Observing the earth from above is a great way of understanding our world better. From space, many complex patterns and relationships on the ground can be identified through high-quality satellite data. The quality and availability of this data in combination with recent advancement in various deep learning techniques allows us to find these patterns more effectively then ever. In this thesis, we will analyze satellite imagery by using deep neural networks in an attempt to find road networks in different cities around the world. Once we have located networks of roads in the cities we will represent them as graphs and deploy the Dijkstra shortest path algorithm to find optimal routes within these networks. Having the ability to efficiently use satellite imagery for near real-time road detection and optimal route prediction has many possible applications, especially from a humanitarian and commercial point of view. For example, in the humanitarian realm, the frequency of natural disasters is unfortunately increasing due to climate change and the need for emergency real-time mapping for relief organisations in the case of a severe flood or similar is growing.  The state-of-the-art deep neural network models that will be implemented, compared and contrasted for this task are mainly based on the U-net and ResNet architectures. However, before introducing these architectures the reader will be given a comprehensive introduction and theoretical background of deep neural networks to distinctly formulate the mathematical groundwork. The final results demonstrates an overall strong model performance across different metrics and data sets, with the highest obtained IoU-score being approximately 0.7 for the segmentation task. For some models we can also see a high degree of similarity between the predicted optimal paths and the ground truth optimal paths. / Att betrakta jorden från ovan är ett bra tillvägagångsätt för att förstå vår egen värld bättre. Från rymden, många komplexa mönster och samband på marken går att urskilja genom hög-upplöst satellitdata. Kvalitén och tillgängligheten av denna data, i kombination med de senaste framstegen inom djupa inlärningstekniker, möjliggör oss att hissa dessa mönster mer effektivt än någonsin. I denna avhandling kommer vi analysera satellitbilder med hjälp av djupa neurala nätverk i ett försök att hitta nätverk av vägar i olika städer runtom i världen. Efter vi har lokaliserat dessa nätverk av vägar så kommer vi att representera nätverken som grafer och använda oss av Dijkstras algoritm för att hitta optimala rutter inom dessa nätverk.  Att ha förmågan att kunna effektivt använda sig av satellitbilder för att i nära realtid kunna identifiera vägar och optimala rutter har många möjliga applikationer. Speciellt ur ett humant och kommersiellt perspektiv. Exempelvis, inom det humanitära området, så ökar dessvärre frekvensen av naturkatastrofer på grund av klimatförändringar och därmed är behovet av nödkartläggning i realtid för hjälporganisationer större än någonsin. En effektiv nödkartläggning skulle exempelvis kunna underlätta enormt vid en allvarlig översvämning eller dylikt.  Dem toppmoderna djupa neurala nätverksmodellerna som kommer implementeras, jämföras och nyanseras för denna uppgift är i huvudsak baserad på U-net och ResNet arkitekturerna. Innan vi presenterar dessa arkitekturer i denna avhandling så kommer läsaren att få en omfattande teoretisk bakgrund till djupa neurala nätverk för att tydligt formulera dem matematiska grundpelarna. Dem slutgiltiga resultaten visar övergripande stark prestanda för samtliga av våra modeller. Både på olika datauppsättningar samt utvärderingsmått. Den högste IoU poängen som uppnås är cirka 0,7 och vi kan även se en hög grad av likhet mellan vissa av våra förutsagda optimala rutter och mark sanningens optimala rutter.

