1,721,032 research outputs found
PAC nearest neighbor queries: Approximate and controlled search in high-dimensional and metric spaces
No full textIn high-dimensional and complex metric spaces, determining the nearest neighbor (NN) of a query object q can be a very expensive task, because of the poor partitioning operated by index structures - the so-called `curse of dimensionality'. This also affects approximately correct (AC) algorithms, which return as result a point whose distance from q is less than (1+ε) times the distance between q and its true NN. In this paper we introduce a new approach to approximate similarity search, called PAC-NN queries, where the error bound ε can be exceeded with probability δ and both ε and δ parameters can be tuned at query time to trade the quality of the result for the cost of the search. We describe sequential and index-based PAC-NN algorithms that exploit the distance distribution of the query object in order to determine a stopping condition that respects the error bound. Analysis and experimental evaluation of the sequential algorithm confirm that, for moderately large data sets and suitable ε and δ values, PAC-NN queries can be efficiently solved and the error controlled. Then, we provide experimental evidence that indexing can further speed-up the retrieval process by up to 1-2 orders of magnitude without giving up the accuracy of the result
WARP: Accurate Retrieval of Shapes Using Phase of Fourier Descriptors and Time Warping Distance
No full textEffective and efficient retrieval of similar shapes from large image databases is still a challenging problem in spite of the high relevance that shape information can have in describing image contents. In this paper, we propose a novel Fourier-based approach, called WARP, for matching and retrieving similar shapes. The unique characteristics of WARP are the exploitation of the phase of Fourier coefficients and the use of the Dynamic Time Warping (DTW) distance to compare shape descriptors. While phase information provides a more accurate description of object boundaries than using only the amplitude of Fourier coefficients, the DTW distance permits us to accurately match images even in the presence of (limited) phase shiftings. In terms of classical precision/recall measures, we experimentally demonstrate that WARP can gain, say, up to 35 percent in precision at a 20 percent recall level with respect to Fourier-based techniques that use neither phase nor DTW distance
Windsurf: the best way to SURF: (and SIFT/BRISK/ORB/FREAK, too).
No full textDespite their popularity, approaches based on salient point descriptors have yet to be proven effective for content-based image retrieval. In this paper, we show how the Windsurf library can be effectively exploited to assess a fair comparison among the existing alternative approaches based on salient points, which can be contrasted on aspects of both effectiveness and efficiency. Our extensive experimental evaluation, performed on four different image benchmarks, indeed, shows that techniques based on salient point descriptors have effectiveness not better than other existing techniques and are less amenable to be indexed, and thus, their efficiency remains questionable
Comparing performances of big data stream processing platforms with RAM3S
No full textNowadays, Big Data platforms allow the analysis of massive data streams in an efficient way. However, the services they provide are often too raw, thus the implementation of advanced real-world applications requires a non-negligible effort for interfacing with such services. This also complicates the task of choosing which one of the many available alternatives is the most appropriate for the application at hand. In this paper, we present a comparative study of the three major open-source Big Data platforms for stream processing, as performed by using our novel RAM^3S framework. Although the results we present are specific for our use case (recognition of suspect people from massive video streams), the generality of the RAM^3S framework allows both considering such results as valid for similar applications and implementing different use cases on top of Big Data platforms with very limited effort
A general framework for real-time analysis of massive multimedia streams.
