Laboratoire d'Informatique Fondamentale de Lille

Extending beyond the boundaries of science, art, and culture, content-based multimedia information retrieval provides new paradigms and methods for searching through the myriad variety of media all over the world. This survey reviews 100+ recent articles on content-based multimedia information retrieval and discusses their role in current research directions which include browsing and search paradigms, user studies, affective computing, learning, semantic queries, new features and media types, high performance indexing, and evaluation techniques. Based on the current state of the art, we discuss the major challenges for the future.

International audience

Vehicular networks are likely to become the most relevant form of mobile ad hoc networks. In this paper, we address the security of these networks. We provide a detailed threat analysis and devise an appropriate security architecture. We also describe some major design decisions still to be made, which in some cases have more than mere technical implications. We provide a set of security protocols, we show that they protect privacy and we analyze their robustness, and we carry out a quantitative assessment of the proposed solution.

In recent years, neural network accelerators have been shown to achieve both high energy efficiency and high performance for a broad application scope within the important category of recognition and mining applications.

Network coding is a new research area that may have interesting applications in practical networking systems. With network coding, intermediate nodes may send out packets that are linear combinations of previously received information. There are two main benefits of this approach: potential throughput improvements and a high degree of robustness. Robustness translates into loss resilience and facilitates the design of simple distributed algorithms that perform well, even if decisions are based only on partial information. This paper is an instant primer on network coding: we explain what network coding does and how it does it. We also discuss the implications of theoretical results on network coding for realistic settings and show how network coding can be used in practice

Large scale distributed systems such as Grids are difficult to study from theoretical models and simulators only. Most Grids deployed at large scale are production platforms that are inappropriate research tools because of their limited reconfiguration, control and monitoring capabilities. In this paper, we present Grid'5000, a 5000 CPU nation-wide infrastructure for research in Grid computing. Grid'5000 is designed to provide a scientific tool for computer scientists similar to the large-scale instruments used by physicists, astronomers, and biologists. We describe the motivations, design considerations, architecture, control, and monitoring infrastructure of this experimental platform. We present configuration examples and performance results for the reconfiguration subsystem.

We advocate the use of Gaussian Process Dynamical Models (GPDMs) for learning human pose and motion priors for 3D people tracking. A GPDM provides a lowdimensional embedding of human motion data, with a density function that gives higher probability to poses and motions close to the training data. With Bayesian model averaging a GPDM can be learned from relatively small amounts of data, and it generalizes gracefully to motions outside the training set. Here we modify the GPDM to permit learning from motions with significant stylistic variation. The resulting priors are effective for tracking a range of human walking styles, despite weak and noisy image measurements and significant occlusions.

The 1 € filter ("one Euro filter") is a simple algorithm to filter noisy signals for high precision and responsiveness. It uses a first order low-pass filter with an adaptive cutoff frequency: at low speeds, a low cutoff stabilizes the signal by reducing jitter, but as speed increases, the cutoff is increased to reduce lag. The algorithm is easy to implement, uses very few resources, and with two easily understood parameters, it is easy to tune. In a comparison with other filters, the 1 € filter has less lag using a reference amount of jitter reduction.

BACKGROUND: In general, the construction of trees is based on sequence alignments. This procedure, however, leads to loss of informationwhen parts of sequence alignments (for instance ambiguous regions) are deleted before tree building. To overcome this difficulty, one of us previously introduced a new and rapid algorithm that calculates dissimilarity matrices between sequences without preliminary alignment. RESULTS: In this paper, HIV (Human Immunodeficiency Virus) and SIV (Simian Immunodeficiency Virus) sequence data are used to evaluate this method. The program produces tree topologies that are identical to those obtained by a combination of standard methods detailed in the HIV Sequence Compendium. Manual alignment editing is not necessary at any stage. Furthermore, only one user-specified parameter is needed for constructing trees. CONCLUSION: The extensive tests on HIV/SIV subtyping showed that the virus classifications produced by our method are in good agreement with our best taxonomic knowledge, even in non-coding LTR (Long Terminal Repeat) regions that are not tractable by regular alignment methods due to frequent duplications/insertions/deletions. Our method, however, is not limited to the HIV/SIV subtyping. It provides an alternative tree construction without a time-consuming aligning procedure.

