Archive ouverte de Centrale Lyon
Not a member yet
32420 research outputs found
Sort by
SCONet: Convolutional Occupancy Networks for Multi-Organ Segmentation
No full textInternational audienceConvolutional neural networks are the de facto standard for 3D multi-organ segmentation but still exhibit significant limitations, especially regarding their computational cost, with high running time and often prohibitive memory footprint for large 3D volumes. To overcome these limitations, we propose to replace the image voxel grid with a more compact point cloud representation. Recently, in the field of 3D object reconstruction, networks learning implicit functions from an input point cloud, such as Convolutional Occupancy Networks (ConvONet), have proven their good surface representation capabilities. We therefore propose SCONet (Segmentation Convolutional Occupancy Network), a lightweight ConvONet-based network adapted to the specific task of multi-organ segmentation. SCONet takes as input a point cloud extracted from the original volume with a standard contour detection algorithm, and enriches it with geometric and photometric features. Thanks to its ability to query per organ occupancy probabilities for any point in space, SCONet can be used to predict a multi-organ segmentation map at arbitrary resolution. We evaluate our method on an abdominal CT image dataset and compare its performances with those of discrete and implicit baselines. Our implementation is available at https://github.com/maylis-j/SCONet.</div
Regular singular Mahler equations and Newton polygons
No full textInternational audienceThough Mahler equations have been introduced nearly one century ago, the study of their solutions is still a fruitful topic for research. In particular, the Galois theory of Mahler equations has been the subject of many recent papers. Nevertheless, long is the way to a complete understanding of relations between solutions of Mahler equations. One step along this way is the study of singularities. Mahler equations with a regular singularity at 0 have rather "nice" solutions: they can be expressed with the help of Puiseux series and solutions of equations with constant coefficients. In a previous paper, the authors described an algorithm to determinewhether an equation is regular singular at 0 or not. Exploiting information from the Frobenius method and Newton polygons, we improve this algorithm by significantly reducing its complexity, by providing some simple criterion for an equation to be regular singular at 0, and by extending its scope to equations with Puiseux coefficients
Does learning to write allographs on a touchscreen tablet stimulate the development of the abstract identity of letters in 5-year-old pre-readers ?
No full textInternational audienceThe aim is to study the effect of a visual-motor exploration of allographs on a touch tablet to develop the abstract identity of letters in 5-year-old pre-readers. The scores in visual allograph recognition are compared between a pre- and post-test according to the exploration of the allographs on the tablet (visuomotor: writing with a stylus; and visual: seeing the writing of a trace). In this task, we manipulated allographic variability (low -e.g. c-, high -e.g. b-) and calligraphy (uppercase script, lowercase script, lowercase cursive). While writing letters on the tablet, static and dynamic parameters are recorded (e.g. total time). Repeated measures ANOVA shows superior progression with visual exploration, particularly for allographs with high variability. These results are discussed in light of environment-subject interactions. No correlation is shown between the gain in visual recognition of allographs and the three motor components identified by principal component analysis, which suggests that motor processes do not contribute to visual processes when the shapes of letters are learned on a tablet
Quantum dots-doped microlenses made by photolithography
No full textInternational audienceMicrolenses doped with quantum dot emitters offer promising potential for photonics applications, but they require complex fabrication. In this work, we present a method to fabricate microlenses incorporating core/shell CdSe/CdxZn1-xS quantum dot red emitters using photolithography. The quantum dots were first synthesized and dispersed in propylene glycol monomethyl ether acetate solvent using suitable ligand exchange. This quantum dot-loaded solvent was then combined with a transparent commercial photoresist and patterned via ultraviolet lithography to create micro-disks with diameters ranging from 20 to 5 µm. The microstructures were then annealed, forming microlenses through the thermal reflow of the photoresist. Micro-photoluminescence and lifetime measurements confirm the homogeneous and efficient emission of the quantum dots within the microlenses. This work demonstrates the feasibility of fabricating microlenses doped with CdSe-based core/shell quantum dots through photolithography and thermal reflow, offering a straightforward approach for photonics applications
Antiferroelectric fluorite-based capacitors for ultra-high energy storage density applications
No full textInternational audienceFor over a decade, ferroelectric and antiferroelectric ultra-thin films of fluorite-structured materials have drawn significant attention for applications requiring high integration density [1]. Among all other dopants, it has been shown that zirconium allows the lowest annealing temperature to crystallize the orthorhombic phase in polycrystalline Hf1-x ZrxO2 (HZO) thin films, responsible for ferroelectricity [2]. Decreasing the annealing temperature allows the growth of CMOS-compatible thin films. At high doping, HZO layers can become antiferroelectric (AFE) [3]. Antiferroelectric HZO holds significant promise for a large range of applications owing to its potential for high energy storage density (ESD) and high efficiency. The performance and properties of Metal/Insulator/Metal (MIM) capacitors will be presented. The AFE insulator is grown by Plasma-Enhanced Atomic Layer Deposition (PEALD). Several processing conditions are investigated (annealing temperature, thin film thickness, measurement conditions) to reach high energy storage properties. Our results show one of the highest, if not the highest, ESD ever reported so far of 125 mJ/cm3 for a ZrO2 thin film of 10.3 nm. Efficiency and endurance of the MIM stack can also be optimized through material engineering and a wake-up effect is also observed for ZrO2 thin films, which is rarely observed for ZrO2 layers. The performances are extremely encouraging for attaining ultra-high energy storage density in multilayer stacks.References: [1] T. S. Böscke, J. Müller, D. Bräuhaus, U. Schröder, U. Böttger, Appl Phys Lett. 99, 102903 (2011).[2] J. Müller et al., Appl Phys Lett. 99, 112901 (2011).[3] Park, M. H., Lee, Y. H., Kim, H. J., Kim, Y. J., Moon, T., Kim, K. D., Müller, J., Kersch, A., Schroeder, U., Mikolajick, T., & Hwang, C. S., Advanced Materials, 27(11), 1811-1831 (2015)
