Global Optimality Guarantees for Nonconvex Unsupervised Video Segmentation

@article{Anderson2019GlobalOG,
  title={Global Optimality Guarantees for Nonconvex Unsupervised Video Segmentation},
  author={Brendon G. Anderson and Somayeh Sojoudi},
  journal={2019 57th Annual Allerton Conference on Communication, Control, and Computing (Allerton)},
  year={2019},
  pages={965-972}
}
  • Brendon G. AndersonS. Sojoudi
  • Published 9 July 2019
  • Computer Science
  • 2019 57th Annual Allerton Conference on Communication, Control, and Computing (Allerton)
In this paper, we consider the problem of unsupervised video object segmentation via background subtraction. Specifically, we pose the nonsemantic extraction of a video’s moving objects as a nonconvex optimization problem via a sum of sparse and low-rank matrices. The resulting formulation, a nonnegative variant of robust principal component analysis, is more computationally tractable than its commonly employed convex relaxation, although not generally solvable to global optimality. In spite of… 

Figures from this paper

Noisy Low-rank Matrix Optimization: Geometry of Local Minima and Convergence Rate

This paper develops a new mathematical framework to solve the above-mentioned problem with a far less restrictive RIP constant and offers the strongest RIP bound and provides a complete theoretical analysis on the global and local optimization landscapes of general low-rank optimization problems under random corruptions from any finite-variance family.

Nonsmooth rank-one matrix factorization landscape

This work identifies a necessary and sufficient condition for the absence of spurious local minima in the rank-one case and exploits the subdifferential regularity of the objective function in order to eliminate the existence quantifier from the first-order optimality condition known as Fermat’s rule.

References

SHOWING 1-10 OF 23 REFERENCES

Nonconvex Optimization Meets Low-Rank Matrix Factorization: An Overview

This tutorial-style overview highlights the important role of statistical models in enabling efficient nonconvex optimization with performance guarantees and reviews two contrasting approaches: two-stage algorithms, which consist of a tailored initialization step followed by successive refinement; and global landscape analysis and initialization-free algorithms.

A theory on the absence of spurious solutions for nonconvex and nonsmooth optimization

This work shows that a class of non-differentiable nonconvex optimization problems arising in tensor decomposition applications are global functions, the first result concerning nonconcex methods for nonsmooth objective functions, and provides a theoretical guarantee for the widely-used $\ell_1$ norm to avoid outliers in nonconventus optimization.

A neural network approach to bayesian background modeling for video object segmentation

It is shown how Probabilistic Neural Network architecture can be extended to form an unsupervised Bayesian classifier for the domain of video object segmentation, and how such a system would be able to achieve real-time segmentation of high-resolution image sequences.

Global Optimality of Local Search for Low Rank Matrix Recovery

It is shown that there are no spurious local minima in the non-convex factorized parametrization of low-rank matrix recovery from incoherent linear measurements, which yields a polynomial time global convergence guarantee for stochastic gradient descent.

Robust principal component analysis?

It is proved that under some suitable assumptions, it is possible to recover both the low-rank and the sparse components exactly by solving a very convenient convex program called Principal Component Pursuit; among all feasible decompositions, this suggests the possibility of a principled approach to robust principal component analysis.

An efficacious medical video segmentation approach with a moving camera

A robust and efficiently background subtraction algorithm that is able to cope with local scene changes, as well as globe scene changes caused by camera motion, is developed.

Adaptive background mixture models for real-time tracking

  • C. StaufferW. Grimson
  • Computer Science
    Proceedings. 1999 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (Cat. No PR00149)
  • 1999
This paper discusses modeling each pixel as a mixture of Gaussians and using an on-line approximation to update the model, resulting in a stable, real-time outdoor tracker which reliably deals with lighting changes, repetitive motions from clutter, and long-term scene changes.

A Benchmark Dataset and Evaluation Methodology for Video Object Segmentation

This work presents a new benchmark dataset and evaluation methodology for the area of video object segmentation, named DAVIS (Densely Annotated VIdeo Segmentation), and provides a comprehensive analysis of several state-of-the-art segmentation approaches using three complementary metrics.

Exact Guarantees on the Absence of Spurious Local Minima for Non-negative Robust Principal Component Analysis

This work shows that the low-dimensional formulation of the symmetric and asymmetric positive rank-1 RPCA based on the Burer-Monteiro approach has benign landscape, and provides strong deterministic and probabilistic guarantees for the exact recovery of the true principal components.