Portail des publications scientifiques IMT Mines Alès
Not a member yet
5198 research outputs found
Sort by
Influence of rhythm features on beat/movement synchronization using a low-cost vision system
No full textInternational audienceThis study examines how musical expertise, tempo, and beat division influence synchronization accuracy and regularity in two movement tasks: finger tapping (discrete movements) and arm swing (continuous movements). Using a markerless motion capture system, we analyzed synchronization metrics across different rhythmic conditions. Motion data were extracted via AI-based pose estimation, and synchronization was computed by aligning movement peaks with beat times detected from audio stimuli. Results show that musicians exhibit higher synchronization accuracy and consistency than non-musicians, particularly in finger tapping tasks. Furthermore, simpler beat structures (binary rhythms) and moderate tempos facilitate better synchronization, whereas increased rhythmic complexity and tempo variability reduce performance. Interestingly, finger tapping leads to more precise synchronization than arm swing, suggesting that movement type significantly impacts rhythmic alignment. These findings support applications in therapy, training, and interactive systems, and demonstrate the value of AI-based motion tracking for scalable rhythm analysis
Gas Emissions during Smoldering in Biobased Insulation: Experimental Study of the Role of Wood Fiber Board Density
No full textInternational audienceBio-based insulating materials are increasingly used in construction due to their environmental benefits. However, these materials are particularly susceptible to smoldering fires, a phenomenon of slow, flameless and self-sustaining combustion, that is very difficult to detect. This study includes two experimental approaches to analyze smoldering fires in wood fiber boards with low (50 kg/m³) and high (140 kg/m³) densities. The tests were conducted using a cone calorimeter (ISO 5660-1) and a smoldering fire test device (EN 16733). The cone calorimeter uses small-sized samples with continuous thermal exposure, whereas the smoldering fire test bench involves discontinuous thermal exposure and larger sample dimensions. The objective is to better understand the differences in thermal behavior, gas emissions, and material degradation characteristics, while considering key factors influencing the propagation of smoldering fires, such as density or additives. Despite several studies addressing these key factors, a full understanding of the underlying mechanisms has yet to be achieved. The results show that the low-density wood fiber board degrades more rapidly, reaching high combustion temperatures in a shorter period. Incontrast, the high-density fiber board has a greater thermal inertia and prolonged combustion. Regarding gas emissions, concentrations of CO, CO₂, and methane vary depending on fiber density, with low-density samples producing higher yield of CO. These findings aim to enhance the understanding of the smoldering behavior of bio-based materials and to emphasize the importance of chemical aspects such as toxic gasemissions, The study contributes to inform future improvements in fire safety practices and may serve as a basis for revisiting or complementing existing fire safety guidelines. The results contribute to a better understanding of the smoldering combustion behavior of bio-based materials, particularly in relation to material density and its influence on fire dynamics. This knowledge is essential for informing practical fire safety strategies, such as early detection systems and material selection in low-ventilation environments and could help to refine fire performance assessment methods within the built environment
Thermally conductive composites based on h-BN and LDPE with low environmental impact
No full textInternational audienceAs the electrification of various applications continues to advance, the efficient dissipation of heat in electronic devices has become increasingly critical. For example, heat sinks are devices in electronic cards with high thermal conductivity (~ 5 W/m.K) that allow to evacuate heat. In a first phD thesis the effect of the process on the microstructure (and especially the orientation) of boron nitride platelets (h-BN) into LDPE matrix was studied. The final thermal conductivity reached 4.25 W/m·K with 50 wt.% of h-BN. In a second phD thesis, the objective is to elaborate high thermally conductive and environmentally friendly composites. To date, a life cycle assessment (LCA) of the production of 1kg of h-BN by carbothermic process was performed and compared to that of another thermally conductive particle and that of a LDPE polymer matrix. The next step will be to draw up and compare the LCA of the production of 1kg of several composites with a thermal conductivity of 4.25 W/m.K
Correlation Between Fire Tests on FR Polymers: What Can be Expected?
