Dans cette thèse, nous nous intéressons au problème de la détection d’objets faiblement supervisée. Le but est de reconnaître et de localiser des objets dans les images, n’ayant à notre disposition durant la phase d’apprentissage que des images partiellement annotées au niveau des objets. Pour cela, nous avons proposé deux méthodes basées sur des modèles différents. Pour la première méthode, nous avons proposé une amélioration de l’approche ”Deformable Part-based Models” (DPM) faiblement supervisée, en insistant sur l’importance de la position et de la taille du filtre racine initial spécifique à la classe. Tout d’abord, un ensemble de candidats est calculé, ceux-ci représentant les positions possibles de l’objet pour le filtre racine initial, en se basant sur une mesure générique d’objectness (par region proposals) pour combiner les régions les plus saillantes et potentiellement de bonne qualité. Ensuite, nous avons proposé l’apprentissage du label des classes latentes de chaque candidat comme un problème de classification binaire, en entrainant des classifieurs spécifiques pour chaque catégorie afin de prédire si les candidats sont potentiellement des objets cible ou non. De plus, nous avons amélioré la détection en incorporant l’information contextuelle à partir des scores de classification de l’image. Enfin, nous avons élaboré une procédure de post-traitement permettant d’élargir et de contracter les régions fournies par le DPM afin de les adapter efficacement à la taille de l’objet, augmentant ainsi la précision finale de la détection. Pour la seconde approche, nous avons étudié dans quelle mesure l’information tirée des objets similaires d’un point de vue visuel et sémantique pouvait être utilisée pour transformer un classifieur d’images en détecteur d’objets d’une manière semi-supervisée sur un large ensemble de données, pour lequel seul un sous-ensemble des catégories d’objets est annoté avec des boîtes englobantes nécessaires pour l’apprentissage des détecteurs. Nous avons proposé de transformer des classifieurs d’images basés sur des réseaux convolutionnels profonds (Deep CNN) en détecteurs d’objets en modélisant les différences entre les deux en considérant des catégories disposant à la fois de l’annotation au niveau de l’image globale et l’annotation au niveau des boîtes englobantes. Cette information de différence est ensuite transférée aux catégories sans annotation au niveau des boîtes englobantes, permettant ainsi la conversion de classifieurs d’images en détecteurs d’objets. Nos approches ont été évaluées sur plusieurs jeux de données tels que PASCAL VOC, ImageNet ILSVRC et Microsoft COCO. Ces expérimentations ont démontré que nos approches permettent d’obtenir des résultats comparables à ceux de l’état de l’art et qu’une amélioration significative a pu être obtenue par rapport à des méthodes récentes de détection d’objets faiblement supervisées. / In this dissertation we address the problem of weakly supervised object detection, wherein the goal is to recognize and localize objects in weakly-labeled images where object-level annotations are incomplete during training. To this end, we propose two methods which learn two different models for the objects of interest. In our first method, we propose a model enhancing the weakly supervised Deformable Part-based Models (DPMs) by emphasizing the importance of location and size of the initial class-specific root filter. We first compute a candidate pool that represents the potential locations of the object as this root filter estimate, by exploring the generic objectness measurement (region proposals) to combine the most salient regions and “good” region proposals. We then propose learning of the latent class label of each candidate window as a binary classification problem, by training category-specific classifiers used to coarsely classify a candidate window into either a target object or a non-target class. Furthermore, we improve detection by incorporating the contextual information from image classification scores. Finally, we design a flexible enlarging-and-shrinking post-processing procedure to modify the DPMs outputs, which can effectively match the approximate object aspect ratios and further improve final accuracy. Second, we investigate how knowledge about object similarities from both visual and semantic domains can be transferred to adapt an image classifier to an object detector in a semi-supervised setting on a large-scale database, where a subset of object categories are annotated with bounding boxes. We propose to transform deep Convolutional Neural Networks (CNN)-based image-level classifiers into object detectors by modeling the differences between the two on categories with both image-level and bounding box annotations, and transferring this information to convert classifiers to detectors for categories without bounding box annotations. We have evaluated both our approaches extensively on several challenging detection benchmarks, e.g. , PASCAL VOC, ImageNet ILSVRC and Microsoft COCO. Both our approaches compare favorably to the state-of-the-art and show significant improvement over several other recent weakly supervised detection methods.

