Workshop on Robust Subspace Learning and Computer Vision

Recent research on robust subspace learning and tracking by decomposition into low-rank plus additive matrices/tensors provides a suitable framework for computer vision applications such as video coding, key frame extraction, hyper-spectral video processing, dynamic MRI, motion saliency detection and background/foreground separation. In this context, decomposition into low rank plus sparse matrices has been developed in different types of problem formulation such as robust principal component analysis, robust non-negative matrix factorization, robust matrix completion, subspace tracking, and low-rank minimization.

In this context, decomposition into low rank plus sparse matrices has been developed in different types of problem formulation such as robust principal component analysis, robust non-negative matrix factorization, robust matrix completion, subspace tracking, and low-rank minimization. These different approaches differ from the decomposition, the corresponding optimization problem and the solvers. The optimization problem can be NP-hard in its original formulation, and it can be convex or not following the constraints and the loss functions used. Thus, the key challenges concern the design of efficient relaxed models and solvers which have to be with iterations as few as possible, and as efficient as possible.

The aim of RSL-CV 2015 are three-fold: 1) proposing robust subspace learning and tracking for computer vision applications, 2) proposing new adaptive and incremental algorithms for robust subspace learning and tracking to reach the requirements of real-time applications such as background/foreground separation, motion saliency and video coding, and 3) proposing robust algorithms to tackle key challenges in applications such as dynamic backgrounds and illumination changes for background/foreground separation.

Papers are solicited to address robust subspace learning and tracking based on matrix/tensor decomposition, to be applied in computer vision, including but not limited to the followings:

  • Robust Principal Component Analysis (RPCA)
  • Decomposition in low-rank plus additive matrices/tensors
  • Solvers (ALM, ADM, etc…)
  • Efficient SVD algorithms
  • Incremental RPCA
  • Real time implementation on GPU
  • Embedded implementation
  • Compressive Sensing
  • Robust Matrix/Tensor Factorization (RMF)
  • Robust Matrix/Tensor Completion (RMC)
  • Subspace Tracking (ST)
  • Low rank minimization (LRM)
  • Structured Sparsity, Dynamic Group Sparsity
  • Dictionary Learning
  • Sparse Representation
  • Linear Approximation

We encourage authors to evaluate their approach on at least one of the reference datasets for each application (Please see the Computer Vision Datasets).

Other resources are available here.

The printable Call for Papers  for RSL-CV 2015 can be downloaded here.