No full textInternational audienc
Dynamic Entity-Masked Graph Diffusion Model for Histopathological Image Representation Learning
No full textDOI en erreur : 10.1609/aaai.v39i10.33202International audienceSignificant disparities between the features of natural images and those inherent to histopathological images make it challenging to directly apply and transfer pre-trained models from natural images to histopathology tasks. Moreover, the frequent lack of annotations in histopathology patch images has driven researchers to explore self-supervised learning methods like mask reconstruction for learning rep-resentations from large amounts of unlabeled data. Crucially, previous mask-based efforts in self-supervised learning have often overlooked the spatial interactions among entities, which are essential for constructing accurate representations of pathological entities. To address these challenges, constructing graphs of entities is a promising approach. In addition, the diffusion reconstruction strategy has recently shown superior performance through its random intensity noise addition technique to enhance the robust learned representation. Therefore, we introduce H-MGDM, a novel self-supervised Histopathology image representation learning method through the Dynamic Entity-Masked GraphDiffusion Model. Specifically, we propose to use complementary subgraphs as latent diffusion conditions and self-supervised targets respectively during pre-training. We note that the graph can embed entities’ topological relationships and enhance representation. Dynamic conditions and targets can improve pathological fine reconstruction. Our model has conducted pretraining experiments on three large histopathological datasets. The advanced predictive performance and interpretability of H-MGDM are clearly evaluated on comprehensive downstream tasks such as classification and survival analysis on six datasets
Coupling high resolution meteorological models with neural networks for flash flood forecasting: implementation on a Southern France basin
No full textInternational audienceFlash floods are an important hazard that particularly affects the Mediterranean region. Flood forecasting using simulation tools adapted to this context is therefore a crucial issue. In exposed regions, the difficulty of measuring and forecasting the spatial variability and intensity of rainfall, as well as the difficulty of identifying processes at the necessary time and space scales, has often led to the use of highly conceptual - or even statistical - models that make few assumptions about hydrological processes. Among these, neural networks have proven their relevance for flash flood forecasting. However, without hydrometeorological coupling, flow forecasting is often limited to the response time of the basin, i.e. a few hours in general. The purpose is to find a way of increasing this lead time, which is often too short for crisis management.A flood forecasting model for the Gardon de Mialet basin (Southern France) is being developed as part of the HydIA joint laboratory funded by the ANR (French National Research Agency) and the Synapse company, with the aim of developing a range of hydrometeorological forecasting services based on artificial intelligence approaches. The use of gridded observed data, like in a meteorological model, has enabled the neural network model implemented (Multilayer Perceptron) to reduce its sensitivity to support change.In the absence of rainfall forecasts, performance decreases with the lead time. With perfect forecasts (observed data used as future data), performance remains high for lead times up to 24h. The model has been coupled with two high resolution weather models, AROME and ARPEGE (2.5km and 10km respectively), implemented by Météo-France for short-range numerical weather prediction. The use of forecasts from these meteorological models for the 49 events in the database enables us to identify the error generated by the hydrological model and that generated by the meteorological model, in comparison with perfect forecasts. Analysis of these errors opens operational perspectives for crisis management. It also makes it possible to improve model training based on perfectible forecast data, and to correct rainfall forecasting biases to achieve higher performance
Development of polyfunctionalized chelating resin for efficient recovery of uranium and thorium from waste materials
No full textInternational audienc
