Oscar David Robles

Static Multi-device Load Balancing for OpenCL

This paper presents the Load Balancing for OpenCL (lbcl) library, devoted to automatically solve load balancing issues on both multi-platform and heterogeneous environments. Using this library, a single kernel can be executed on a set of heterogeneous devices, giving each device an amount of work proportional to its computing power. A wrapper has been developed so the library can balance the workload of an existing application not only without introducing any changes into its source code, but without any recompilation stage. Also a general OpenCL profiler has been developed to easily do a detailed profiling of the obtained results.

Parallel CBIR implementations with load balancing algorithms

The purpose of content-based information retrieval (CBIR) systems is to retrieve, from real data stored in a database, information that is relevant to a query. When large volumes of data are considered, as it is very often the case with databases dealing with multimedia data, it may become necessary to look for parallel solutions in order to store and gain access to the available items in an efficient way.Among the range of parallel options available nowadays, clusters stand out as flexible and cost effective solutions, although the fact that they are composed of a number of independent machines makes it easy for them to become heterogeneous. This paper describes a heterogeneous cluster-oriented CBIR implementation. First, the cluster solution is analyzed without load balancing, and then, a new load balancing algorithm for this version of the CBIR system is presented.The load balancing algorithm described here is dynamic, distributed, global and highly scalable. Nodes are monitored through a load index which allows the estimation of their total amount of workload, as well as the global system state. Load balancing operations between pairs of nodes take place whenever a node finishes its job, resulting in a receptor-triggered scheme which minimizes the systems communication overhead. Globally, the CBIR cluster implementation together with the load balancing algorithm can cope effectively with varying degrees of heterogeneity within the cluster; the experiments presented within the paper show the validity of the overall strategy.Together, the CBIR implementation and the load balancing algorithm described in this paper span a new path for performant, cost effective CBIR systems which has not been explored before in the technical literature.

A study of Zernike invariants for content-based image retrieval

This paper presents a study about the application of Zernike invariants to content-based Image Retrieval for 2D color images. Zernike invariants have been chosen because of their good performance for object recognition. Taking into account the good results achieved in previous CBIR experiments with color based primitives using a multiresolution representation of the visual contents, this paper presents the application of a wavelet transform to the images in order to obtain a multiresolution representation of the shape based features studied. Experiments have been performed using two databases: the first one is a small self-made 2D color database formed by 298 RGB images and a test set with 1655 query images that has been used for preliminary tests; the second one is Also experiments using the Amsterdam Library of Object Images (ALOI), a free access database. Experimental results show the feasibility of this new approach.

Shot boundary detection using Zernike moments in multi-GPU multi-CPU architectures

This paper presents an analysis of a Multi-GPU Multi-CPU environment, along with the different possible hybrid combinations. The analysis has been performed for a shot boundary detection application, based on Zernike moments, although it is general enough to be applied to many different application areas. A deep study of the performance, bottlenecks and design challenges is carried out showing the validity of this approach and achieving very high frame per second rates. In this paper, Zernike calculations are carried out on GPUs, taking advantage of a packing strategy proposed to minimize host-device communication time.

CBIR on grids

From the computational point of view, Content-based Im- age Retrieval systems are potentially expensive and have user response times growing with the ever-increasing sizes of the databases associated to them This paper presents a grid implementation of a Content-based Image Retrieval system that offers a good cost/performance ratio to solve this problem due to their excellent flexibility, scalability and fault tolerance This approach allows a dynamic management of specific data- bases that can be incorporated to or removed from the CBIR system in function of the desired user query Experimental performance results are collected in order to show the feasibility of this solution.

A load index and load balancing algorithm for heterogeneous clusters

This paper presents a load balancing algorithm specifically designed for heterogeneous clusters, composed of nodes with different computational capabilities. The method is based on a new index, which takes into consideration two levels of processors heterogeneity: the number of cores per node and the computational power of each core. The experimental results show that this index allows achieving balanced workload distributions even on those clusters where heterogeneity can not be neglected.

Combining activity and temporal coherence with low-level information for summarization of video rushes

This paper describes the work performed by the GMRV-URJC team as part of the TRECVid 2008 Rushes Summarization benchmark. The main goal of our approach is to obtain good results by only using low-level techniques. Using only this kind of features also has the advantage of achieving fast processing times. The work presented in this paper is a keyframe-based approach which means that all the process is built around keyframes. The proposed solution can be decomposed in three stages which are the candidate selection, candidate filtering and summary construction. We present an analysis of the results achieved in the official TRECVid tests as well as in some additional runs.

Technical Section: MSRS: A fast linear solver for the real-time simulation of deformable objects

Nowadays many interactive computer graphics applications deal with soft-bodies. These applications are always demanding higher and higher levels of realism, increasing the complexity of the animated scenarios. Therefore, the techniques used to animate the objects should be very efficient to grant interactivity. Most often, an accurate answer is not needed and a visually plausible one is enough. One of the most important issues in physically based simulations is to keep the model stable under all simulation conditions. In this paper we present a technique called matrix system reduction solver (MSRS) to highly accelerate the real-time simulation of a finite element-based technique. Our method is fast, visually plausible and robust. In the first stage of the algorithm an initial solution is estimated. Then, only those equations showing a large error are solved using a more precise technique. In the final stage the linear and angular momenta are corrected. Experimental results show the feasibility of this new approach from the point of view of performance compared to the results provided by the stiffness warping method.

Efficient Grid-Based Video Storage and Retrieval

Different fields, such as news broadcasting (news, shows, series, etc), advertising, and medical applications, require to store a large amount of video data which can be used later on. Content-Based Video Retrieval (CBVR) systems are very attractive since they can help users to retrieve video sequences over large video databases with respect to some specific topic, character or place. In most cases, stored videos belong to different organizations and have a quite demanding storage capacity. These features fit in a natural way into the concept of grid computing, that provides a great computing and storage capacity thanks to the use of heterogeneous resources put together by the cooperation and resource sharing among different institutions. This paper presents a Grid and Content-based VIdeo Retrieval (GCViR) system that offers a good cost/performance ratio to select the most suitable grid resources for data storage in order to store and retrieve large video data providing flexibility and scalability. An evaluation shows the feasibility and benefits of this approach.

Video shot extraction on parallel architectures

One of the main objectives of Content-based Multimedia Retrieval systems is the automation of the information extraction process from raw data. When dealing with video data, the first step is to perform a temporal video segmentation in order to make a shot decomposition of the video content. From a computational point of view, this is a very high demanding task and algorithm optimization must be seeked. This paper presents a comparison between two different parallel programming paradigms: shared-memory communication and distributed memory processing using the message passing paradigm. Taking into account the software solutions, experimental results are collected over two alternative parallel architectures: a shared-memory symmetric multiprocessor and a cluster. This paper analyzes the performance achieved from the viewpoints of speed and scalability.