The propulsion mechanics driving the movement of living cells constitutes one of the most incredible engineering works of nature. Active cell motility via the controlled movement of a flagellum beating is among the phylogentically oldest forms of motility, and has been retained in higher level organisms for spermatozoa transport. Despite this ubiquity and importance, the details of how each structural component within the flagellum is orchestrated to generate bending waves, or even the elastic material response from the sperm flagellum, is far from fully understood. By using microbiomechanical modelling and simulation, we develop bio-inspired mathematical models to allow the exploration of sperm motility and the material response of the sperm flagellum. We successfully construct a simple biomathematical model for the human sperm movement by taking into account the sperm cell and its interaction with surrounding fluid, through resistive-force theory, in addition to the geometrically non-linear response of the flagellum elastic structure. When the surrounding fluid is viscous enough, the model predicts that the sperm flagellum may buckle, leading to profound changes in both the waveforms and the swimming cell trajectories. Furthermore, we show that the tapering of the ultrastructural components found in mammalian spermatozoa is essential for sperm migration in high viscosity medium. By reinforcing the flagellum in regions where high tension is expected this flagellar accessory complex is able to prevent tension-driven elastic instabilities that compromise the spermatozoa progressive motility. We equally construct a mathematical model to describe the structural effect of passive link proteins found in flagellar axonemes, providing, for the first time, an explicit mathematical demonstration of the counterbend phenomenon as a generic property of the axoneme, or any cross-linked filament bundle. Furthermore, we analyse the differences between the elastic cross-link shear and pure material shear resistance. We show that pure material shearing effects from Cosserat rod theory or, equivalently, Timoshenko beam theory or are fundamentally different from elastic cross-link induced shear found in filament bundles, such as the axoneme. Finally, we demonstrate that mechanics and modelling can be utilised to evaluate bulk material properties, such as bending stiffness, shear modulus and interfilament sliding resistance from flagellar axonemes its constituent elements, such as microtubules.

Deformable image registration (DIR) is an essential tool in medical image processing. It provides a means to combine image datasets, allowing for intra-subject, inter-subject, multi-modality, and multi-instance analysis, as well as motion detection and compensation. One of the most popular DIR algorithms models the displacement vector field (DVF) as B-splines, a sum of piecewise polynomials with coefficients that enable local shape control. B-splines have many advantageous properties in the context of DIR, but they often struggle to adequately model steep local gradients and discontinuities. This dissertation addresses that limitation by proposing the replacement of conventional B-splines with a generalized formulation known as a Non-Uniform Rational B-Splines (NURBS). Beginning with the 1D fitting, heuristic rules are developed to determine the values of the additional free parameters introduced by NURBS. These rules are subsequently modified and extended to the 2D and 3D fitting of anonymized and publicly available patient DVFs. Based on the lessons learned from these increasingly complex test cases, a 2D DIR scheme is developed and tested on slices from a thoracic computed tomography (CT) scan. Finally, an automatic, non-uniform scheme is presented, and its registration performance is compared to the conventional uniform methods.

Facial modelling and animation has been an active research subject in computer graphics since the 1970s. Due to extremely complex biomechanical structures of human faces and people’s visual familiarity with human faces, modelling and animating realistic human faces is still one of greatest challenges in computer graphics. Since we are so familiar with human faces and very sensitive to unnatural subtle changes in human faces, it usually requires a tremendous amount of artistry and manual work to create a convincing facial model and animation. There is a clear need of developing automatic techniques for facial modelling in order to reduce manual labouring. In order to obtain a realistic facial model of an individual, it is now common to make use of 3D scanners to capture range scans from the individual and then fit a template to the range scans. However, most existing template-fitting methods require manually selected landmarks to warp the template to the range scans. It would be tedious to select landmarks by hand over a large set of range scans. Another way to reduce repeated work is synthesis by reusing existing data. One example is expression cloning, which copies facial expression from one face to another instead of creating them from scratch. This aim of this study is to develop a fully automatic framework for template-based facial modelling, facial expression transferring and facial expression tracking from range scans. In this thesis, the author developed an extension of the iterative closest points (ICP) algorithm, which is able to match a template with range scans in different scales, and a deformable model, which can be used to recover the shapes of range scans and to establish correspondences between facial models. With the registration method and the deformable model, the author proposed a fully automatic approach to reconstructing facial models and textures from range scans without re-quiring any manual interventions. In order to reuse existing data for facial modelling, the author formulated and solved the problem of facial expression transferring in the framework of discrete differential geometry. The author also applied his methods to face tracking for 4D range scans. The results demonstrated the robustness of the registration method and the capabilities of the deformable model. A number of possible directions for future work were pointed out.