Eco‐Friendly Deposition of Catalyst‐Free Graphene on Diverse Substrates (Adv. Sustainable Syst. 4/2025)
No full textInternational audienceIn article number 2500105, Mustapha Jouiad and co-workers present an environmentally friendly and catalyst-free method for depositing high-quality graphene on various substrates, using ethylene as the sole carbon source. Using low-pressure electrochemical vapor deposition at room temperature, followed by flash annealing, the approach allows direct deposition without a complex transfer process. The resulting graphene films, 3–10 layers, exhibit excellent conductivity, opening promising prospects for sustainable graphene applications
A Multi-Criteria Approach for Gaze Analysis Similarity in Paintings
No full textInternational audienceIn the fields of art history and visual semiotics, analysing gazes in paintings is important to understand the artwork, and to find semantic relationships between several paintings. Thanks to digitization and museum initiatives, the volume of datasets on artworks continues to expand, enabling new avenues for exploration and research. Artificial neural networks, trained on large datasets are able to extract complex features, and visually compare artworks. This comparison could be done by focusing on the objects present in the paintings, and matching paintings with high object co-occurrence. Our research takes this further by studying the way objects are viewed by characters in the scene. This study proposes a new approach that combines methods for gazebased and visual-based similarity, to encode and use gaze information for finding similar paintings, while maintaining a close visual aspect. Experimental results which integrate the opinions of domain experts, show that these methods complement each other. Quantitative and qualitative assessments confirm the results from the combination of gaze and visual analysis. Thus, this method improves existing visual similarity queries and opens up new possibilities to retrieve similar paintings according to user-specific criteria
Comment se rassembler (avec certitude) quand on est confus et désorienté
No full textInternational audienceNous considérons le problème du rassemblement en temps fini et au même endroit, non connu à l’avance, d’un ensemble de robots déterministes semi-synchrones, partant d’une configuration initiale arbitraire qui peut même être bivalente (c’est-à-dire une configuration où les robots sont répartis également sur deux emplacements différents). Ce problème est connu pour être insoluble lorsque les robots sont oublieux, c’est-à-dire lorsqu’ils ne peuvent pas se souvenir de leurs actions passées. Nous présentons un algorithme de rassemblement déterministe où les robots peuvent retenir et communiquer un bit de mémoire. Ce bit peut être arbitrairement défini dans la configuration initiale. Notre solution est donc optimale en termes de mémoire et auto-stabilisante. Sa preuve de correction est formellement certifiée par l’assistant de preuve Coq en utilisant la bibliothèque Pactole
Rendez-vous au Point-Dont-On-Ne-Doit-Pas-Calculer-Les-Coordonnées
No full textInternational audienceLe point Weber, aussi appelé médian géométrique, a toujours eu mauvaise réputation dans les travaux sur les robots mobiles. La raison invoquée est qu’il n’est pas constructible, mais tout le monde sait que la vraie raison est qu’il offre une solution triviale au problème du regroupement, supprimant ainsi la possibilité de publier des protocoles compliqués dans des journaux réputés. Est-ce cependant vraiment le cas ? Maintenant que les protocoles compliqués sont publiés, nous montrons qu’en fait, le point Weber, bien que non constructible par un robot muni d’un compas et d’une règle, n’est pas une solution triviale (une destination inconditionnelle). Dans cet article, nous présentons un protocole non trivial utilisant le point Weber pour regrouper un nombre quelconque de robots ne se trouvant pas initialement dans une configuration impossible à résoudre (bivalente) pour un modèle d’exécution asynchrone (sans hypothèses d’ordonnancement des cycles ni sur l’achèvement des déplacements) et non rigide (les déplacements peuvent être interrompus). Le point Weber et ses propriétés, ainsi que la correction de notre protocole sont prouvés en Coq, grâce à la bibliothèque PACTOLE
About Berge–Füredi’s conjecture on the chromatic index of hypergraphs
No full textInternational audienceWe prove that the chromatic index of a hypergraph H satisfies the Berge-Füredi conjectured bound q(H ≤ ∆ [H 2 ] + 1 under certain hypotheses on the antirank ar(H ) or on the maximum degree ∆(H ). This provides sharp information in connection with the Erdős-Faber-Lovász conjecture which deals with the coloring of a family of cliques that intersect pairwise in at most one vertex.Résumé. Nous montrons que l'indice chromatique d'un hypergraphe satisfait la borne conjecturée indépendamment par Berge et Füredi, q(H ≤ ∆ [H 2 ] + 1 sous certaines conditions portant sur son antirang ar(H ) ou son degré maximum ∆(H ). Ces résultats fournissent des informations sur la conjecture de Erdős, Faber and Lovász qui traite de la coloration d'une famille de cliques se coupant deux à deux en au plus un sommet.</div