HAL - Lille 3
Not a member yet
5967 research outputs found
Sort by
A Novel Space-Time Representation on the Positive Semidefinite Cone for Facial Expression Recognition
No full textInternational audienceIn this paper, we study the problem of facial expression recognition using a novel space-time geometric representation. We describe the temporal evolution of facial landmarks as parametrized trajectories on the Riemannian man-ifold of positive semidefinite matrices of fixed-rank. Our representation has the advantage to bring naturally a second desirable quantity when comparing shapes – the spatial covariance – in addition to the conventional affine-shape representation. We derive then geometric and computational tools for rate-invariant analysis and adaptive re-sampling of trajectories, grounding on the Riemannian geometry of the manifold. Specifically, our approach involves three steps: 1) facial landmarks are first mapped into the Riemannian manifold of positive semidefinite matrices of rank 2, to build time-parameterized trajectories; 2) a temporal alignment is performed on the trajectories, providing a geometry-aware (dis-)similarity measure between them; 3) finally, pairwise proximity function SVM (ppfSVM) is used to classify them, incorporating the latter (dis-)similarity measure into the kernel function. We show the effectiveness of the proposed approach on four publicly available benchmarks (CK+, MMI, Oulu-CASIA, and AFEW). The results of the proposed approach are comparable to or better than the state-of-the-art methods when involving only facial landmarks
Sista: Saving Optimized Code in Snapshots for Fast Start-Up
No full textInternational audienceModern virtual machines for object-oriented languages such as Java HotSpot, Javascript V8 or Python PyPy reach high performance through just-in-time compilation techniques, involving on-the-fly optimization and deoptimization of the executed code. These techniques require a warm-up time for the virtual machine to collect information about the code it executes to be able to generate highly optimized code. This warm-up time required before reaching peak performance can be considerable and problematic. In this paper, we propose an approach, Sista (Speculative Inlining SmallTalk Architecture) to persist optimized code in a platform-independent representation as part of a snapshot. After explaining the overall approach, we show on a large set of benchmarks that the Sista virtual machine can reach peak performance almost immediately after start-up when using a snapshot where optimized code was persisted
Effets macroéconomiques de la mutualisation de la dette en union monétaire
No full textInternational audienc
TOWARD AN ADVANCED INFORMATIONAL FRAMEWORK FOR IMPROVING MINIMALISTIC VISUAL ODOMETRY
No full textInternational audienc
Effective Lyapunov Level Set for Nonlinear Optimal Control. Application to Turbocharged Diesel Engine Model
No full textInternational audienceWe develop an optimization algorithm in order to improve the nonlinear optimal robust controller for diesel engine which equipped by Exhaust Gas Recirculation and Variable Geometry Turbocharger systems. For this goal we present a new way to search the optimal shape of control Lyapunov function which permits to calculate the parameters to achieve stabilization and optimal control criteria. The idea is to verify different form of Lyapunov functions in order to desire better performance
INCREMENTAL CONDUCTANCE ALGORITHM BASED ON T-S FUZZY LOGI1 C 2 CONTROLLER FOR PHOTOVOLTAIC SYSTEM MPPT
No full textInternational audienc
Distributed formation control of fractional-order multi-agent systems with relative damping and communication delay
No full textInternational audienc
The Smart Home for the Elderly: Perceptions, Technologies and Psychological Accessibilities : The Requirements Analysis for the Elderly in Thailand
No full textInternational audienceVarious studies of assistive technology, embedded into the house for helping elderly persons have been introduced. There are many terms used to call this assistive technology, such as smart home, intelligent home, and home networking. The current technologies for smart home system can be divided into two basic categories: passive-intervention devices and active-intervention devices. Passive-intervention devices monitor an elderly person's condition and safety without intervening in his or her care, such as detecting heart rate, respiration, and restlessness by using electrode-slipping device, inserted under mattress pads. Active-intervention devices play more active role in an elderly person care, such as all sensors equipped with alarms, reminder systems, and medication administration system. The smart home technologies enhance a quality of living and allow the elderly to live independently in their homes, while their adult