Recognizing human actions in 3-D video sequences is an important open problem that is currently at the heart of many research domains including surveillance, natural interfaces and rehabilitation. However, the design and development of models for action recognition that are both accurate and efficient is a challenging task due to the variability of the human pose, clothing and appearance. In this paper, we propose a new framework to extract a compact representation of a human action captured through a depth sensor, and enable accurate action recognition. The proposed solution develops on fitting a human skeleton model to acquired data so as to represent the 3-D coordinates of the joints and their change over time as a trajectory in a suitable action space. Thanks to such a 3-D joint-based framework, the proposed solution is capable to capture both the shape and the dynamics of the human body, simultaneously. The action recognition problem is then formulated as the problem of computing the similarity between the shape of trajectories in a Riemannian manifold. Classification using k-nearest neighbors is finally performed on this manifold taking advantage of Riemannian geometry in the open curve shape space. Experiments are carried out on four representative benchmarks to demonstrate the potential of the proposed solution in terms of accuracy/latency for a low-latency action recognition. Comparative results with state-of-the-art methods are reported.

<font color=#ff0000><b>BEST PAPER AWARD</b></font><br><br>In this paper, we propose and evaluate three techniques for optimizing network performance in the Xen virtualized environment. Our techniques retain the basic Xen architecture of locating device drivers in a privileged `driver' domain with access to I/O devices, and providing network access to unprivileged `guest' domains through virtualized network interfaces. First, we redefine the virtual network interfaces of guest domains to incorporate high-level network offfload features available in most modern network cards. We demonstrate the performance benefits of high-level offload functionality in the virtual interface, even when such functionality is not supported in the underlying physical interface. Second, we optimize the implementation of the data transfer path between guest and driver domains. The optimization avoids expensive data remapping operations on the transmit path, and replaces page remapping by data copying on the receive path. Finally, we provide support for guest operating systems to effectively utilize advanced virtual memory features such as superpages and global page mappings. The overall impact of these optimizations is an improvement in transmit performance of guest domains by a factor of 4.4. The receive performance of the driver domain is improved by 35% and reaches within 7% of native Linux performance. The receive performance in guest domains improves by 18%, but still trails the native Linux performance by 61%. We analyse the performance improvements in detail, and quantify the contribution of each optimization to the overall performance.

We propose a novel geometric framework for analyzing 3D faces, with the specific goals of comparing, matching, and averaging their shapes. Here we represent facial surfaces by radial curves emanating from the nose tips and use elastic shape analysis of these curves to develop a Riemannian framework for analyzing shapes of full facial surfaces. This representation, along with the elastic Riemannian metric, seems natural for measuring facial deformations and is robust to challenges such as large facial expressions (especially those with open mouths), large pose variations, missing parts, and partial occlusions due to glasses, hair, and so on. This framework is shown to be promising from both--empirical and theoretical--perspectives. In terms of the empirical evaluation, our results match or improve upon the state-of-the-art methods on three prominent databases: FRGCv2, GavabDB, and Bosphorus, each posing a different type of challenge. From a theoretical perspective, this framework allows for formal statistical inferences, such as the estimation of missing facial parts using PCA on tangent spaces and computing average shapes.

Norine is the first database entirely dedicated to nonribosomal peptides (NRPs). In bacteria and fungi, in addition to the traditional ribosomal proteic biosynthesis, an alternative ribosome-independent pathway called NRP synthesis allows peptide production. It is performed by huge protein complexes called nonribosomal peptide synthetases (NRPSs). The molecules synthesized by NRPS contain a high proportion of nonproteogenic amino acids. The primary structure of these peptides is not always linear but often more complex and may contain cycles and branchings. In recent years, NRPs attracted a lot of attention because of their biological activities and pharmacological properties (antibiotic, immunosuppressor, antitumor, etc.). However, few computational resources and tools dedicated to those peptides have been available so far. Norine is focused on NRPs and contains more than 700 entries. The database is freely accessible at http://bioinfo.lifl.fr/norine/. It provides a complete computational tool for systematic study of NRPs in numerous species, and as such, should permit to obtain a better knowledge of these metabolic products and underlying biological mechanisms, and ultimately to contribute to the redesigning of natural products in order to obtain new bioactive compounds for drug discovery.

