Results 1  10
of
37
Discriminative Subvolume Search for Efficient Action Detection
"... Actions are spatiotemporal patterns which can be characterized by collections of spatiotemporal invariant features. Detection of actions is to find the reoccurrences (e.g. through pattern matching) of such spatiotemporal patterns. This paper addresses two critical issues in pattern matchingbase ..."
Abstract

Cited by 45 (7 self)
 Add to MetaCart
Actions are spatiotemporal patterns which can be characterized by collections of spatiotemporal invariant features. Detection of actions is to find the reoccurrences (e.g. through pattern matching) of such spatiotemporal patterns. This paper addresses two critical issues in pattern matchingbased action detection: (1) efficiency of pattern search in 3D videos and (2) tolerance of intrapattern variations of actions. Our contributions are twofold. First, we propose a discriminative pattern matching called naiveBayes based mutual information maximization (NBMIM) for multiclass action categorization. It improves the stateoftheart results on standard KTH dataset. Second, a novel search algorithm is proposed to locate the optimal subvolume in the 3D video space for efficient action detection. Our method is purely datadriven and does not rely on object detection, tracking or background subtraction. It can well handle the intrapattern variations of actions such as scale and speed variations, and is insensitive to dynamic and clutter backgrounds and even partial occlusions. The experiments on versatile datasets including KTH and CMU action datasets demonstrate the effectiveness and efficiency of our method. 1.
Towards Parallel Programming by Transformation: The FAN Skeleton Framework
, 2001
"... A Functional Abstract Notation (FAN) is proposed for the specification and design of parallel algorithms by means of skeletons  highlevel patterns with parallel semantics. The main weakness of the current programming systems based on skeletons is that the user is still responsible for finding the ..."
Abstract

Cited by 21 (10 self)
 Add to MetaCart
A Functional Abstract Notation (FAN) is proposed for the specification and design of parallel algorithms by means of skeletons  highlevel patterns with parallel semantics. The main weakness of the current programming systems based on skeletons is that the user is still responsible for finding the most appropriate skeleton composition for a given application and a given parallel architecture. We describe a transformational framework for the development of skeletal programs which is aimed at filling this gap. The framework makes use of transformation rules which are semantic equivalences among skeleton compositions. For a given problem, an initial, possibly inefficient skeleton specification is refined by applying a sequence of transformations. Transformations are guided by a set of performance prediction models which forecast the behavior of each skeleton and the performance benefits of different rules. The design process is supported by a graphical tool which locates applicable transformations and provides performance estimates, thereby helping the programmer in navigating through the program refinement space. We give an overview of the FAN framework and exemplify its use with performancedirected program derivations for simple case studies. Our experience can be viewed as a first feasibility study of methods and tools for transformational, performancedirected parallel programming using skeletons.
Improved algorithms for the kmaximum subarray problem for small k
 In Proceedings of the 11th Annual International Conference on Computing and Combinatorics, volume 3595 of LNCS
, 2005
"... Abstract. The maximum subarray problem for a one or twodimensional array is to find the array portion that maiximizes the sum of array elements in it. The Kmaximum subarray problem is to find the K subarrays with largest sums. We improve the time complexity for the onedimensional case from O(min ..."
Abstract

Cited by 15 (5 self)
 Add to MetaCart
Abstract. The maximum subarray problem for a one or twodimensional array is to find the array portion that maiximizes the sum of array elements in it. The Kmaximum subarray problem is to find the K subarrays with largest sums. We improve the time complexity for the onedimensional case from O(min{K + n log 2 n, n √ K}) for 0 ≤ K ≤ n(n − 1)/2 to O(n log K + K 2) for K ≤ n. The latter is better when K ≤ √ n log n. If we simply extend this result to the twodimensional case, we will have the complexity of O(n 3 log K + K 2 n 2).We improve this complexity to O(n 3) for K ≤ √ n. 1
Experiments with a Task Partitioning Model for Heterogeneous Computing
, 1992
"... One potentially promising approach for exploiting the best features of a variety of different computer architectures is to partition an application program to simultaneously execute on two or more different machines interconnected with a highspeed network. A fundamental problem with this heterogene ..."
Abstract

Cited by 13 (4 self)
 Add to MetaCart
One potentially promising approach for exploiting the best features of a variety of different computer architectures is to partition an application program to simultaneously execute on two or more different machines interconnected with a highspeed network. A fundamental problem with this heterogeneous computing, however, is the difficulty of partitioning an application program across the machines. This paper presents a partitioning strategy that relates the relative performance of two heterogeneous machines to the communication cost of transferring partial results across their interconnection network. Experiments are described that use this strategy to partition two different application programs across the sequential frontend processor of a Connection Machine CM200, and its parallel backend array.
Learning To Count Objects in Images
 In NIPS
, 2010
"... We propose a new supervised learning framework for visual object counting tasks, such as estimating the number of cells in a microscopic image or the number of humans in surveillance video frames. We focus on the practicallyattractive case when the training images are annotated with dots (one dot p ..."
Abstract

