-
[show abstract]
[hide abstract]
ABSTRACT: Multimedia data is produced in massive quantities. The applications that drive this massive data influx today span a large spectrum, from entertainment, surveillance, and e-commerce to Web content and social media. This data flood brings forth a need for highly parallelizable frameworks for scalable processing and efficient analysis of large media collections.In this article, we have observed that a major obstacle in executing large media-analysis operations in a scalable manner is that, when implemented naively, a significant portion of the work might be wasted. RanKloud is a system we are developing for share-nothing, batched parallelization of large media-analysis operations in a waste-avoiding manner. RanKloud achieves this by treating the utility of the data and features to the given analysis as an integral part of the process and by developing data partitioning and resource strategies that consider the ranked semantics of the analysis operations as well as data and utility characteristics discovered at runtime. Future work will include extension of the RanKloud framework with new primitives and investigation of the application of RanKloud to various large-scale multimedia data analysis applications.
IEEE Multimedia 02/2011; · 0.44 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: Complex media fusion operations can be costly in terms of the time they need to process input objects. If data arrive faster to fusion nodes than the speed with which they can consume the inputs, this will result in some input objects not being processed. In this paper, we develop load shedding mechanisms which take into consideration both data quality and expensive nature of media fusion operators. In particular, we present quality assessment models for objects and multistream fusion operators and highlight that such quality assessments may impose partial orders on objects. We highlight that the most effective load control approach for fusion operators involves shedding of (not the individual input objects but) combinations of objects. Yet, identifying suitable combinations of objects in real time will not be possible if efficient combination selection algorithms do not exist. We develop efficient combination selection schemes for scenarios with different quality assessment and target characteristics. We first develop efficient combination-based load shedding when the fusion operator has unambiguously monotone semantics. We then extend this to the more general ambiguously monotone case and present experimental results that show the performance gains using quality-aware combination-based load shedding strategies under the various fusion scenarios.
IEEE Transactions on Knowledge and Data Engineering 02/2010; · 1.66 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: Most of the Web interfaces are primarily designed for people with sight, with visually rich features that makes effective use of the tools to enhance visual usability but in process making it impossible for users who are blind or visually impaired to use them. In this work, our goal is to improve participation to NSF's National Science Digital Library (NSDL) by teachers, librarians, and learners who are blind. The middleware for accessible information spaces on NSDL (MAISON) is enhancing the accessibility of NSDL, its internal and external resources and existing services (such as strand maps of educational benchmarks). Relying on cutting-edge, context-aware graph segmentation, filtering and summarization, and concept propagation techniques, the middleware provides information space adaptation, reduction, and preview services through open Web-based service APIs to enable implementation of informative navigation interfaces that are able to reduce the complexity of the information space and provide previews to prevent user disorientation.
Multimedia and Expo, 2009. ICME 2009. IEEE International Conference on; 08/2009
-
[show abstract]
[hide abstract]
ABSTRACT: The tasks in the physical environments are mainly information centric processes, such as search and exploration of physical objects. We have developed an informational environment, AURA that supports object searches in the physical world. The goal of AURA is to enable individuals to use the environment in which they function as a living (short-term) memory of their activities and of the objects with which they interact in this environment. To support physical searches, the environment that the user is occupying must be transparently embedded with relevant information and made accessible by in-situ search mechanisms. We achieve this through innovative algorithms that re-imagine a collection of environmentally distributed RFID tags to act as a distributed storage cloud that encodes the required information for attribute-based object search. Since RFID tags lack radio transmitters and, thus, cannot communicate among each other, auraProp and auraSearch leverage the movements of the humans in the environment to propagate information: as they move in the environment, users not only leave traces (or auras) of their own activities, but also help further disseminate auras of prior activities in the same space. This scheme creates an information-gradient in the physical environment which AURA then leverages to direct the user toward the object of interest. auraSearch significantly reduces the number of steps that the user has to walk while searching for a given object.
Mobile Data Management: Systems, Services and Middleware, 2009. MDM '09. Tenth International Conference on; 06/2009
-
[show abstract]
[hide abstract]
ABSTRACT: In this paper, we introduce AURA, a novel framework for enriching the physical environment with information about objects and activities in order to support searches in the physical world. The goal is to enable individuals to use the environment in which they function as a living (short-term) memory of their activities and of the objects with which they interact in this environment. In order to act as a memory, the physical environment must be transparently embedded with relevant information and made accessible by in-situ search mechanisms. We achieve this embedding through innovative algorithms that leverage a collection of parasitic RFID tags distributed in the environment to act as a distributed storage cloud. Information about the activities of the users and objects with which they interact are encoded and stored, in a decentralized way, on these RFID tags to support attribute-based search. A novel auraProp algorithm disseminates information in the environment and a complementary auraSearch algorithm implements spatial searches for physical objects in the environment. Parasitic RFID tags are not self-powered and thus cannot communicate among each other. AURA leverages human movement in the environment to propagate information: as they move in the environment, users not only leave traces (or auras) of their own activities, but also help further disseminate auras of prior activities in the same space. AURA relies on a novel signature based information dissemination mechanism and a randomized information erasure scheme to ensure that the extremely limited storage spaces available on the RFID tags are used effectively. The erasure scheme also helps create an information gradient in the physical environment, which the auraSearch algorithm uses to direct the user towards the object of interest.
Data Engineering, 2009. ICDE '09. IEEE 25th International Conference on; 05/2009
-
[show abstract]
[hide abstract]
ABSTRACT: Media-rich ubiquitous distributed media processing workflow systems continuously sense users' needs, status, and the context, filter and fuse a multitude of real-time media data, and react by adapting the environment to the user. One challenge facing these systems is that they need to process real-time data arriving continuously from the sensors and data rates and qualities may vary dramatically. Thus, reducing the amount of unqualified data objects that need to be processed within the underlying media processing workflows can enhance the performance significantly. In this paper, we first focus on the prediction of the output qualities at the actuator, especially in the presence of fusion operators in the workflow. We then present quality-aware early object elimination schemes to enable informed resource savings in continuous real-time media processing workflow middleware.
Multimedia and Expo, 2007 IEEE International Conference on; 08/2007
-
[show abstract]
[hide abstract]
ABSTRACT: In situ wireless sensor networks, not only have to route sensed data from sources to destinations, but also have to filter and fuse observations to eliminate potentially irrelevant data. If data arrive faster to such fusion nodes than the speed with which they can consume the inputs, this will result in an overflow of input buffers. In this paper, we develop load shedding mechanisms which take into consideration both data quality and expensive nature of fusion operators. In particular, we present quality assessment models for objects and fusion operators and we highlight that such quality assessments may impose partial orders on objects.
Data Engineering, 2007. ICDE 2007. IEEE 23rd International Conference on; 05/2007