Microblogging sites are a unique and dynamic Web 2.0 communica-tion medium. Understanding the information flow in these systems can not only provide better insights into the underlying sociology, but is also crucial for applications such as content ranking, recommendation and filtering, spam detection and viral marketing. In this paper, we characterize the propagation of URLs in the social network of Twitter, a popular microblogging site. We track 15 million URLs exchanged among 2.7 million users over a 300 hour period. Data analysis uncov-ers several statistical regularities in the user activity, the social graph, the structure of the URL cascades and the communication dynamics. Based on these results we propose a propagation model that predicts which users are likely to mention which URLs. The model correctly accounts for more than half of the URL mentions in our data set, while maintaining a false positive rate lower than 15%. 1

The proliferation of hotspots based on IEEE 802.11 wireless LANs brings the promise of seamless Internet access from a large number of public locations. However, as the number of users soars, so does the risk of possible misbehavior; to protect themselves, wireless ISPs already make use of a number of security mechanisms, and require mobile stations to authenticate themselves at the Access Points (APs). However, IEEE 802.11 works properly only if the stations also respect the MAC protocol. We show in this paper that a greedy user can substantially increase his share of bandwidth, at the expense of the other users, by slightly modifying the driver of his network adapter. We explain how easily this can be performed, in particular with the new generation of adapters. We then present DOMINO (System for Detection Of greedy behavior in the MAC layer of IEEE 802.11 public NetwOrks), a piece of software to be installed in the Access Point. DOMINO can detect and identify greedy stations, without requiring any modification of the standard protocol at the AP and without revealing its own presence. We illustrate these concepts by simulation results and by the description of our prototype.

In this work we compare the use of a particle swarm optimization (PSO) and a genetic algorithm (GA) (both augmented with support vector machines SVM) for the classification of high dimensional microarray data. Both algorithms are used for finding small samples of informative genes amongst thousands of them. A SVM classifier with 10- fold cross-validation is applied in order to validate and evaluate the provided solutions. A first contribution is to prove that PSOsvm is able to find interesting genes and to provide classification competitive performance. Specifically, a new version of PSO, called Geometric PSO, is empirically evaluated for the first time in this work using a binary representation in Hamming space. In this sense, a comparison of this approach with a new GAsvm and also with other existing methods of literature is provided. A second important contribution consists in the actual discovery of new and challenging results on six public datasets identifying significant in the development of a variety of cancers (leukemia, breast, colon, ovarian, prostate, and lung).

This paper presents a general class of gossip-based averaging algorithms, which are inspired from Uniform Gossip. While Uniform Gossip works synchronously on complete graphs, weighted gossip algorithms allow asynchronous rounds and converge on any connected, directed or undirected graph. Unlike most previous gossip algorithms, Weighted Gossip admits stochastic update matrices which need not be doubly stochastic. Double-stochasticity being very restrictive in a distributed setting, this novel degree of freedom is essential and it opens the perspective of designing a large number of new gossip-based algorithms. To give an example, we present one of these algorithms, which we call One-Way Averaging. It is based on random geographic routing, just like Path Averaging, except that routes are one way instead of round trip. Hence in this example, getting rid of double stochasticity allows us to add robustness to Path Averaging.

This paper presents the most exhaustive study of synchronization to date. We span multiple layers, from hardware cache-coherence protocols up to high-level concurrent software. We do so on different types of architectures, from single-socket -uniform and nonuniform -to multi-socket -directory and broadcastbased -many-cores. We draw a set of observations that, roughly speaking, imply that scalability of synchronization is mainly a property of the hardware.

Cloud computing in general, and Infrastructure-as-a-Service (IaaS) in particular, are becoming ever more popular. Unfortunately, performance interference (and the resulting unpredictability in the delivered performance) across virtual machines (VMs) co-located on the same physical machine (PM) threatens to make cloud computing inadequate for performance-sensitive customers and more expensive than necessary for all customers. We describe the design and implementation of DeepDive, a system for transparently identifying and managing interference. DeepDive successfully addresses several important challenges, including limiting overhead and requiring no performance information from applications. We first show that it is possible to use easily obtainable, low-level metrics to clearly discern when interference is occurring and what resource is causing it. Using realistic workloads, we demonstrate that DeepDive quickly learns about interference across co-located VMs. Moreover, we show DeepDive’s ability to deal efficiently with interference when it is detected, by using a low-overhead approach to selecting an alternative PM for a VM that is causing interference at its current PM.

In this paper, we propose a method for three-dimensional (3D)-model indexing based on two-dimensional (2D) views, which we call adaptive views clustering (AVC). The goal of this method is to provide an "optimal" selection of 2D views from a 3D model, and a probabilistic Bayesian method for 3D-model retrieval from these views. The characteristic view selection algorithm is based on an adaptive clustering algorithm and uses statistical model distribution scores to select the optimal number of views. Starting from the fact that all views do not have equal importance, we also introduce a novel Bayesian approach to improve the retrieval. Finally, we present our results and compare our method to some state-of-the-art 3D retrieval descriptors on the Princeton 3D Shape Benchmark database and a 3D-CAD-models database supplied by the car manufacturer Renault