This thesis consists of a summary of two papers dealing with questions related to Legendrian submanifolds of contact manifolds together with exact Lagrangian cobordisms between Legendrian submanifolds. The focus is on studying Legendrian submanifolds from the perspective of their handle decompositions. The techniques used are mainly from Symplectic Field Theory. In Paper I, a series of examples of Legendrian surfaces in standard contact 5-space are studied. For every g > 0, we produce g+1 Legendrian surfaces of genus g, all with g+1 transverse Reeb chords, which lie in distinct Legendrian isotopy classes. For each g, exactly one of the constructed surfaces has a Legendrian contact homology algebra admitting an augmentation. Moreover, it is shown that the same surface is the only one admitting a generating family. Legendrian contact homology with Novikov coefficients is used to classify the different Legendrian surfaces. In particular, we study their augmentation varieties. In Paper II, the effect of a Legendrian ambient surgery on a Legendrian submanifold is studied. Given a Legendrian submanifold together which certain extra data, a Legendrian ambient surgery produces a Legendrian embedding of the manifold obtained by surgery on the original submanifold. The construction also provides an exact Lagrangian handle-attachment cobordism between the two submanifolds. The Legendrian contact homology of the submanifold produced by the Legendrian ambient surgery is then computed in terms of pseudo-holomorphic disks determined by data on the original submanifold. Also, the cobordism map induced by the exact Lagrangian handle attachment is computed. As a consequence, it is shown that a sub-critical standard Lagrangian handle attachment cobordism induces a one-to-one correspondence between the augmentations of the Legendrian contact homology algebras of its two ends.

In the most recent years, many deep learning models have been used to identify and classify diseases of plant leaves by inputting plant leaf images as input to the model. However, there is still a gap in research on how to improve the accuracy of the deep learning models of plant leaf diseases. This thesis is about investigating various techniques for improving the MobileNetV2 model's accuracy for plant disease detection in leaves and classification. These techniques involved adjusting the learning rate, adding additional layers, and various data-augmented operations. The results of this thesis have shown that these techniques can significantly improve the accuracy of the model, and the best results can be achieved by using random rotation and crop data augmentation. After adding random rotation and crop data augmentation to the model, it achieved an accuracy of 94%, a precision of 91%, a recall of 96%, and an F1-score of 95%. This shows that the proposed techniques can be used to improve the accuracy of plant leaf disease detection and classification models, which can help farmers identify and treat plant diseases.

This thesis examines the evidence for the buildings and precincts of the five friaries of late medieval London: Black Friars, Grey Friars, White Friars, Austin Friars and Crossed (or Crutched) Friars. Virtually nothing survives, at least above ground, of these once-famous institutions and so documentary and archaeological evidence form the core of the research. Using a technique of historic map regression – working backwards from the modern Ordnance Survey map and carrying out a succession of ‘digital tracings' of historic maps – the early modern street plan of each friary was drawn. Then, evidence from dozens of archaeological excavations (small and large, antiquarian and modern) could be pasted onto the base map of each friary. Finally, documentary evidence was brought in, primarily a series of surveys (‘particulars for grant') by the Court of Augmentations, the Crown body supervising the Dissolution of the Monasteries in the 1530s and ‘40s. After setting out the historiography of research into monastic London, five chapters examine the five friaries in turn, discussing the church, cloister, precinct walls and gardens, and illustrating the evidence with a series of reconstructed plans. The chapters also examine the fate of the friary buildings in the mid-sixteenth century, after the Dissolution. In a concluding chapter, the churches and precincts are compared, looking at size, status and the use of space. The limited evidence for the economy of the friaries – both income and expenditure – is also examined. The gradual ‘secularisation' of the friaries in the fifteenth and early sixteenth centuries is also considered, before studying the purchasers of the old friary buildings in the 1540s and the uses they made of their new properties.

La dextérité d'un système de Réalité Augmentée (RA) mobile appliqué à la maintenance industrielle repose sur sa flexibilité, sa précision et sa robustesse intrinsèque face aux verrous scientifique et technologique. Nous introduisons d'abord une architecture orientée composants flexible, innovante et satisfaisant les contraintes de temps réel des systèmes de RA. Sur cette architecture est bâti un système de localisation par la vision utilisant des cibles codées. De nouveaux algorithmes d'estimation de pose sont proposés. Notre architecture comporte un Système de Gestion des Augmentations (SGA) utilisant des procédures de maintenance numériques. Chacune est liée à des documents multimédia l'illustrant, dont des modèles 3D animés. Le projet RNTL-Assistance à la Maintenance en RA intègre ce SGA. Enfin, le système est élargi à la RA en vision directe pour laquelle nous combinons une technique de prédiction du point de vue basée sur le filtre particulaire et une méthode de post-rendering.

