HAL-UNILIM
Not a member yet
39506 research outputs found
Sort by
High-temperature thermal conductivity of yttrium and rare-earth iron garnets
No full textInternational audienceYttrium and rare-earth iron garnets (R3Fe3O12) are ferrimagnetic insulators that have been widely studied for magnetic and spintronic applications. In this study, we report the thermal conductivity (κ) between 300 and 773 K for the single crystals of R3Fe3O12, where R = Y, Gd, Dy, and Yb. For Y3Fe3O12, the κ up to the Curie temperature (TRC ≈ 555 K) can be described well with a pure phononic model, without considering conduction or scattering by the magnons. The iron garnets containing magnetic rare-earth ions exhibit smaller κ, with Dy3Fe3O12 showing the smallest values due to the strong interactions of heat-carrying phonons with the crystal field excitations of Dy3+ ions
A Hybrid Deep Learning and Knowledge Graph Approach for Intelligent Image Indexing and Retrieval
No full textInternational audienceTechnological advancements have enabled users to digitize and store an unlimited number of multimedia documents, including images and videos. However, the heterogeneous nature of multimedia content poses significant challenges in efficient indexing and retrieval. Traditional approaches primarily focus on visual features, often neglecting the semantic context, which limits retrieval efficiency. This paper proposes a hybrid deep learning and knowledge graph approach for intelligent image indexing and retrieval. By integrating deep learning models such as EfficientNet and Vision Transformer (ViT) with structured knowledge graphs, the proposed framework enhances semantic understanding and retrieval performance. The methodology incorporates feature extraction, concept classification, and hierarchical knowledge graph structuring to facilitate effective multimedia retrieval. Experimental results on benchmark datasets, including TRECVID, Corel, and MSCOCO, demonstrate significant improvements in precision, robustness, and query expansion techniques. The findings highlight the potential of combining deep learning with knowledge graphs to bridge the semantic gap and optimize multimedia indexing and retrieval
Participation à la table-ronde « La sociologie dans l'espace médiatique : critiques, réponses et enjeux de la diffusion de la connaissance sociologique »
No full textInternational audienceAutres participants : Emmanuelle Carinos Vasquez, docteure en sociologie, CRESPPA, Journaliste à l’Abcdr du Son et Romaric Godin, Journaliste à Mediapart, auteur, co-directeur de la collection Économie politique (La Découverte).Animation : Romain Renier, doctorant en sociologie, et Séléna Chauré, doctorante en sociologie, GRESCO, Université de Limoges
Tungsten implantation of VO2 nanobeams for neuromorphic applications
No full textInternational audienc
Remote Current Sensing Using Reflectometry for Bioelectric Applications
No full textInternational audienc
L'acquisition de la compréhension et de la production des pronoms personnels chez les enfants francophones : vers la capacité à incarner les perspectives des personnages dans diverses interactions verbales illustrées
No full textInternational audienceLanguage acquisition involves the ability to switch perspectives according to the partners’ roles in the speech context. This ability, involving morphosyntactic and pragmatic aspects, is fundamental in the acquisition process of personal pronouns (PersP). Our aim is to analyze, in spoken French, the acquisition of comprehension and production of PersP according to different perspectives (1p, 2p, 3p) and functions [subject (S); direct object (DO) and indirect object (IO) clitics]. 110 children aged from 6 to 13 participated. Comprehension and production of PersP were evaluated through an experimental protocol consisting of various ecological situations of the children’s daily life, in a comic strip format. Sentences had two levels of difficulty: D1, involving S and DO; and D2 involving S, DO and IO PersP. We found an age effect on the scores for the two tasks. Scores for expected answers were higher in the comprehension (1) than in the production task (2). We identified a significant increase in the mean scores, after an estimated breakpoint of 8 years for (1), and after the two estimated breakpoints of 8 and 10 years for (2). In (2) we found significantly better performances for D1 than for D2 sentences. In D1, unexpected answers occur more significantly on the DO than on the S, and in D2 they occur first on the DO, then on the IO and finally on the S. In D1 sentences, the DO is pronominalized but challenging for gender; in D2 sentences it is mainly omitted, like for IO. After the age of 6, children’s performances in the acquisition of PersP’s comprehension and production increase with age. At the age of 8 for the comprehension task, and at the ages of 8 and 10 for the production task we identified a significant increase in the children’s performances. In this last task, the difficulties could be explained by its computational complexity in terms of morphosyntactic and pragmatic constraints The difficulties were focused on the use of object pronouns and mainly DO. We suggest that, in D2 sentences, children focus primarily on the partners of the interaction involved in the speech context (designated by S and IO), rather than on the object of the interaction (DO)