Cited by 12 (1 self)
 Add to MetaCart
We propose a new supervised learning framework for visual object counting tasks, such as estimating the number of cells in a microscopic image or the number of humans in surveillance video frames. We focus on the practicallyattractive case when the training images are annotated with dots (one dot per object). Our goal is to accurately estimate the count. However, we evade the hard task of learning to detect and localize individual object instances. Instead, we cast the problem as that of estimating an image density whose integral over any image region gives the count of objects within that region. Learning to infer such density can be formulated as a minimization of a regularized risk quadratic cost function. We introduce a new loss function, which is wellsuited for such learning, and at the same time can be computed efficiently via a maximum subarray algorithm. The learning can then be posed as a convex quadratic program solvable with cuttingplane optimization. The proposed framework is very flexible as it can accept any domainspecific visual features. Once trained, our system provides accurate object counts and requires a very small time overhead over the feature extraction step, making it a good candidate for applications involving realtime processing or dealing with huge amount of visual data. 1
A Linear Time Algorithm for the k Maximal Sums Problem
"... Abstract. Finding the subvector with the largest sum in a sequence of n numbers is known as the maximum sum problem. Finding the k subvectors with the largest sums is a natural extension of this, and is known as the k maximal sums problem. In this paper we design an optimal O(n+k) time algorithm f ..."
Abstract

Cited by 6 (2 self)
 Add to MetaCart
Abstract. Finding the subvector with the largest sum in a sequence of n numbers is known as the maximum sum problem. Finding the k subvectors with the largest sums is a natural extension of this, and is known as the k maximal sums problem. In this paper we design an optimal O(n+k) time algorithm for the k maximal sums problem. We use this algorithm to obtain algorithms solving the twodimensional k maximal sums problem in O(m 2 ·n+k) time, where the input is an m ×n matrix with m ≤ n. We generalize this algorithm to solve the ddimensional problem in O(n 2d−1 +k) time. The space usage of all the algorithms can be reduced to O(n d−1 + k). This leads to the first algorithm for the k maximal sums problem in one dimension using O(n + k) time and O(k) space. 1
Bichromatic separability with two boxes: a general approach
"... Let S be a point set in general position on the plane such that its elements are colored red or blue. We study the following problem: Remove as few points as possible from S such that the remaining points can be enclosed by two isothetic rectangles, one containing all the red points, the other all t ..."
Abstract

Cited by 6 (5 self)
 Add to MetaCart
Let S be a point set in general position on the plane such that its elements are colored red or blue. We study the following problem: Remove as few points as possible from S such that the remaining points can be enclosed by two isothetic rectangles, one containing all the red points, the other all the blue points, and such that each rectangle contains only points of one color. We prove that this problem can be solved in O(n 2 log n) time and O(n) space. We show how our techniques can be generalized to solve other variants of the given problem such as the 3dimensional problem and the trichromatic problem. 1
A note on ranking k maximum sums
, 2005
"... In this paper, we design a fast algorithm for ranking the k maximum sum subsequences. Given a sequence of real numbers 〈x1, x2, · · · , xn 〉 and an integer parameter k, the problem is to compute k subsequences of consecutive elements with the sums of their elements being the largest, second large ..."
Abstract

Cited by 3 (0 self)
 Add to MetaCart
In this paper, we design a fast algorithm for ranking the k maximum sum subsequences. Given a sequence of real numbers 〈x1, x2, · · · , xn 〉 and an integer parameter k, the problem is to compute k subsequences of consecutive elements with the sums of their elements being the largest, second largest,..., and the k th largest among all possible range sums. For any value of k, 1 ≤ k ≤ n(n + 1)/2, our algorithm takes O(n + k log n) time in the worst case to rank all such subsequences. Our algorithm is optimal for k ≤ n.
Partitioning Tasks Between a Pair of Interconnected Heterogeneous Processors: a Case Study
 PRACTICE AND EXPERIENCE
, 1995
"... With the variety of computer architectures available today, it often is difficult to determine which particular type of architecture will provide the best performance on a given application program. In fact, one type of architecture may be well suited to executing one section of a program while anot ..."
Abstract

Cited by 3 (2 self)
 Add to MetaCart
With the variety of computer architectures available today, it often is difficult to determine which particular type of architecture will provide the best performance on a given application program. In fact, one type of architecture may be well suited to executing one section of a program while another architecture may be better suited to executing another section of the same program. One potentially promising approach for exploiting the best features of different computer architectures is to partition an application program to simultaneously execute on two or more types of machines interconnected with a highspeed communication network. A fundamental difficulty with this heterogeneous computing, however, is the problem of determining how to partition the application program across the interconnected machines. The goal of this paper is to show how a programmer or a compiler can use a model of a heterogeneous system to determine the machine on which each subtask should be executed. This technique is illustrated with a simple model that relates the relative performance of two heterogeneous machines to the communication time required to transfer partial results across their interconnection network. Experiments with a Connection Machine CM200 demonstrate how to apply this model to partition two different application programs across the sequential frontend processor and the parallel backend array.
Computing maximumscoring segments in almost linear time
 IN PROCEEDINGS OF THE 12TH ANNUAL INTERNATIONAL COMPUTING AND COMBINATORICS CONFERENCE, VOLUME 4112 OF LNCS
, 2006
"... Given a sequence, the problem studied in this paper is to find a set of k disjoint continuous subsequences such that the total sum of all elements in the set is maximized. This problem arises naturally in the analysis of DNA sequences. The previous best known algorithm requires Θ(n log n) time in th ..."
Abstract

Cited by 3 (1 self)
 Add to MetaCart
Given a sequence, the problem studied in this paper is to find a set of k disjoint continuous subsequences such that the total sum of all elements in the set is maximized. This problem arises naturally in the analysis of DNA sequences. The previous best known algorithm requires Θ(n log n) time in the worst case. For a given sequence of length n, we present an almost lineartime algorithm for this problem. Our algorithm uses a disjointset data structure and requires O(nα(n, n)) time in the worst case, where α(n, n) is the inverse Ackermann function.