Simultaneous Localization and Mapping (SLAM) is the crucial ability for many autonomous systems to operate in unknown environments. In recent years SLAM development has focused on achieving robustness regarding the challenges the field still faces e.g. dynamic environments. During this thesis work different existing approaches to tackle dynamics with Visual SLAM systems were analyzed by surveying the recent literature within the field. The goal was to define the advantages and drawbacks of the approaches to provide further insight into the field of dynamic SLAM. Furthermore, two methods of different approaches were chosen for experiments and their implementation was documented. Key conclusions from the literature survey and experiments are the following. The exclusion of dynamic objects with regard to camera pose estimation presents promising results. Tracking of dynamic objects provides valuable information when combining SLAM with other tasks e.g. path planning. Moreover, dynamic reconstruction with SLAM offers better scene understanding and analysis of objects’ behavior within an environment. Many solutions rely on pre-processing and heavy hardware requirements due to the nature of the object detection methods. Methods of motion confirmation of objects lack consideration of camera movement, resulting in static objects being excluded from feature extraction. Considerations for future work within the field include accounting for camera movement for motion confirmation and producing available benchmarks that offer evaluation of the SLAM result as well as the dynamic object detection i.e. ground truth for both camera and objects within the scene. / Simultaneous Localization and Mapping (SLAM) är för många autonoma system avgörande för deras förmåga att kunna verka i tidigare outforskade miljöer. Under de senaste åren har SLAM-utvecklingen fokuserat på att uppnå robusthet när det gäller de utmaningar som fältet fortfarande står inför, t.ex. dynamiska miljöer. I detta examensarbete analyserades befintliga metoder för att hantera dynamik med visuella SLAM-system genom att kartlägga den senaste litteraturen inom området. Målet var att definiera för- och nackdelar hos de olika tillvägagångssätten för att bidra med insikter till området dynamisk SLAM. Dessutom valdes två metoder från olika tillvägagångssätt ut för experiment och deras implementering dokumenterades. De viktigaste slutsatserna från litteraturstudien och experimenten är följande. Uteslutningen av dynamiska objekt vid uppskattning av kamerans position ger lovande resultat. Spårning av dynamiska objekt ger värdefull information när SLAM kombineras med andra uppgifter, t.ex. path planning. Dessutom ger dynamisk rekonstruktion med SLAM bättre förståelse om omgivningen och analys av objekts beteende i den kringliggande miljön. Många lösningar är beroende av förbehandling samt ställer höga hårdvarumässiga krav till följd av objektdetekteringsmetodernas natur. Metoder för rörelsebekräftelse av objekt tar inte hänsyn till kamerarörelser, vilket leder till att statiska objekt utesluts från funktionsextraktion. Uppmaningar för framtida studier inom området inkluderar att ta hänsyn till kamerarörelser under rörelsebekräftelse samt att ta ändamålsenliga riktmärken för att möjliggöra tydligare utvärdering av SLAM-resultat såväl som för dynamisk objektdetektion, dvs. referensvärden för både kamerans position såväl som för objekt i scenen.

An investigation into using attention mechanisms for better feature extraction in wildfire spread prediction models. This research examines the U-net architecture to achieve image segmentation, a process that partitions images by classifying pixels into one of two classes. The deep learning models explored in this research integrate modern deep learning architectures, and techniques used to optimize them. The models are trained on 12 distinct observational variables derived from the Google Earth Engine catalog. Evaluation is conducted with accuracy, Dice coefficient score, ROC-AUC, and F1-score. This research concludes that when augmenting U-net with attention mechanisms, the attention component improves feature suppression and recognition, improving overall performance. Furthermore, employing ensemble modeling reduces bias and variation, leading to more consistent and accurate predictions. When inferencing on wildfire propagation at 30-minute intervals, the architecture presented in this research achieved a ROC-AUC score of 86.2% and an accuracy of 82.1%.