Influence of fiber direction on the tribological behavior of carbon reinforced PEEK for application in gas turbine engines
No full textInternational audienceThe conditions under which materials are expected to perform in the next generation of gas turbine engines are becoming increasingly challenging. Consequently, proper material selection is essential for extending the service life of engine components. Polymer matrix composites (PMCs) offer an alternative to traditional metals and alloys due to their high strength-to-weight ratios; however, their use in harsh environments is still limited. Previous studies have extensively characterized carbon-reinforced polymer composites, highlighting the effects of parallel and anti-parallel fiber orientations on their mechanical properties and tribological behavior. However, recent advancements in PMC manufacturing have enabled the production of carbon-reinforced polyetheretherketone (PEEK) with fibers oriented normal to the surface. While this innovative technology shows promise with respect to mechanical properties, the tribological behavior of PMCs featuring normal fiber orientation has not been thoroughly investigated. This study aims to enhance the understanding of how fiber orientation influences the tribological performance of carbon-reinforced polymers. Friction and wear tests were conducted on CF-PEEK films with fibers aligned in the normal direction, and the results were compared to those of standard modulus carbon fiber/PEEK unidirectional tape (UD) and unreinforced, pure PEEK films under various contact conditions. Worn surface characterization was performed through ex situ analysis, utilizing techniques such as scanning electron microscopy (SEM), confocal laser scanning microscopy (CLSM), and atomic force microscopy (AFM) to reveal interfacial phenomena. A correlation between fiber orientation and wear and friction performance was established based on the different experimental parameters
Effect of BaO/Al2O3 ratio on crystallization and properties of glass-ceramic sealants
No full textInternational audienc
Contributions à l'indexation et à la recherche d'information : application aux données multimodales généralistes et médicales
No full textThis university habilitation summarizes my contributions in the field of content-based information indexing and retrieval, applied to both general and medical data. It is structured into four chapters. The first chapter explores multimodal indexing systems that combine visual, audio, and textual information. The second chapter focuses on semantic multimedia retrieval at multiple levels, incorporating deep learning techniques. The third chapter presents my work on integrating AI for managing diabetic foot ulcers, through systems like DFU-SIAM and DFU-Helper, which assist clinicians in monitoring and decision-making. Finally, the fourth chapter is dedicated to applying AI in the analysis of thoracic radiographs, with the development of a CBMIR (Content-Based Medical Image Retrieval) model based on deep learning, aimed at automatically identifying pulmonary diseases. These contributions aim to bridge the gap between low-level descriptors and their semantic interpretation, while developing innovative solutions for medical imaging and disease monitoring.Cette habilitation universitaire retrace mes contributions dans le domaine de l'indexation et de la recherche d'information par le contenu, appliquées aux données généralistes et médicales. Elle est structurée en quatre chapitres. Dans le premier, j'explore les systèmes d'indexation multimodale qui combinent des informations visuelles, sonores et textuelles. Le second chapitre se concentre sur la recherche sémantique multimédia à plusieurs niveaux, incluant l’apprentissage profond. Le troisième chapitre présente mes travaux sur l'intégration de l'IA pour la gestion des ulcères du pied diabétique, à travers des systèmes comme DFU-SIAM et DFU-Helper, qui assistent les cliniciens dans le suivi et la prise de décision. Enfin, le quatrième chapitre est dédié à l’application de l’IA dans l’analyse des radiographies thoraciques, avec la conception d’un modèle CBMIR basé sur le deep learning, pour l’identification automatique des pathologies pulmonaires. Ces contributions visent à combler le fossé entre les descripteurs bas-niveau et leur interprétation sémantique, tout en développant des solutions innovantes pour l'imagerie médicale et le suivi de maladies