No full textBig Data platforms provide opportunities for the management and analysis of large quantities of information, but the services they provide are often too raw, since they focus on issues of fault-tolerance, increased parallelism, and so on. An additional software layer is, therefore, needed to effectively use such architectures for advanced applications in several important real-world domains, such as scientific and health care sensors, user-generated data, supply chain systems and financial companies, to name a few. In this paper, we present RAM(Formula presented.)S, a framework for the real-time analysis of massive multimedia streams, where data come from multiple data sources (such as sensors and cameras) that are widely located on the territory, with the final goal to discovery new and hidden information from the output of data sources as they occur, thus with very limited latency. We apply RAM^3S to the use case of automatic detection of suspected people from several concurrent video streams, and instantiate it on top of three different open source engines for the analysis of streaming Big Data (i.e., Apache Spark, Apache Storm, and Apache Flink). The effectiveness and scalability of RAM^3S instantiation is experimentally evaluated on real data, also comparing the performance of the three considered Big Data platforms. Such comparison is performed both on a cluster of physical machines in our datalab and on the Google Cloud Platform
Data Mining and Machine Learning for Condition-based Maintenance
Full text linkComplex production systems may count thousands of parts and components, subjected to multiple physical and logical connections and interdependencies. This level of complexity inhibits the traditional and statistically-based approach to reliability engineering, failure prediction and maintenance planning. The existing ICT solutions simplify the on-field collection of large amount of data, but require models and tools able to create knowledge from these data. Key questions on how to predict in advance the performance of the production system and the associated failure events could be finally addressed. This paper introduces a set of data analytics models and methods that can be profitably used for decision making in general, and, specifically, in maintenance engineering. These classification models, specifically decision trees, random forests, and neural networks, are applied to a real-world case study, and the resulting accuracy on predicting faults is quantified and compared. We used the historical profiles of the energy variables of an high-speed packaging machine to find out some strategies for the prediction of a given failure. The conducted experiments demonstrate that the accuracy of the random forest is slightly better than the other methods, but even increases the probability of false alarm, which would result in unwanted production break-down. Even though the obtained results are promising, they leave room for further experiments based on the application of other classifiers, rather than the definition of customized methods able to embrace such complexity
A stream processing abstraction framework
Full text linkReal-time analysis of large multimedia streams is nowadays made efficient by the existence of several Big Data streaming platforms, like Apache Flink and Samza. However, the use of such platforms is difficult due to the fact that facilities they offer are often too raw to be effectively exploited by analysts. We describe the evolution of RAM3S, a software infrastructure for the integration of Big Data stream processing platforms, to SPAF, an abstraction framework able to provide programmers with a simple but powerful API to ease the development of stream processing applications. By using SPAF, the programmer can easily implement real-time complex analyses of massive streams on top of a distributed computing infrastructure, able to manage the volume and velocity of Big Data streams, thus effectively transforming data into value
The Power of Distance Distributions: Cost Models and Scheduling Policies for Quality-Controlled Similarity Queries
No full textApproximate similarity queries are a practical way to obtain good, yet suboptimal, results from large data sets without having to pay high execution costs. In this paper we analyze the problem of understanding how the strategy for searching through an index tree, also called scheduling policy, can influence costs. We consider quality-controlled similarity queries, in which the user sets a quality (distance) threshold \theta ̧ and the system halts as soon as it finds k objects in the data set at distance \theta ̧ from the query object. After providing experimental evidence that the scheduling policy might indeed have a high impact on paid costs, we characterize the policies' behavior through an analytical cost model, in which a major role is played by parameterized local distance distributions. Such distributions are also the key to derive new scheduling policies, which we show to be optimal in a simplified, yet relevant, scenario
Multimedia Queries in Digital Libraries
No full textThe intrinsic complexity and diversity of data in multimedia digital libraries (MDLs) require devising techniques and solutions that are inherently different from those usually adopted in traditional information retrieval and database (DB) systems. Moreover, the size and the dynamicity of MDLs force researchers to strive for efficiency, so as to guarantee real-time results to the users. Finally, semantics should be also brought into context, in order to facilitate users’ experience in querying, browsing, and consuming multimedia information. This chapter will present an approach toward the efficient, effective, and semantically rich data retrieval in MDLs. With respect to the commonly used holistic approach, where the multimedia datum is considered as an atomic entity, our reductionist strategy considers the multimedia information as a complex combination of component subparts and eases the fulfillment of the three above properties of efficiency, effectiveness, and semantic richness. Indeed, by decomposing multimedia information into simpler and smaller component objects, we are able to index such components without giving up the ability of querying the original information as a whole
- …