La formation aux gestes en obstétrique réalisée directement auprès du patient pose des problèmes éthiques et médico-légaux. C'est pourquoi une grande partie de cette formation repose sur l'observation in-situ. Cette approche ne permet pas de prendre en charge l'ensemble des dimensions nécessaires à la formation telles que l'extraction instrumentale, l'acquisition d'un niveau de dextérité suffisant ou encore la capacité de prise de décision face à une situation à risque.Les systèmes d'entrainement par simulation consituent une réponse à ce problème de formation. Toutefois, les simulateurs haptiques actuels ne permettent qu'une évaluation qualitative du geste obstétrique et ne fournissent pas d'informations quantitatives sur les efforts subis par les différents organes pelviens. Ces informations quantitatives ne sont accessibles que par des simulation biomécaniques de la descente fœtale. Par contre, pour pouvoir être utilisées conjointement avec un simulateur haptique dans un but d'apprentissage, ces simulations doivent être interactives.Dans ce manuscrit, nous proposons tout d'abord un état de l'art des différents modèles biomécaniques existants permettant de modéliser les organes pelviens et la descente fœtale. Après avoir identifié la tête foetale comme la structure la plus susceptible de subir des dommages importants lors de l'accouchement, nous présentons un modèle de la tête foetale basé sur des éléments finis de coque CST-DKT que nous étendons avec un formulation co-rotationelle et une contrainte de volume permettant de prendre en compte la matière intra-crânienne. Une implémentation GPU de ce modèle est proposée pour permettre des simulations interactives. Ce modèle est validé sur une simulation de la pression intra-utérine subie par la tête fœtale lors de la deuxième phase de l'accouchement. Enfin, nous proposons une ébauche de modèle complet pour la simulation interactive de la descente fœtale. / The training to obstetrical gestures performed directly on the patient raise ethical and medico-legal problems. That is why most of the formation is based on in-situ observations. This approach can not handle all the necessary dimensions to the formation such as instrumental extraction, a good level of dexterity or the capacity to take decision in high risk situations.Simulation based training systems can address this formation problems. Neverteless, today's haptic simulators don't provide quantitative informations about the efforts undergone by pelvic organs. This informations are available only through biomechanical simulations of the foetal descent. To be used in conjunction with an haptic simulator, such simulations must be interactive.In this manuscript, first we propose a state of the art of the existing biomechanical models allowing to model the pelvic organs. After having identified the foetal head as the structure that can potentially undergo important damages during childbirth, we present a model of the foetal head based on shell finite elements CST-DKT that we extend to a co-rotationnal formulation and a volume constraint allowing to take in count the intra-cranial matter. A GPU implementation of this model is also porposed to allow interactive simulations. This model is validated on a simulation of intra-uterine pressure undergone by the foetal head during the second phase of childbirth. Finally, we propose a draft of a complete model intended to simulate interactively the feotal descent.

"A number of outstanding problems in robotic motion and manipulation involve tasks where degrees of freedom (DoF), be they part of the robot, an object being manipulated, or the surrounding environment, cannot be accurately controlled by the actuators of the robot alone. Rather, they are also controlled by physical properties or interactions - contact, robot dynamics, actuator behavior - that are influenced by the actuators of the robot. In particular, we focus on two important areas of poorly controlled robotic manipulation: motion planning for deformable objects and in deformable environments; and manipulation with uncertainty. Many everyday tasks we wish robots to perform, such as cooking and cleaning, require the robot to manipulate deformable objects. The limitations of real robotic actuators and sensors result in uncertainty that we must address to reliably perform fine manipulation. Notably, both areas share a common principle: contact, which is usually prohibited in motion planners, is not only sometimes unavoidable, but often necessary to accurately complete the task at hand. We make four contributions that enable robot manipulation in these poorly controlled tasks: First, an efficient discretized representation of elastic deformable objects and cost function that assess a ``cost of deformation' for a specific configuration of a deformable object that enables deformable object manipulation tasks to be performed without physical simulation. Second, a method using active learning and inverse-optimal control to build these discretized representations from expert demonstrations. Third, a motion planner and policy-based execution approach to manipulation with uncertainty which incorporates contact with the environment and compliance of the robot to generate motion policies which are then adapted during execution to reflect actual robot behavior. Fourth, work towards the development of an efficient path quality metric for paths executed with actuation uncertainty that can be used inside a motion planner or trajectory optimizer."

