An Efficient Decomposition Framework for Discriminative Segmentation with Supermodular Losses

Several supermodular losses have been shown to improve the perceptual quality of image segmentation in a discriminative framework such as a structured output support vector machine (SVM). These loss functions do not necessarily have the same structure as the one used by the segmentation inference algorithm, and in general, we may have to resort to generic submodular minimization algorithms for loss augmented inference. Although these come with polynomial time guarantees, they are not practical to apply to image scale data. Many supermodular losses come with strong optimization guarantees, but are not readily incorporated in a loss augmented graph cuts procedure. This motivates our strategy of employing the alternating direction method of multipliers (ADMM) decomposition for loss augmented inference. In doing so, we create a new API for the structured SVM that separates the maximum a posteriori (MAP) inference of the model from the loss augmentation during training. In this way, we gain computational efficiency, making new choices of loss functions practical for the first time, while simultaneously making the inference algorithm employed during training closer to the test time procedure. We show improvement both in accuracy and computational performance on the Microsoft Research Grabcut database and a brain structure segmentation task, empirically validating the use of several supermodular loss functions during training, and the improved computational properties of the proposed ADMM approach over the Fujishige-Wolfe minimum norm point algorithm.

[1]  Ben Taskar,et al.  Discriminative learning of Markov random fields for segmentation of 3D scan data , 2005, 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05).

[2]  Maurice Queyranne,et al.  Minimizing symmetric submodular functions , 1998, Math. Program..

[3]  Thorsten Joachims,et al.  Training structural SVMs when exact inference is intractable , 2008, ICML '08.

[4]  Alexander Schrijver,et al.  Combinatorial optimization. Polyhedra and efficiency. , 2003 .

[5]  Christoph H. Lampert,et al.  Learning to Localize Objects with Structured Output Regression , 2008, ECCV.

[6]  Sebastian Nowozin,et al.  Structured Learning and Prediction in Computer Vision , 2011, Found. Trends Comput. Graph. Vis..

[7]  Vladimir Kolmogorov,et al.  Minimizing a sum of submodular functions , 2010, Discret. Appl. Math..

[8]  Ben Taskar,et al.  Max-Margin Markov Networks , 2003, NIPS.

[9]  Matthew B. Blaschko,et al.  A Convex Surrogate Operator for General Non-Modular Loss Functions , 2016, AISTATS.

[10]  Patrick Pérez,et al.  Interactive Image Segmentation Using an Adaptive GMMRF Model , 2004, ECCV.

[11]  Sebastian Nowozin,et al.  Optimal Decisions from Probabilistic Models: The Intersection-over-Union Case , 2014, 2014 IEEE Conference on Computer Vision and Pattern Recognition.

[12]  Wojciech Zaremba,et al.  Discriminative training of CRF models with probably submodular constraints , 2016, 2016 IEEE Winter Conference on Applications of Computer Vision (WACV).

[13]  Andrew McCallum,et al.  Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data , 2001, ICML.

[14]  H. Damasio,et al.  IEEE Transactions on Pattern Analysis and Machine Intelligence: Special Issue on Perceptual Organization in Computer Vision , 1998 .

[15]  Pravesh Kothari,et al.  Provable Submodular Minimization using Wolfe's Algorithm , 2014, NIPS.

[16]  Guillaume Charpiat,et al.  Exhaustive family of energies minimizable exactly by a graph cut , 2011, CVPR 2011.

[17]  Dimitri P. Bertsekas,et al.  Nonlinear Programming , 1997 .

[18]  Andreas Krause,et al.  Efficient Minimization of Decomposable Submodular Functions , 2010, NIPS.

[19]  Matthew B. Blaschko,et al.  The Lovász Hinge: A Convex Surrogate for Submodular Losses , 2015, ArXiv.

[20]  Tamir Hazan,et al.  Efficient Training of Structured SVMs via Soft Constraints , 2015, AISTATS.

[21]  Satoru Fujishige,et al.  Lexicographically Optimal Base of a Polymatroid with Respect to a Weight Vector , 1980, Math. Oper. Res..

[22]  Nikos Komodakis,et al.  MRF Optimization via Dual Decomposition: Message-Passing Revisited , 2007, 2007 IEEE 11th International Conference on Computer Vision.

[23]  Matthew B. Blaschko,et al.  Efficient Learning for Discriminative Segmentation with Supermodular Losses , 2016, BMVC.

[24]  Pushmeet Kohli,et al.  Learning Low-order Models for Enforcing High-order Statistics , 2012, AISTATS.

[25]  Satoru Fujishige,et al.  Submodular functions and optimization , 1991 .

[26]  Nikos Paragios,et al.  Discrete Multi Atlas Segmentation using Agreement Constraints , 2014, BMVC.

[27]  Olga Veksler,et al.  Fast approximate energy minimization via graph cuts , 2001, Proceedings of the Seventh IEEE International Conference on Computer Vision.

[28]  Richard S. Zemel,et al.  Structured Output Learning with High Order Loss Functions , 2012, AISTATS.

[29]  Torsten Rohlfing,et al.  Image Similarity and Tissue Overlaps as Surrogates for Image Registration Accuracy: Widely Used but Unreliable , 2012, IEEE Transactions on Medical Imaging.

[30]  Thomas Hofmann,et al.  Large Margin Methods for Structured and Interdependent Output Variables , 2005, J. Mach. Learn. Res..

[31]  Pushmeet Kohli,et al.  Perceptually Inspired Layout-Aware Losses for Image Segmentation , 2014, ECCV.

[32]  R. A. Silverman,et al.  Introductory Real Analysis , 1972 .

[33]  Richard S. Zemel,et al.  HOP-MAP: Efficient Message Passing with High Order Potentials , 2010, AISTATS.

[34]  C. SIAMJ. A FASTER SCALING ALGORITHM FOR MINIMIZING SUBMODULAR FUNCTIONS∗ , 2001 .

[35]  Stefano Soatto,et al.  Class segmentation and object localization with superpixel neighborhoods , 2009, 2009 IEEE 12th International Conference on Computer Vision.

[36]  Michael I. Jordan,et al.  On the Convergence Rate of Decomposable Submodular Function Minimization , 2014, NIPS.

[37]  Vladimir Kolmogorov,et al.  Optimizing Binary MRFs via Extended Roof Duality , 2007, 2007 IEEE Conference on Computer Vision and Pattern Recognition.

[38]  Andrew Blake,et al.  Geodesic star convexity for interactive image segmentation , 2010, 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[39]  Stephen P. Boyd,et al.  Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers , 2011, Found. Trends Mach. Learn..

[40]  Thorsten Joachims,et al.  Cutting-plane training of structural SVMs , 2009, Machine Learning.

[41]  Matthew B. Blaschko,et al.  Learning Submodular Losses with the Lovasz Hinge , 2015, ICML.

[42]  Francis R. Bach,et al.  Learning with Submodular Functions: A Convex Optimization Perspective , 2011, Found. Trends Mach. Learn..

[43]  Kathrin Klamroth,et al.  Biconvex sets and optimization with biconvex functions: a survey and extensions , 2007, Math. Methods Oper. Res..

[44]  Derek Hoiem,et al.  Learning CRFs Using Graph Cuts , 2008, ECCV.

[45]  Andreas Krause,et al.  SFO: A Toolbox for Submodular Function Optimization , 2010, J. Mach. Learn. Res..

[46]  Vladimir Kolmogorov,et al.  An experimental comparison of min-cut/max- flow algorithms for energy minimization in vision , 2001, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[47]  James B. Orlin,et al.  A faster strongly polynomial time algorithm for submodular function minimization , 2007, Math. Program..