163 research outputs found
EDM-Research/DIMO_ObjectDetection: v1.0
No full textObject detection for the DIMO dataset. Uses the Mask-RCNN model. This is the official implementation of Analysis of Training Object Detection Models with Synthetic Data, published in BMVC: British Machine Vision Conference, 2022.
Source code for the following scientific publication:
Vanherle, B., Moonen, S., Van Reeth, F., and Michiels, N. (2022). Analysis of Training Object Detection Models with Synthetic Data. 33rd British Machine Vision Conference 2022, BMVC 2022, London, UK, November 21-24, 2022. Retrieved from https://bmvc2022.mpi-inf.mpg.de/0833.pdfPILS SBO: Product Inspection with Little Supervision. Flanders Make (Belgium). awardNumber:null. 02ndjfz59BOF Special Research Fund. Hasselt University. awardNumber:null. 10.13039/50110000955
Positron scattering by equivalent one-electron models of helium
No full textLow-energy positron-helium scattering is investigated using three different one-electron models of helium in order to find out if such a model is capable of yielding accurate results for elastic scattering, positronium formation and positron-electron annihilation. Comparisons are made with the accurate results obtained from ab initio variational calculations of Van Reeth and Humberston (Van Reeth P and Humberston J W 1999 J. Phys. B: At. Mel. Opt. Phys. 32 3651). The most elaborate helium model used here gives accurate values for the elastic scattering phase shifts throughout most of the energy range up to the positronium formation threshold. However, near-threshold resonances associated with this model have substantial effects on all the partial wave contributions to the positronium formation cross section that have been investigated. The other two helium models yield rather less accurate elastic scattering phase shifts but the positronium formation cross sections are in good qualitative, and even reasonably good quantitative, agreement with the accurate results. None of the models yields very accurate results for the annihilation cross section
VATr++: Choose Your Words Wisely for Handwritten Text Generation
No full textStyled Handwritten Text Generation (HTG) has received significant attention in recent years,propelled by the success of learning-based solutions employing GANs,Transformers,and,preliminarily,Diffusion Models. Despite this surge in interest,there remains a critical yet understudied aspect - the impact of the input,both visual and textual,on the HTG model training and its subsequent influence on performance. This work extends the VATr [1] Styled-HTG approach by addressing the pre-processing and training issues that it faces,which are common to many HTG models. In particular,we propose generally applicable strategies for input preparation and training regularization that allow the model to achieve better performance and generalization capabilities. Moreover,in this work,we go beyond performance optimization and address a significant hurdle in HTG research - the lack of a standardized evaluation protocol. In particular,we propose a standardization of the evaluation protocol for HTG and conduct a comprehensive benchmarking of existing approaches. By doing so,we aim to establish a foundation for fair and meaningful comparisons between HTG strategies,fostering progress in the field
Urban architecture, hybrid buildings: Studio Beveren
No full textMet als bijlage: A0 posterArchitectur
Modeling Performances and Competitive Balance in Professional Road Cycling
No full textIn the economics of professional team sports leagues, the concept of competitive balance is well documented. It postulates the necessity of equilibrium between the teams in a league in order to guarantee uncertainty of outcome and thus generate public demand. By contrast, performances and competitive balance are not easy to define in road cycling. This is because cycling can be seen as a team sport but the global team performance usually is of minor importance or even not taken into account at all. A large proportion of cyclists are in support of another rider, meaning that they do not care about their personal result but instead try to help their team leader(s). Moreover, a team leader generally has one specific objective amongst a range of possible ones. This chapter deals with the complex issue of modeling performances and competitive balance in professional road cycling. After a brief review of the literature on modeling performances and competitive balance in cycling, an innovative measure is introduced: competitive intensity in cycling. We illustrate this measure with two stages of the 2013 and 2014 Tour de France, and we discuss its implications
