Introduction S-t Graph cuts Extension to multi-label problems Compare simulated annealing and alpha-

Similar presentations

Ads by Google