Detecting water under vegetation is critical to tracking the status of geological ecosystems like wetlands. Researchers use different methods to estimate water presence, avoiding costly on-site measurements. Optical satellite imagery allows the automatic delineation of water using the concept of the Normalised Difference Water Index (NDWI). Still, optical imagery is subject to visibility conditions and cannot detect water under the vegetation, a typical situation for wetlands. Synthetic Aperture Radar (SAR) imagery works under all visibility conditions. It can detect water under vegetation but requires deep network algorithms to segment water presence, and manual annotation work is required to train the deep models. This project uses DEEPAQUA, a cross-modal knowledge distillation method, to eliminate the manual annotation needed to extract water presence from SAR imagery with deep neural networks. In this method, a deep student model (e.g., UNET) is trained to segment water in SAR imagery. The student model uses the NDWI algorithm as the non-parametric, cross-modal teacher. The key prerequisite is that NDWI works on the optical imagery taken from the exact location and simultaneously as the SAR. Three different deep architectures are tested in this project: UNET, SegNet, and UNET++, and the Otsu method is used as the baseline. Experiments on imagery from Swedish wetlands in 2020-2022 show that cross-modal distillation consistently achieved better segmentation performances across architectures than the baseline. Additionally, the UNET family of algorithms performed better than SegNet with a confidence of 95%. The UNET++ model achieved the highest Intersection Over Union (IOU) performance. However, no statistical evidence emerged that UNET++ performs better than UNET, with a confidence of 95%. In conclusion, this project shows that cross-modal knowledge distillation works well across architectures and removes tedious and expensive manual work hours when detecting water from SAR imagery. Further research could evaluate performances on other datasets and student architectures. / Att upptäcka vatten under vegetation är avgörande för att hålla koll på statusen på geologiska ekosystem som våtmarker. Forskare använder olika metoder för att uppskatta vattennärvaro vilket undviker kostsamma mätningar på plats. Optiska satellitbilder tillåter automatisk avgränsning av vatten med hjälp av konceptet Normalised Difference Water Index (NDWI). Optiska bilder fortfarande beroende av siktförhållanden och kan inte upptäcka vatten under vegetationen, en typisk situation för våtmarker. Synthetic Aperture Radar (SAR)-bilder fungerar under alla siktförhållanden. Den kan detektera vatten under vegetation men kräver djupa nätverksalgoritmer för att segmentera vattennärvaro, och manuellt anteckningsarbete krävs för att träna de djupa modellerna. Detta projekt använder DEEPAQUA, en cross-modal kunskapsdestillationsmetod, för att eliminera det manuella annoteringsarbete som behövs för att extrahera vattennärvaro från SAR-bilder med djupa neurala nätverk. I denna metod tränas en djup studentmodell (t.ex. UNET) att segmentera vatten i SAR-bilder semantiskt. Elevmodellen använder NDWI, som fungerar på de optiska bilderna tagna från den exakta platsen och samtidigt som SAR, som den icke-parametriska, cross-modal lärarmodellen. Tre olika djupa arkitekturer testas i detta examensarbete: UNET, SegNet och UNET++, och Otsu-metoden används som baslinje. Experiment på bilder tagna på svenska våtmarker 2020-2022 visar att cross-modal destillation konsekvent uppnådde bättre segmenteringsprestanda över olika arkitekturer jämfört med baslinjen. Dessutom presterade UNET-familjen av algoritmer bättre än SegNet med en konfidens på 95%. UNET++-modellen uppnådde högsta prestanda för Intersection Over Union (IOU). Det framkom dock inga statistiska bevis för att UNET++ presterar bättre än UNET, med en konfidens på 95%. Sammanfattningsvis visar detta projekt att cross-modal kunskapsdestillation fungerar bra över olika arkitekturer och tar bort tidskrävande och kostsamma manuella arbetstimmar vid detektering av vatten från SAR-bilder. Ytterligare forskning skulle kunna utvärdera prestanda på andra datamängder och studentarkitekturer.