Surveillance and control of influenza in pigs
No full textInfluenza viruses are members of the family Orthomyxoviridae. They are enveloped, single stranded ribonucleic acid (RNA) viruses with a segmented genome and are grouped into 3 types, designated A, B and C (Wright and Webster 2001). Influenza viruses of the C type are found exclusively in humans and are not considered a public health concern. Influenza B viruses cause sporadic outbreaks of mild respiratory disease in humans. They have also been isolated from pigs (Takátsy et al. 1967) but they are of no veterinary interest. Of greater importance are influenza A viruses. They have a spherical or filamentous morphology and their size ranges from 80 to 120 nm (Wright and Webster 2001). Their genome consists of 8 RNA segments, which encode 10 proteins (Table 1). These include two transmembrane “spike-like” glycoproteins: the haemagglutinin (HA) and the neuramindase (NA), a third transmembrane protein referred to as matrix protein M2, the underlying matrix protein M1, which forms a layer below the lipid envelope and gives structure to the virus, two non structural proteins, the NS1 and NS2 and the ribonucleoprotein (RNP) complexes, which consist of four additional proteins, the nucleoprotein (NP) and the three polymerases: PA, PB1 and PB2 (Figure 1). The HA and NA are of particular interest because (a) they facilitate the entry and release of the virus into and out of the host cell and (b) they are the primary targets recognized by the immune system after infection or vaccination. So far, 16 different types of haemagglutinin (H1 to H16) and 9 of neuraminidase (N1 to N9) have been recognized (Webster and Bean 1998, Fouchier et al. 2005). Influenza A viruses are named using the following convention: A/species of origin/location of isolation/isolate number/year of isolation (in the case of human isolates the species is not mentioned), i.e. A/Swine/Belgium/1/98 or A/New York/55/04 (WHO 1980).
Furthermore, influenza A viruses are subtyped based on the nature of the HA and NA, their combination defines their subtype accordingly: H1N1, H3N2, H5N1. Because of their segmented single stranded RNA genome, influenza viruses have a high mutation rate (genetic drift) and the possibility to undergo reassortment. Reassortment may occur when more than one virus co-infect the same cell, exchange genes and provide a novel influenza virus, which combines gene segments from the original viruses (genetic shift) (Scholtissek 1998, Wright and Webster 2001)
Klimaat
No full textHoofdlijnen:
De ontwikkelde Vlaamse klimaatscenario’s wijzen eenduidig op een stijging van de gemiddelde omgevingstemperatuur tegen 2100 met 1,5°C tot 4,4°C in de winter en met 2,4°C tot 7,2°C in de zomer en op meer neerslag tijdens de winter.
De meeste klimaatscenario’s tonen een daling van de gemiddelde zomerneerslag voor Vlaanderen.
Achttien soorten broedvogels en zestien soorten dagvlinders, waaronder een aantal algemene soorten, lopen een verhoogd risico om tegen 2100 uit Vlaanderen te verdwijnen door de temperatuurstijging.
Mogelijk kunnen nieuwe broedvogel- en dagvlindersoorten zich in Vlaanderen vestigen dankzij de temperatuurstijging.status: Publishe
Design d'impulsions RF par contrôle optimal pour l'optimisation du contraste en IRM : applications in vivo
