Vision and Learning Seminar SeriesWeekly on Tuesdays at 2pmVENUE: CSE E305Coordinators: Prof. Baba C. Vemuri, Ted Ha
Date: Tuesday October 30, 2012
Cosegmentation is typically defined as the task of jointly segmenting "something similar" in a given set of images. Existing methods are too generic and so far have not demonstrated competitive results for any specific task. In this paper we overcome this limitation by adding two new aspects to cosegmentation: (1) the "something" has to be an object, and (2) the "similarity" measure is learned. In this way, we are able to achieve excellent results on the recently introduced iCoseg dataset, which contains small sets of images of either the same object instance or similar objects of the same class. The challenge of this dataset lies in the extreme changes in viewpoint, lighting, and object deformations within each set. We are able to considerably outperform several competitors. To achieve this performance, we borrow recent ideas from object recognition: the use of powerful features extracted from a pool of candidate objectlike segmentations. We believe that our work will be beneficial to several application areas, such as image retrieval.
Date: Tuesday October 23, 2012
Compressed sensing is a new data acquisition paradigm enabling universal, simple, and reduced-cost acquisition, by exploiting a sparse signal model. Most notably, recovery of the signal by computationally efficient algorithms is guaranteed for certain randomized acquisition systems. However, there is a discrepancy between the theoretical guarantees and practical applications. In applications, including Fourier imaging in various modalities, the measurements are acquired by inner products with vectors selected randomly (sampled) from a frame. Currently available guarantees are derived using a so-called restricted isometry property (RIP), which has only been shown to hold under ideal assumptions. For example, the sampling from the frame needs to be independent and identically distributed with the uniform distribution, and the frame must be tight. In practice though, one or more of the ideal assumptions is typically violated and none of the existing guarantees applies. Motivated by this discrepancy, we propose two related changes in the existing framework: (i) a generalized RIP called the restricted biorthogonality property (RBOP); and (ii) correspondingly modified versions of existing greedy pursuit algorithms, which we call oblique pursuits. Oblique pursuits are guaranteed using the RBOP without requiring ideal assumptions; hence, the guarantees apply to practical acquisition schemes. Numerical results show that oblique pursuits also perform competitively with, or sometimes better than their conventional counterparts.
Date: Tuesday October 2, 2012
The Dirichlet process is a stochastic proces used in Bayesian nonparametric models of data, particularly in Dirichlet process mixture models (also known as infinite mixture models). It is a distribution over distributions, i.e. each draw from a Dirichlet process is itself a distribution. It is called a Dirichlet process because it has Dirichlet distributed finite dimensional marginal distributions, just as the Gaussian process, another popular stochastic process used for Bayesian nonparametric regression, has Gaussian distributed finite dimensional marginal distributions. Distributions drawn from a Dirichlet process are discrete, but cannot be described using a finite number of parameters, thus the classification as a nonparametric model.
Date: Tuesday September 11, 2012
In recent years, several methods have been proposed to combine multiple kernels instead of using a single one. These different kernels may correspond to using different notions of similarity or may be using information coming from multiple sources (different representations or different feature subsets). In trying to organize and highlight the similarities and differences between them, we give a taxonomy of and review several multiple kernel learning algorithms. We perform experiments on real data sets for better illustration and comparison of existing algorithms. We see that though there may not be large differences in terms of accuracy, there is difference between them in complexity as given by the number of stored support vectors, the sparsity of the solution as given by the number of used kernels, and training time complexity. We see that overall, using multiple kernels instead of a single one is useful and believe that combining kernels in a nonlinear or data-dependent way seems more promising than linear combination in fusing information provided by simple linear kernels, whereas linear methods are more reasonable when combining complex Gaussian kernels. Abstract (Lanckriet) : Kernel-based learning algorithms work by embedding the data into a Euclidean space, and then searching for linear relations among the embedded data points. The embedding is performed implicitly, by specifying the inner products between each pair of points in the embedding space. This information is contained in the so-called kernel matrix, a symmetric and positive definite matrix that encodes the relative positions of all points. Specifying this matrix amounts to specifying the geometry of the embedding space and inducing a notion of similarity in the input space -- classical model selection problems in machine learning. In this paper we show how the kernel matrix can be learned from data via Semi-Definite Programming techniques. When applied to a kernel matrix associated with both training and test data this gives a powerful transductive algorithm -- using the labelled part of the data one can learn an "optimal" embedding also for the unlabelled part. The induced similarity between test points is learned by using training points and their labels. Importantly, these learning problems are convex, so we obtain a method for learning both the model class and the function without local minima. Finally, the novel approach presented in the paper is supported by positive empirical results.
Date: Tuesday September 4, 2012
Surface reconstruction from gradient fields is an important final step in several applications involving gradient manipulations and estimation. Typically, the resulting gradient field is non-integrable due to linear/non-linear gradient manipulations, or due to presence of noise/outliers in gradient estimation. In this paper, we analyze integrability as error correction, inspired from recent work in compressed sensing, particulary l0 - l1 equivalence. We propose to obtain the surface by finding the gradient field which best fits the corrupted gradient field in l1 sense. We present an exhaustive analysis of the properties of l1 solution for gradient field integration using linear algebra and graph analogy. We consider three cases: (a) noise, but no outliers (b) no-noise but outliers and (c) presence of both noise and outliers in the given gradient field. We show that l1 solution performs as well as least squares in the absence of outliers. While previous l0 - l1 equivalence work has focused on the number of errors (outliers), we show that the location of errors is equally important for gradient field integration. We characterize the l1 solution both in terms of location and number of outliers, and outline scenarios where l1 solution is equivalent to l0 solution. We also show that when l1 solution is not able to remove outliers, the property of local error confinement holds: i.e., the errors do not propagate to the entire surface as in least squares. We compare with previous techniques and show that l1 solution performs well across all scenarios without the need for any tunable parameter adjustments. |
Home | People | Research | Publications | Software/Demos | Seminars | Courses | Awards | Sponsors | Contact | CISE
© University of Florida, Gainesville, FL 32611; (352) 392-3261.
This page was last updated on September 04 2007 EST.