children are working. A successful promotion of smart home technologies does not depend only on the intelligence itself, but it also requires needs, concerns, perceptions, and psychological accessibilities for elderly persons. This paper proposes the smart home model for the elderly in Thailand, based on the stakeholders' requirements, concerns, and perceptions. By given the selected technologies in passive-intervention devices and active-intervention devices, this paper shows the perceptions, psychological accessibility factors, needs, and concerns of specific smart home model for elderly care, using a questionnaire and survey. The interpreting of the results of statistical analysis relies on both qualitative and quantitative methods, therefore the smart home model for the elderly in Thailand can be designed and implemented, practically based on the stakeholders' perspectives
Solutions d'amélioration des études de métagénomique ciblée
No full textTargeted metagenomics is the study of the composition of microbial communities in diverse biological samples, based on the sequencing of a genomic locus. This application has boomed over the last decade thanks to the democratisation of high-throughput sequencing, and has allowed substantial progress in the study of microbial evolution and diversity. However, new problems have emerged with high-throughput sequencing : the exponential generation of data must be properly analyzed with bioinformatics tools fitted to the experimental designs and associated biological questions. This dissertation provides solutions to improve targeted metagenomics studies, by the development of new tools and methods allowing a better understanding of analytical biases, and a better design of experiments. Firstly, an expert assessment of the analytical pipeline used on the PEGASE-biosciences plateform has been performed. This assessment revealed the need of a formal evaluation method of analytical pipelines used for targeted metagenomics analyses. This method has been developed with simulated and real datasets, and adequate evaluation metrics. It has been used on several analytical pipelines commonly used by the scientific community, as well as on new analytical methods which have never been used in such a context before. This evaluation allowed to better understand experimental design biases, which can affect the results and biological conclusions. One of those major biases is the design of amplification primers to target the genomic locus of interest. A primer design software, adaptable to different experimental designs, has been specifically developed to minimize this bias. Finally, analytical guidelines and experimental design recommendations have been formulated to improve targeted metagenomics studies.La métagénomique ciblée, étude de la composition et de la diversité des communautés microbiennes présentes dans différents échantillon biologiques sur la base d'un marqueur génomique, a connu un véritable essor lors de cette dernière décennie grâce à l'arrivée du séquençage haut-débit. Faisant appel à des outils de biologie moléculaire et de bioinformatique, elle a été à l’origine de substantiels progrès dans les domaines de l’évolution et de la diversité microbienne. Cependant, de nouvelles problématiques sont apparues avec le séquençage haut-débit : la génération exponentielle de données soulève des problèmes d'analyse bioinformatique, qui doit être adaptée aux plans d'expérience et aux questions biologiques associées. Cette thèse propose des solutions d'amélioration des études de métagénomique ciblée par le développement d'outils et de méthodes innovantes, apportant une meilleure compréhension des biais d'analyse inhérents à de telles études, et une meilleure conception des plans d'expérience. Tout d'abord, une expertise du pipeline d'analyse utilisé en production sur la plate-forme PEGASE-biosciences a été menée. Cette évaluation a révélé la nécessité de mettre en place une méthode d'évaluation formelle de pipelines d'analyses de données de métagénomique ciblée, qui a été développée sur la base de données simulées et réelles, et de métriques d'évaluation adaptées. Cette méthode a été utilisée sur plusieurs pipelines d'analyse couramment utilisés par la communauté, tout comme sur de nouvelles approches d'analyse jamais utilisées dans un tel contexte. Cette évaluation a permis de mieux comprendre les biais du plan d'expérience qui peuvent affecter les résultats et les conclusions biologiques associées. Un de ces biais majeurs est le choix des amorces d'amplification de la cible ; un logiciel de design d'amorces adaptées au plan d'expérience a été spécifiquement développé pour minimiser ce biais. Enfin, des recommandations de montage de plan d'expérience et d'analyse ont été émises afin d'améliorer la robustesse des études de métagénomique ciblée