Articulation-Invariant Representation of Non-planar Shapes

Gopalan, Raghuraman; Turaga, Pavan; Chellappa, Rama

doi:10.1007/978-3-642-15558-1_21

Cited by 80 publications

(137 citation statements)

References 25 publications

Supporting

Mentioning

136

Contrasting

Order By: Relevance

“…The shape descriptors were: curvature (as an example descriptor of one dimension for each point), BAS [2] (four dimensiones) and the shape contexts (SC) [5,6] (60 dimensions). The results achieved with these descriptors, and in particular the ones with the shape contexts, can be applied to other ones of similar characteristics from the bibliography [3,4,[6][7][8]. We also used a synthetic corpus of sequences of several dimensions (1, 5, 10, 20 and 60).…”

Section: Methodsmentioning

confidence: 99%

“…Among the methods to solve this problem the ones related to Dynamic Time Warping (DTW) [1] and descriptors of the contour with sequences of components of several dimensions have had a significant presence [2][3][4][5][6][7][8]. In general, these shape descriptors aim to have information from all of the contour with respect to each point, that is the reason for their large size (see Figure 1 for an example of the shape descriptor used in [6]).…”

Section: Introductionmentioning

confidence: 99%

“…In this work, they do not use the algorithm of [10], but they make clusters of sequences based on their similarity, treating every possible starting point as a different sequence and using indexing methods with lower bounds of these clusters. This solution seems to be suitable just for shape descriptors with only one dimension (such as the curvature) and not for much more dimensions [2][3][4][5][6][7][8]. For instance, in [5], 60 dimensions are required for each point (Figure 1).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A Heuristic Based on the Intrinsic Dimensionality for Reducing the Number of Cyclic DTW Comparisons in Shape Classification and Retrieval Using AESA

Palazón-González

Marzal

2012

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Cyclic Dynamic Time Warping (CDTW) is a good dissimilarity of shape descriptors of high dimensionality based on contours, but it is computationally expensive. For this reason, to perform recognition tasks, a method to reduce the number of comparisons and avoid an exhaustive search is convenient. The Approximate and Eliminate Search Algorithm (AESA) is a relevant indexing method because of its drastic reduction of comparisons, however, this algorithm requires a metric distance and that is not the case of CDTW. In this paper, we introduce a heuristic based on the intrinsic dimensionality that allows to use CDTW and AESA together in classification and retrieval tasks over these shape descriptors. Experimental results show that, for descriptors of high dimensionality, our proposal is optimal in practice and significantly outperforms an exhaustive search, which is the only alternative for them and CDTW in these tasks.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Heuristic Based on the Intrinsic Dimensionality for Reducing the Number of Cyclic DTW Comparisons in Shape Classification and Retrieval Using AESA

Palazón-González

Marzal

2012

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…Note that we do not require to have a complete object segmented and thus will also address partial shape matching. Note also that most of the well established shape-based approaches do not consider deformations and articulated movements, while we do [16,19]. Contrarily to classical skeletal-based representations, ours is not overly sensitive to small boundary deformations and furthermore gives high response in those regions where the object has high curvature with large boundary support and in the vicinity of joints (between well-delineated parts, such as the limbs of an animal).…”

Section: Introductionmentioning

confidence: 95%

“…Also, rather than comparing our method only with readily available texture-based well-known methods, such as SIFT and SURF which are well suited to deal with large databases of images, we ought to also compare with more closely related methods such as the Inner Distance Shape Context method -one of the few other published method which is capable to explicitly deal with planar articulations [16] -and other recent related methods (e.g. [19,55,46]); but notice that it is unclear how these other shape-based techniques scale with a database size (as they are often only tested on relatively small datasets of at most a few hundred targets).…”