Une méthode basée modèles, données et patrons pour l’ingénierie d’un Système Jumeau Numérique
No full textNational audienceUn Jumeau Numérique (JN) est classiquement vu comme la réplique virtuelle d’une entité physique (e.g. un système de production, un réseau logistique ou un produit) dans le but de servir un usage particulier, e.g. contrôler en temps réel l’entité physique, analyser et anticiper ses possibles dérives comportementales, organiser et gérer sa maintenance ou former des opérateurs. Les standards se sont récemment étoffés et mettent en avant le concept de Système Jumeau Numérique (SJN). C’est un système fournissant des fonctionnalités pour le JN, composé d'entités cibles interopérables, d'entités numériques, de connexions de données et de modèles, données et interfaces impliqués dans le processus de connexion de données. Les fonctionnalités d’un SJN sont de même précisées et complétées par une liste de capabilités dont un SJN doit être pourvu selon les usages souhaités. Il s’agit de capabilités en termes de Data Services (DS), d’Integration (IR), d’Intelligence (IN), d’User Expérience (UX), de Management (MN) et de Trustworthiness (TW). Cependant, l’ingénierie d’un SJN comme sa maintenance en conditions opérationnelles (MCO) en restant en phase avec les évolutions de l’entité physique qu’il représente, reste un travail difficile, long, coûteux, et mobilisateur de ressources. Pour faciliter cette ingénierie et cette maintenance, les travaux présentés ici posent plusieurs hypothèses. Tout d’abord, l’ingénierie d’un SJN suit les processus et principes d’une Ingénierie Système Basée Modèles. Cette façon de mener à bien promeut et recommande la création et l’utilisation de modèles, ici du SJN, pour arriver à mieux maitriser la complexité du sujet. De même, cette ingénierie gagnerait à s’inspirer et tirer parti du concept de patron et d’approches de type Ingénierie de Systèmes Basée Patrons (PBSE) pour faciliter la réutilisabilité donc réduire les efforts des parties prenantes impliquées. Un patron est en effet un principe associant la description d’un problème rencontré en cours de conception, de développement ou de MCO, à un espace de solution(s), l’un et l’autre étant valides dans un contexte bien défini. Ces travaux visent donc à proposer une méthode d’ingénierie de SJN qui associe ces deux hypothèses, basée Modèles et Patrons. En synthèse, cette méthode doit promouvoir des principes, activités et outils de modélisation en phase avec l’hypothèse MBSSE, puis alimenter continuellement, assurer la cohérence et mettre à disposition dans tous les projets d’ingénierie d’un SJN d’un catalogue de patrons structurés en fonction : - De la nature même du système physique, des besoins spécifiques des parties prenantes (e.g. concepteurs, futur utilisateurs et mainteneurs) ; - De classes et de types de problèmes déjà rencontrés au cours d’autres projets, exprimés à un niveau générique ou spécifiques e.g. d’ordre conceptuel, méthodologique ou technique et de solutions à ces problèmes jugées satisfaisantes et éprouvées. Le but est bien ici de promouvoir la réutilisation et la capitalisation d’expérience et de culture d’entreprise, de réduire les délais et les efforts de recherche et développement sans pour autant entraver mais en guidant la créativité des parties prenantes, et détecter au plus tôt les erreurs et omissions ; - De l’usage prévu pour le SJN en cours d’ingénierie et des capabilités qui sont alors recommandées. Cette méthode est donc définie par 5 composantes que sont 1) les concepts/attributs/relations définissant ainsi un vocabulaire unifié et suffisant pour l’ingénierie de SJN, 2) les langages de modélisation mais aussi de simulation, de programmation, d’analyse ou autres qui utiliseront alors ces concepts et permettront de construire et analyser, améliorer et optimiser des modèles du SJN, 3) une démarche opératoire expliquant comment la méthode est mise en oeuvre et, pour cela, il s’agit de définir quatre processus maîtres (Acculturation, Déploiement, Amélioration et Application s’inspiran
Detecting fast-ripples on both micro- and macro-electrodes in epilepsy: A wavelet-based CNN detector
No full textInternational audienceFast-ripples (FR) are short (∼10 ms) high-frequency oscillations (HFO) between 200 and 600 Hz that are helpful in epilepsy to identify the epileptogenic zone. Our aim is to propose a new method to detect FR that had to be efficient for intracerebral EEG (iEEG) recorded from both usual clinical macro-contacts (millimeter scale) and microwires (micrometer scale)