Submitted by Silvana Teresinha Dornelles Studzinski (sstudzinski) on 2016-05-04T15:43:49Z No. of bitstreams: 1 Francisco de Assis da Silva Cavalcante Filho_.pdf: 3681280 bytes, checksum: 92b2f9dcc668fbbbd3c9feae15c9d569 (MD5) / Made available in DSpace on 2016-05-04T15:43:49Z (GMT). No. of bitstreams: 1 Francisco de Assis da Silva Cavalcante Filho_.pdf: 3681280 bytes, checksum: 92b2f9dcc668fbbbd3c9feae15c9d569 (MD5) Previous issue date: 2016-03-01 / IFRR - Instituto Federal de Educação Ciências e Tecnologia de Roraima / A Articulação Temporomandibular é uma das articulações mais complexas do corpo humano. Ela é composta de duas articulações, uma em cada lado da mandíbula, que trabalham juntas para realizar os movimentos de abertura e fechamento da boca, assim como os movimentos de mastigação. É uma das articulações que está constantemente sob pressão e a sobrecarga nessa articulação pode resultar em inúmeras condições médicas. Cerca de 30% da população apresenta algum sintoma de DTM, sendo o desalinhamento dos discos temporomandibulares o mais comum entre eles, atingindo 70% dos pacientes. Os trabalhos desenvolvidos até o momento tiveram como foco o estudo de características específicas da ATM, como a trajetória da mandíbula em determinados movimentos, os músculos que são ativados nestes movimentos ou a tensão sofrida pelos discos articulares. Estes modelos apresentam uma representação limitada da ATM e estruturas adjacentes ou utilizam técnicas que demandam muito poder computacional, restringindo sua utilização a pesquisas direcionadas. Este trabalho propõe um modelo para visualização dos movimentos da mandíbula de forma interativa com base nas restrições do disco articular e estruturas adjacentes. O modelo emprega técnicas de simulação física para obter maior realismo na visualização dos movimentos, permitindo sua aplicação em várias áreas da medicina. / The temporomandibular joint is one of the most complex joints in the human body. It is composed of two joints, one at each side of the jaw, that work together to perform the opening and closing movements of the mouth, as well as chewing movements. It is a joint that is constantly under pressure and the loading in this joint can result in numerous medical conditions. About 30% of the population have some symptom of TMD, and the temporomandibular discs displacement is the most common between them, reaching 70% of patients. The work carried out so far have focused on the study of specific features of ATM, as the trajectory of the jaw in certain movements, the muscles that are activated in these movements or stress suffered by the articular disc. These models have a limited representation of ATM and adjacent structures or use techniques that requires too much computational power, restricting its use to directed research. This paper presents a model for visualization of the jaw movements interactively based on disk restrictions and adjacent structures. The model uses physical simulation techniques to achieve more realism on movements visualization, allowing its application in various areas of medicine.