No full textNational audienceBut : Le design d’impulsions radio-fréquence (RF) par contrôle optimal pour l’optimisation du contraste a récemment été proposé [1]. Ces impulsions garantissent un certain degré d’optimalité sur le contraste obtenu, qui s’approche de la borne théorique. Jusqu’alors, leur utilisation a été validée sur IRM lors d’expériences in vitro uniquement [2]. Cette étude étend ces résultats in vivo sur cerveaux de rat et souris.Méthode : Le problème de contrôle optimal, via la résolution du Principe du Maximum de Pontryagin [3] (PMP), consiste à calculer le contrôle qui optimise une fonction de coût, connaissant la dynamique du système à optimiser. Appliqué à l’IRM, il s’agit de calculer les composantes réelles et imaginaires du champ RF qui amène l’aimantation, dont l’évolution est régie par les équations de Bloch, dans un état souhaité. Pour l’optimisation du contraste, l’état d’aimantation souhaité depend du contraste désiré. Soient Ma(t) et Mb(t) l’aimantation des spins a et b. L’optimisation du contraste (maximisation du signal du spin b et minimisation de a), revient à minimiser : C(w) = ||Ma(tf)||2 - ||Mb(tf)||2 avec w = (wx, wy) le champ RF et tf la durée du champ RF au bout de laquelle le contraste souhaité est atteint. Cette durée est généralement assez longue pour permettre la combinaison des phénomènes d’excitation et de relaxation (centaine de ms). La résolution numérique de ces équations est effectuée grâce à l’algorithme GRAPE [4], qui réduit itérativement la fonction de coût par descente de gradient tout en respectant les contraintes imposées par le PMP. Enfin, les inhomogénéités de champ sont prises en compte dans la dynamique du système, afin que le champ RF soit robuste aux déviations par rapport à la fréquence de Larmor.Résultats : Les acquisitions in vivo ont été réalisées sur cerveaux de rat et souris, sur un IRM Bruker petit animal 4.7T, avec des antennes en quadrature. Dans les deux expériences, le champ RF optimal est calculé de sorte à préparer le contraste sur l’axe longitudinal MZ (impulsion non sélective). L’aimantation est ensuite basculée dans le plan transverse en utilisant un schéma d’excitation classique, ici une séquence RARE. Le TE est fixé le plus court possible (8 ms) afin d’assurer la préservation du contraste au moment de l’acquisition. Le TR est fixé suffisamment long (5 s) pour assurer la repouse complète de l’aimantation, contrainte de l’implémentation actuelle de l’algorithme GRAPE. Le calcul des champs RF nécessite la connaissance des temps de relaxation des tissus à contraster. Ces derniers sont estimés à partir de régressions exponentielles du pic d’eau dans des voxels de spectroscopie acquis à plusieurs TE et TR. L’expérience sur souris consiste à saturer le signal du cerveau ([T1c T2c] = [920, 66] ms) et maximiser le signal des muscles pariétaux ([T mT2m] = [1011, 30] ms) situés de part et d’autre du cerveau. La Figure 1 montre l’amplitude du champ RF optimal calculé, ainsi que l’évolution de l’aimantation du cerveau et des muscles pendant l’application de ce champ. L’image acquise est montrée en Figure 2. L’expérience sur rat consiste à maximiser le signal de l’hippocampe ([T1h T2h] = [921, 68] ms) et minimiser celui du thalamus ([T1t T2t] = [832, 63] ms). La Figure 3 montre une comparaison entre l’image acquise au TE maximisant le contraste T2 (65.4 ms) et l’image obtenue grâce au champ optimal.Discussion : Notons tout d’abord que les images obtenues avec les champs RF calculés ne souffrent pas d’artefacts dus aux inhomogénéités de champ malgré la proximité des canaux auditifs. La Figure 2 illustre la flexibilité sur le contraste qu’offrent les champs calculés par contrôle optimal. Il est en effet difficile de reproduire ce contraste avec les pondérations classiques T1 ou T2 car (T2c > T2m) et (T1c ≈ T m). La Figure 3 montre que l’on peut obtenir un contraste meilleur que le simple contraste T2, et que la visualisation du lemnisque médian est nettement améliorée.Remerciements : ANR-DFG (14-CE35-0013-01), Labex PRIMES (ANR-11-LABX-0063/ANR-11-IDEX-0007).1. Lapert M, et al. Exploring the physical limits of saturation contrast in magnetic resonance imaging” Scientific Reports, Nature Publishing Group, 2012, 22. Van Reeth E, et al. Optimizing MRI Contrast with B1 pulses using optimal control theory, IEEE 12th International Symposium on Biomedical Imaging (ISBI), 20163. Pontryagin L S, Mathematical theory of optimal processes, CRC Press, 1987.4. Khaneja N, et al. “Optimal control of coupled spin dynamics: design of nmr pulse sequences by gradient ascent algorithms,” Journal of Magnetic Resonance, 172(2), 2005
- …