Section: Fig 27mentioning

confidence: 99%

Point-based medialness for 2D shape description and identification

Aparajeya

Leymarie

2015

Multimed Tools Appl

View full text Add to dashboard Cite

We propose a perception-based medial point description of a natural form (2D: static or in articulated movement) as a framework for a shape representation which can then be efficiently used in biological species identification and matching tasks. Medialness is defined by adapting and refining a definition first proposed in the cognitive science literature when studying the visual attention of human subjects presented with articulated biological 2D forms in movement, such as horses, dogs and humans (walking, running). In particular, special loci of high medialness for the interior of a form in movement, referred to as "hot spots", prove most attractive to the human perceptual system. We propose an algorithmic process to identify such hot spots. In this article we distinguish exterior from interior shape representation. We further augment hot spots with extremities of medialness ridges identifying significant concavities (from outside) and convexities (from inside). Our representation is strongly footed in results from cognitive psychology, but also inspired by know-how in art and animation, and the algorithmic part is influenced by techniques from more traditional computer vision. A robust shape matching algorithm is designed that finds the most relevant targets from a database of templates by comparing feature points in a scale, rotation and translation invariant way. The performance of our method has been tested on several databases. The robustness of the algorithm is further tested by perturbing the data-set at different levels.

show abstract

Contextual Spaces Re‐Ranking: accelerating the Re‐sort Ranked Lists step on heterogeneous systems

Pisani

Pedronette

Torres

et al. 2016

Concurrency and Computation

View full text Add to dashboard Cite

Re-ranking algorithms have been proposed to improve the effectiveness of content-based image retrieval systems by exploiting contextual information encoded in distance measures and ranked lists. In this paper, we show how we improved the efficiency of one of these algorithms, called Contextual Spaces Re-Ranking (CSRR). One of our approaches consists in parallelizing the algorithm with OpenCL to use the central and graphics processing units of an accelerated processing unit. The other is to modify the algorithm to a version that, when compared with the original CSRR, not only reduces the total running time of our implementations by a median of 1:6 but also increases the accuracy score in most of our test cases. Combining both parallelization and algorithm modification results in a median speedup of 5:4 from the original serial CSRR to the parallelized modified version. Different implementations for CSRR's Re-sort Ranked Lists step were explored as well, providing insights into graphics processing unit sorting, the performance impact of image descriptors, and the trade-offs between effectiveness and efficiency. vectors, which are later used for similarity assessment among images. A CBIR system ranks the collection images by decreasing order of similarity, and because users consider mostly top-ranked images, it is imperative that the rank be as accurate as possible.In recent years, several successful attempts to increase the effectiveness (quality of results) of CBIR systems have been performed [1][2][3][4][5][6]. In particular, re-ranking methods have been used to improve the effectiveness of CBIR systems by exploiting contextual information encoded in similarity scores and ranked lists. These methods are, on the other hand, very costly as they are based on comparing collection images multiple times. In a real-world scenario, CBIR systems require both good effectiveness and efficiency (response time), so re-ranking methods must be improved.Central processing units (CPUs) no longer have just one core, and graphics processing units (GPUs) are now being used as general purpose processors due to having evolved into massive parallel architectures capable of executing hundreds of operations per cycle [7]. These devices have been successfully used to accelerate re-ranking [7, 8] and retrieval [9] systems, obtaining good speedups.Therefore, alternatives that increase performance with parallelization seem to be a possible fit for the Contextual Spaces Re-Ranking (CSRR) algorithm [2], which we discuss in this paper. Another possible approach consists in analyzing the compromises between accuracy and performance that come out of modifying existing algorithms, because this can lead to eliminating demanding work.Our solution first exploits the use of parallelization to speed up the more costly steps of the CSRR algorithm, obtaining speedups of up to 3:3 for the Compute Distances step and 5:1 for the Re-sort Ranked Lists step on an accelerated processing unit (APU). We then propose a modification to this algorithm, which by itse...

show abstract

Articulation-Invariant Representation of Non-planar Shapes

Abstract: Abstract. Given a set of points corresponding to a 2D projection of a non-planar shape, we would like to obtain a representation invariant to articulations (under no self-occlusions

Cited by 80 publications

References 25 publications

A Heuristic Based on the Intrinsic Dimensionality for Reducing the Number of Cyclic DTW Comparisons in Shape Classification and Retrieval Using AESA

A Heuristic Based on the Intrinsic Dimensionality for Reducing the Number of Cyclic DTW Comparisons in Shape Classification and Retrieval Using AESA

Point-based medialness for 2D shape description and identification

Contextual Spaces Re‐Ranking: accelerating the Re‐sort Ranked Lists step on heterogeneous systems

Contact Info

Product

Resources

About