719 research outputs found
Grocery product detection and recognition
No full textObject detection and recognition are challenging computer vision tasks receiving great attention due to the large number of applications. This work focuses on the detection/recognition of products in supermarket shelves; this framework has a number of practical applications such as providing additional product/price information to the user or guiding visually impaired customers during shopping. The automatic creation of planograms (i.e., actual layout of products on shelves) is also useful for commercial analysis and management of large stores. Although in many object detection/recognition contexts it can be assumed that training images are representative of the real operational conditions, in our scenario such assumption is not realistic because the only training images available are acquired in well-controlled conditions. This gap between the training and test data makes the object detection and recognition tasks far more complex and requires very robust techniques. In this paper we prove that good results can be obtained by exploiting color and texture information in a multi-stage process: pre-selection, fine-selection and post processing. For fine-selection we compared a classical Bag of Words technique with a more recent Deep Neural Networks approach and found interesting outcomes. Extensive experiments on datasets of varying complexity are discussed to highlight the main issues characterizing this problem, and to guide toward the practical development of a real application
A fast algorithm for nonconvex approaches to sparse recovery problems
No full textThis paper addresses the problem of sparse signal recovery from a lower number of
measurements than those requested by the classical compressed sensing theory. This
problem is formalized as a constrained minimization problem, where the objective
function is nonconvex and singular at the origin. Several algorithms have been recently
proposed, which rely on iterative reweighting schemes, that produce better estimates at
each new minimization step. Two such methods are iterative reweighted l2 and l1
minimization that have been shown to be effective and general, but very computationally
demanding. The main contribution of this paper is the proposal of the algorithm
WNFCS, where the reweighted schemes represent the core of a penalized approach to the
solution of the constrained nonconvex minimization problem. The algorithm is fast, and
succeeds in exactly recovering a sparse signal from a smaller number of measurements
than the l1 minimization and in a shorter time. WNFCS is very general, since it represents
an algorithmic framework that can easily be adapted to different reweighting strategies
and nonconvex objective functions. Several numerical experiments and comparisons
with some of the most recent nonconvex minimization algorithms confirm the capabilities
of the proposed algorithm
Blind cluster structured sparse signal recovery: A nonconvex approach
No full textWe consider the problem of recovering a sparse signal
when its nonzero coefficients tend to cluster into blocks, whose number, dimension and
position are unknown. We refer to this problem as {it blind cluster
structured sparse recovery}. For its solution, differently from the existing methods that consider the problem
in a statistical context, we propose a deterministic neighborhood based approach characterized by the use both of a nonconvex, nonseparable
sparsity inducing function and of a penalized version of the iterative reweighted method. Despite the
high nonconvexity of the approach, a suitable integration of these building elements led to the development of
MB-NFCS ({it Model Based Nonlinear Filtering for Compressed Sensing}), an iterative fast, self-adaptive, and efficient algorithm that, without requiring any information on the
sparsity pattern, adjusts at each iteration the action of the sparsity inducing function in order to strongly encourage the emerging
cluster structure. The effectiveness of the proposed approach is demonstrated by a large set of numerical experiments
that show the superior performance of MB-NFCS to the state-of-the-art algorithms
On the Generation of Synthetic Fingerprint Alterations
No full textIn this paper we propose some techniques to generate synthetic altered fingerprints and prove the utility of the generated datasets for developing, tuning and evaluating algorithms for altered fingerprint detection/matching. Due to the lack of public databases of altered fingerprints the generation tool proposed (and made freely available) can be a valid instrument to boost research on these challenging problems
Location-Based Discovery and Network Handover Management for Heterogeneous IEEE 802.11ah IoT Applications
No full textThis research was funded by the Flemish IDEAL-IoT project (FWO SBO, grant nr. S004017N). The author Serena Santi is funded by the Flemish FWO SB grant (nr. 1S82120N). The author Filip Lemic was supported by the EU MSCA grant (nr. 893760). The computational resources were provided by the VSC (Flemish Supercomputer Center), funded by FWO and the Flemish Government -department EWI
Saliency-based keypoint selection for fast object detection and matching
No full textIn this paper we present a new approach to rank and select keypoints based on their saliency for object detection and matching under moderate viewpoint and lighting changes. Saliency is defined in terms of detectability, repeatability and distinctiveness by considering both the keypoint strength (as returned by the detector algorithm) and the associated local descriptor discriminating power. Our experiments prove that selecting a small amount of available keypoints (e.g., 10%) not only boosts efficiency but can also lead to better detection/matching accuracy thus making the proposed method attractive for real-time applications (e.g., augmented reality)
Filtered wavelet thresholding methods
No full textAbstractWhen working with nonlinear filtering algorithms for image denoising problems, there are two crucial aspects, namely, the choice of the thresholding parameter λ and the use of a proper filter function. Both greatly influence the quality of the resulting denoised image. In this paper we propose two new filters, which are a piecewise quadratic and an exponential function of λ, respectively, arid we show how they can be successfully used instead of the classical Donoho and Johnstone's Soft thresholding filter. We exploit the increased regularity and flexibility of the new filters to improve the quality of the final results. Moreover, we prove that our filtered approximation is a near-minimizer of the functional which has to be minimized to solve the denoising problem. We also show that the quadratic filter, due to its shape, yields good results if we choose λ as the Donoho and Johnstone universal threshold, while the exponential one is more suitable if we use the recently proposed H-curve criterion. Encouraging results in extensive numerical experiments on several test images confirm the effectiveness of our proposal
Erratum: Lack of immunity against rubella among Italian young adults. [BMC Infect Dis., 17, (2017) (199)] Doi: 10.1186/s12879-017-2295-y
No full textAfter publication of this article [1], the authors noted that the given names and family names of all authors had been inverted, and are therefore incorrect in the original article. In the original article, the author names appear as the following: Gallone Maria Serena, Gallone Maria Filomena, Larocca Angela Maria Vittoria, Germinario Cinzia and Tafuri Silvio. However, this is incorrect, and the author names should appear as per the below: Maria Serena Gallone, Maria Filomena Gallone, Angela Maria Vittoria Larocca, Cinzia Germinario, Silvio Tafuri. The author names have been corrected in the author list and the citation for this Erratum
Image denoising using principal component analysis in the wavelet domain
No full textAbstractIn this work we describe a method for removing Gaussian noise from digital images, based on the combination of the wavelet packet transform and the principal component analysis. In particular, since the aim of denoising is to retain the energy of the signal while discarding the energy of the noise, our basic idea is to construct powerful tailored filters by applying the Karhunen–Loéve transform in the wavelet packet domain, thus obtaining a compaction of the signal energy into a few principal components, while the noise is spread over all the transformed coefficients. This allows us to act with a suitable shrinkage function on these new coefficients, removing the noise without blurring the edges and the important characteristics of the images. The results of a large numerical experimentation encourage us to keep going in this direction with our studies
- …