In a world under the constant impact of global warming, wetlands are decreasing in size all across the globe. As the wetlands are a vital part of preventing global warming, the ability to prevent their shrinkage through restorative measures is critical. Continuously orbiting the Earth are satellites that can be used to monitor the wetlands by collecting images of them over time. In order to determine the size of a wetland, and to register if it is shrinking or not, deep learning models can be used. Especially useful for this task is convolutional neural networks (CNNs). This project uses one type of CNN, a U-Net, to segment vegetated water in satellite data. However, this task requires labeled data, which is expensive to generate and difficult to acquire. The model used therefore needs to be able to generate reliable results even on small data sets. Therefore, pre-training of the network is used with a large-scale natural image segmentation data set called Common Objects in Context (COCO). To transfer the satellite data into RGB images to use as input for the pre-trained network, three different methods are tried. Firstly, the commonly used linear transformation method which simply moves the value of radar data into the RGB feature space. Secondly, two convolutional layers are placed before the U-Net which gradually changes the number of channels of the input data, with weights trained through backpropagation during the fine-tuning of the segmentation model. Lastly, a convolutional auto-encoder is trained in the same way as the convolutional layers. The results show that the autoencoder does not perform very well, but that the linear transformation and convolutional layers methods each can outperform the other depending on the data set. No statistical significance can be shown however between the performance of the two latter. Experimenting with including different amounts of polarizations from Sentinel-1 and bands from Sentinel-2 showed that only using radar data gave the best results. It remains to be determined whether one or both of the polarizations should be included to achieve the best result. / I en värld som ständigt påverkas av den globala uppvärmningen, minskar våtmarkerna i storlek över hela världen. Eftersom våtmarkerna är en viktig del i att förhindra global uppvärmning, är förmågan att förhindra att de krymper genom återställande åtgärder kritisk. Kontinuerligt kretsande runt jorden finns satelliter som kan användas för att övervaka våtmarkerna genom att samla in bilder av dem över tid. För att bestämma storleken på en våtmark, i syfte att registrera om den krymper eller inte, kan djupinlärningsmodeller användas. Speciellt användbar för denna uppgift är konvolutionella neurala nätverk (CNN). Detta projekt använder en typ av CNN, ett U-Net, för att segmentera vegeterat vatten i satellitdata. Denna uppgift kräver dock märkt data, vilket är dyrt att generera och svårt att få tag på. Modellen som används behöver därför kunna generera pålitliga resultat även med små datauppsättning. Därför används förträning av nätverket med en storskalig naturlig bildsegmenteringsdatauppsättning som kallas Common Objects in Context (COCO). För att överföra satellitdata till RGB-bilder som ska användas som indata för det förtränade nätverket prövas tre olika metoder. För det första, den vanliga linjära transformationsmetoden som helt enkelt flyttar värdet av radardatan till RGB-funktionsutrymmet. För det andra två konvolutionella lager placerade före U-Net:et som gradvis ändrar mängden kanaler i indatan, med vikter tränade genom bakåtpropagering under finjusteringen av segmenteringsmodellen. Slutligen tränade en konvolutionell auto encoder på samma sätt som de konvolutionella lagren. Resultaten visar att auto encodern inte fungerar särskilt bra, men att metoderna för linjär transformation och konvolutionella lager var och en kan överträffa den andra beroende på datauppsättningen. Ingen statistisk signifikans kan dock visas mellan prestationen för de två senare. Experiment med att inkludera olika mängder av polariseringar från Sentinell-1 och band från Sentinell-2 visade att endast användning av radardata gav de bästa resultaten. Om att inkludera båda polariseringarna eller bara en är den mest lämpliga återstår fortfarande att fastställa.

Histological slides are an important tool in the diagnosis of tumors as well as of other diseases that affect cell shapes and distributions. Until now, the research concerning an optimal staining time has been mainly done empirically. In experimental investigations, it is often not possible to stain an already-stained slide with another stain to receive further information. To overcome these challenges, in the present paper a continuum-based model was developed for conducting a virtual (re-)staining of a scanned histological slide. This model is capable of simulating the staining of cell nuclei with the dye hematoxylin (C.I. 75,290). The transport and binding of the dye are modeled (i) along with the resulting RGB intensities (ii). For (i), a coupled diffusion–reaction equation is used and for (ii) Beer–Lambert’s law. For the spatial discretization an approach based on the finite element method (FEM) is used and for the time discretization a finite difference method (FDM). For the validation of the proposed model, frozen sections from human liver biopsies stained with hemalum were used. The staining times were varied so that the development of the staining intensity could be observed over time. The results show that the model is capable of predicting the staining process. The model can therefore be used to perform a virtual (re-)staining of a histological sample. This allows a change of the staining parameters without the need of acquiring an additional sample. The virtual standardization of the staining is the first step towards universal cross-site comparability of histological slides.