A pele possui um grande número de mecanismos de defesa antioxidante. No entanto, a exposição prolongada à radiação ultravioleta (RUV), é capaz de aumentar a concentração de espécies reativas de oxigênio (EROs) provocando um desequilíbrio antioxidante/oxidante. A geração de EROs pode levar ao fotoenvelhecimento e doenças, como o câncer. A administração tópica de antioxidantes é uma maneira eficiente de enriquecer o sistema protetor cutâneo endógeno e, assim, uma estratégia para reduzir os danos oxidativos causados pela RUV à pele. Entretanto, para fornecer a proteção adequada à pele, estas substâncias devem ultrapassar a barreira imposta pelo estrato córneo. O extrato de Lychnophora salicifolia Mart. possui compostos com atividades antioxidante, anti-inflamatória e analgésica, como o ácido clorogênico e a vicenina-2, sendo, portanto, um possível candidato à agente protetor dos danos induzidos pela radiação na pele. Assim, o objetivo do presente trabalho foi avaliar o potencial fotoquimiopreventivo deste extrato e do ácido clorogênico, bem como de formulações contendo o extrato e a substância isolada encapsulados em lipossomas. Para atingir tal objetivo, foram realizados experimentos in vitro e in vivo de atividade antioxidante e eficácia fotoquimiopreventiva para comprovação do potencial antioxidante e anti-inflamatório, estudos de liberação e estudos de penetração cutânea das formulações utilizando pele humana e pele de orelha de porco. Os resultados obtidos demonstraram que o extrato possui atividade antioxidante in vitro e baixa citotoxicidade. A incorporação do extrato e da substância isolada em lipossomas foi bem sucedida, uma vez que estes demonstraram uma boa estabilidade físico-química e aumentaram a penetração dos compostos na pele de camundongos sem pelos. As formulações contendo lipossomas permitiram que os compostos do extrato e o ácido clorogênico isolado atravessassem o estrato córneo chegando até epiderme e derme na pele humana ex vivo. O acúmulo de partículas nos folículos pilosos parece não exercer influencia neste processo. Os extrato foi capaz de sequestrar EROs intracelulares e evitar a depleção do antioxidante endógeno GSH tanto em cultura de células como nos experimentos in vivo. Também foi observada a inibição in vivo da metaloproteinase de matriz-9 (MMP-9), enzima ativada pela RUV e que tem a capacidade de degradar os componentes da matriz extracelular da pele. Além disso, o extrato apresentou atividade anti-inflamatória superior à do ácido clorogênico, verificada pela diminuição dos níveis de citocinas pró-inflamatórias (IL-1?, IL-6 e TNF-?) e da atividade da enzima mieloperoxidase (MPO) in vivo. Portanto, o extrato parece ser um agente fotoquimiopreventivo promissor para uso tópico e, as formulações desenvolvidas, um sistema viável para liberação destas substâncias na pele. / The skin has several defense mechanisms to avoid oxidant damages. However, a prolonged exposition under ultraviolet radiation (UVR) may increase reactive oxygen species (ROS) concentration causing antioxidant/oxidant imbalance. ROS generation can induce photoaging and skin cancer. Topical administration of antioxidants is an efficient way to fortify the endogenous protection system and to reduce the oxidative damage caused by UVR. Nevertheless, to be effective, these compounds need to pass through the skin barrier imposed by stratum corneum. Lychnophora salicifolia Mart. extract has antioxidant, antiinflammatory and analgesic compounds, such as chlorogenic acid and vicenin-2, and it may be a skin protector agent against UVR damage. Therefore, the aim of this study was to evaluate the photochemopreventive potential of this extract and chlorogenic acid, as well as their formulation containing liposomes. In vitro and in vivo tests were performed to prove antioxidant and antiinflammatory activity. Formulation release and skin penetration studies were also carried out in human and porcine ear skin. The results showed that the extract has antioxidant activity in vitro and low cytotoxicity. The extract and the isolated compound were successfully entrapped in liposomes, demonstrating a good physic and chemical stability and increasing the skin penetration in hairless mice. The formulations containing liposomes allowed that the extract compounds and chlorogenic acid passed through the stratum corneum and reached epidermis and dermis in human skin ex vivo. The hair follicle uptake of the particles do not influenced the penetration process. The extract was able to scavenge intracellular ROS and avoid the depletion of the endogenous GSH both in cell culture and in vivo, even as it inhibited matrix metalloproteinase 9 (MMP-9), enzyme activated by UVR that degrades skin matrix extracellular components. The extract also presented larger antiinflammatory properties than chlorogenic acid, decreasing the cytokines (IL-1?, IL-6 and TNF-?) levels and the mieloperoxidase activity in vivo. Therefore, the results suggest the potential applicability of the extract as a photochemopreventive agent to topical use and the developed formulations seem to be a viable system to release these compounds to the skin.