La navigation par satellite, Global Navigation Satellite System, a été reconnue comme une solution prometteuse afin de fournir des services de navigation aux utilisateurs de l'Aviation Civile. Ces dernières années, le GNSS est devenu l'un des moyens de navigation de référence, son principal avantage étant sa couverture mondiale. Cette tendance globale est visible à bord des avions civils puisqu'une majorité d'entre eux est désormais équipée de récepteurs GNSS. Cependant, les exigences de l'Aviation Civile sont suffisamment rigoureuses et contraignantes en termes de précision de continuité, de disponibilité et d'intégrité pour que les récepteurs GPS seuls ne puissent être utilisés comme unique moyen de navigation. Cette réalité a mené à la définition de plusieurs architectures visant à augmenter les constellations GNSS. Nous pouvons distinguer les SBAS (Satellite Based Augmentation Systems), les GBAS (Ground Based Augmentation Systems), et les ABAS (Aircraft Based Augmentation Systems). Cette thèse étudie le comportement de l'erreur de position en sortie d'architectures de récepteur qui ont été identifiées comme étant très prometteuses pour les applications liées à l'Aviation Civile. / Since many years, civil aviation has identified GNSS as an attractive mean to provide navigation services for every phase of flight due to its wide coverage area. However, to do so, GNSS has to meet relevant requirements in terms of accuracy, integrity, availability and continuity. To achieve this performance, augmentation systems have been developed to correct the GNSS signals and to monitor the quality of the received Signal-In-Space (SIS). We can distinguish GBAS (Ground Based Augmentation Systems), ABAS (Airborne Based Augmentation Systems) SBAS (Satellite Based Augmentation Systems). In this context, the aim of this study is to characterize and evaluate the GNSS position error of various positioning solutions which may fulfil applicable civil aviation requirements for GNSS approaches. In particular, this study focuses on two particular solutions which are: • Combined GPS/GALILEO receivers augmented by RAIM where RAIM is a type of ABAS augmentation. This solution is a candidate to provide a mean to conduct approaches with vertical guidance (APV I, APV II and LPV 200). • GPS L1 C/A receivers augmented by GBAS. This solution should allow to conduct precision approaches down to CAT II/III, thus providing an alternative to classical radio navigation solutions such as ILS. This study deals with the characterization of the statistics of the position error at the output of these GNSS receivers. It is organised as following. First a review of civil aviation requirements is presented. Then, the different GNSS signals structure and the associated signal processing selected are described. We only considered GPS and GALILEO constellations and concentrated on signals suitable for civil aviation receivers. The next section details the GNSS measurement models used to model the measurements made by civil aviation receivers using the previous GNSS signals. The following chapter presents the GPS/GALILEO and RAIM combination model developed as well as our conclusions on the statistics of the resulting position error. The last part depicts the GBAS NSE (Navigation System Error) model proposed in this report as well as the rationales for this model.

Supervised learning requires labeled data which is cumbersome to produce, making it costly and time-consuming. SimCLR is a self-supervising framework that uses data augmentations to learn without labels. This thesis investigates how well cropping and color distorting augmentations work for two datasets, MPI3D and Causal3DIdent. The representations learned are evaluated using representation similarity analysis. The data augmentations were meant to make the model learn invariant representations of the object shape in the images regarding it as content while ignoring unnecessary features and regarding them as style. As a result, 8 models were created, models A-H. A and E were trained using supervised learning as a benchmark for the remaining self-supervised models. B and C learned invariant features of style instead of learning invariant representations of shape. Model D learned invariant representations of shape. Although, it also regarded style-related factors as content. Model F, G, and H managed to learn invariant representations of shape with varying intensities while regarding the rest of the features as style. The conclusion was that models can learn invariant representations of features related to content using self-supervised learning with the chosen augmentations. However, the augmentation settings must be suitable for the dataset. / Övervakad maskininlärning kräver annoterad data, vilket är dyrt och tidskrävande att producera. SimCLR är ett självövervakande maskininlärningsramverk som använder datamodifieringar för att lära sig utan annoteringar. Detta examensarbete utvärderar hur väl beskärning och färgförvrängande datamodifieringar fungerar för två dataset, MPI3D och Causal3DIdent. De inlärda representationerna utvärderas med hjälp av representativ likhetsanalys. Syftet med examensarbetet var att få de självövervakande maskininlärningsmodellerna att lära sig oföränderliga representationer av objektet i bilderna. Meningen med datamodifieringarna var att påverka modellens lärande så att modellen tolkar objektets form som relevant innehåll, men resterande egenskaper som icke-relevant innehåll. Åtta modeller skapades (A-H). A och E tränades med övervakad inlärning och användes som riktmärke för de självövervakade modellerna. B och C lärde sig oföränderliga representationer som bör ha betraktas som irrelevant istället för att lära sig form. Modell D lärde sig oföränderliga representationer av form men också irrelevanta representationer. Modellerna F, G och H lyckades lära sig oföränderliga representationer av form med varierande intensitet, samtidigt som de resterande egenskaperna betraktades som irrelevant. Beskärning och färgförvrängande datamodifieringarna gör således att självövervakande modeller kan lära sig oföränderliga representationer av egenskaper relaterade till relevant innehåll. Specifika inställningar för datamodifieringar måste dock vara lämpliga för datasetet.