Afin d’optimiser la surveillance des travailleuses du nucléaire par anthroporadiamétrie, il est nécessaire de corriger les coefficients d’étalonnage obtenus à l’aide du fantôme physique masculin Livermore. Pour ce faire, des étalonnages numériques basés sur l’utilisation des calculs Monte Carlo associés à des fantômes numériques ont été utilisés. De tels étalonnages nécessitent d’une part le développement de fantômes représentatifs des tailles et des morphologies les plus communes et d’autre part des simulations Monte Carlo rapides et fiables. Une bibliothèque de fantômes thoraciques féminins a ainsi été développée en ajustant la masse des organes internes et de la poitrine suivant la taille et les recommandations de la chirurgie plastique. Par la suite, la bibliothèque a été utilisée pour étalonner le système de comptage du Secteur d’Analyses Médicales d’AREVA NC La Hague. De plus, une équation décrivant la variation de l’efficacité de comptage en fonction de l’énergie et de la morphologie a été développée. Enfin, des recommandations ont été données pour corriger les coefficients d’étalonnage du personnel féminin en fonction de la taille et de la poitrine. Enfin, pour accélérer les simulations, des méthodes de réduction de variance ainsi que des opérations de simplification de la géométrie ont été considérées.Par ailleurs, pour l’étude des cas de contamination complexes, il est proposé de remonter à la cartographie d’activité en associant aux mesures anthroporadiamétriques le calcul Monte Carlo. La méthode développée consiste à réaliser plusieurs mesures spectrométriques avec différents positionnements des détecteurs. Ensuite, il s’agit de séparer la contribution de chaque organe contaminé au comptage grâce au calcul Monte Carlo. L’ensemble des mesures réalisées au LEDI, au CIEMAT et au KIT ont démontré l’intérêt de cette méthode et l’apport des simulations Monte Carlo pour une analyse plus précise des mesures in vivo, permettant ainsi de déterminer la répartition de l’activité à la suite d’une contamination interne. / To optimize the monitoring of female workers using in vivo spectrometry measurements, it is necessary to correct the typical calibration coefficients obtained with the Livermore male physical phantom. To do so, numerical calibrations based on the use of Monte Carlo simulations combined with anthropomorphic 3D phantoms were used. Such computational calibrations require on the one hand the development of representative female phantoms of different size and morphologies and on the other hand rapid and reliable Monte Carlo calculations. A library of female torso models was hence developed by fitting the weight of internal organs and breasts according to the body height and to relevant plastic surgery recommendations. This library was next used to realize a numerical calibration of the AREVA NC La Hague in vivo counting installation. Moreover, the morphology-induced counting efficiency variations with energy were put into equation and recommendations were given to correct the typical calibration coefficients for any monitored female worker as a function of body height and breast size. Meanwhile, variance reduction techniques and geometry simplification operations were considered to accelerate simulations.Furthermore, to determine the activity mapping in the case of complex contaminations, a method that combines Monte Carlo simulations with in vivo measurements was developed. This method consists of realizing several spectrometry measurements with different detector positioning. Next, the contribution of each contaminated organ to the count is assessed from Monte Carlo calculations. The in vivo measurements realized at LEDI, CIEMAT and KIT have demonstrated the effectiveness of the method and highlighted the valuable contribution of Monte Carlo simulations for a more detailed analysis of spectrometry measurements. Thus, a more precise estimate of the activity distribution is given in the case of an internal contamination.

Esta memoria para optar al grado de doctor por la Universidad de Alicante, en el área de Matemática Aplicada, se divide en ocho capítulos. En los cuatro primeros se lleva a cabo una revisión del problema de la rotación de la Tierra rígida y elástica, en formalismo hamiltoniano, y se introducen algunas alternativas originales a la construcción de ciertos elementos de la teoría. En los otros cuatro se presentan nuevas contribuciones al estudio analítico de los movimientos de precesión y nutación de la Tierra deformable. Estas se pueden agrupar en dos categorías: la obtención de fórmulas que describen el movimiento de rotación del eje de figura, que completan las existentes hasta la fecha, y la extensión del formalismo para hacerlo compatible con modelos reológicos de la Tierra más generales.