Recent advances in deep learning have greatly improved the ability to generate analysis models from medical images. In particular, great attention is focused on quickly generating models of the left ventricle from cardiac magnetic resonance imaging (cMRI) to improve the diagnosis and prognosis of millions of patients. However, even state-of-the-art frameworks present challenges, such as discontinuities of the cardiac tissue and excessive jaggedness along the myocardial walls. These geometrical features are often anatomically incorrect and may lead to unrealistic results once the geometrical models are employed in computational analyses. In this research, we propose an end-to-end pipeline for a subject-specific model of the heart's left ventricle from Cine cMRI. Our novel pipeline incorporates the uncertainty originating from the segmentation methods in the estimation of cardiac indices, such as ejection fraction, myocardial volume changes, and global radial and longitudinal strain, during the cardiac cycle. First, we propose an anatomically-guided deep learning model to overcome the common segmentation challenges while preserving the advantages of state-of-the-art frameworks, such as computational efficiency, robustness, and abstraction capabilities. Our anatomically-guided neural networks include a B-spline head, which acts as a regularization layer during training. In addition, the introduction of the B-spline head contributes to achieving a robust uncertainty quantification of the left ventricle inner and outer walls. We validate our approach using human short-axis (SA) cMRI slices and later apply transfer learning to verify its generalization capabilities in swine long-axis (LA) cMRI slices. Finally, we use the SA and LA contours to build a Gaussian Process (GP) model to create inner and outer walls 3D surfaces, which are then used to compute global indices of cardiac functions. Our results show that the proposed pipeline generates anatomically consistent geometries while also providing a robust tool for quantifying uncertainty in the geometry and the derived cardiac indices.

<div> <div> <div> <p>An RGBZ synthetic dataset consisting of five object classes in a variety of virtual environments and orientations was combined with a small sample of real-world image data and used to train the Mask R-CNN (MR-CNN) architecture in a variety of configurations. When the MR-CNN architecture was initialized with MS COCO weights and the heads were trained with a mix of synthetic data and real world data, F1 scores improved in four of the five classes: The average maximum F1-score of all classes and all epochs for the networks trained with synthetic data is F1∗ = 0.91, compared to F1 = 0.89 for the networks trained exclusively with real data, and the standard deviation of the maximum mean F1-score for synthetically trained networks is σ∗ _F1= 0.015, compared to σF 1 = 0.020 for the networks trained exclusively with real data. Various backgrounds in synthetic data were shown to have negligible impact on F1 scores, opening the door to abstract backgrounds and minimizing the need for intensive synthetic data fabrication. When the MR-CNN architecture was initialized with MS COCO weights and depth data was included in the training data, the net- work was shown to rely heavily on the initial convolutional input to feed features into the network, the image depth channel was shown to influence mask generation, and the image color channels were shown to influence object classification. A set of latent variables for a subset of the synthetic datatset was generated with a Variational Autoencoder then analyzed using Principle Component Analysis and Uniform Manifold Projection and Approximation (UMAP). The UMAP analysis showed no meaningful distinction between real-world and synthetic data, and a small bias towards clustering based on image background. </p></div></div></div>

Given satellite images with accompanying pixel classifications and elevation data, we propose different solutions to object detection. The first method uses hierarchical clustering for segmentation and then employs different methods of classification. One of these classification methods used domain knowledge to classify objects while the other used Support Vector Machines. Additionally, a combination of three Support Vector Machines were used in a hierarchical structure which out-performed the regular Support Vector Machine method in most of the evaluation metrics. The second approach is more conventional with different types of Convolutional Neural Networks. A segmentation network was used as well as a few detection networks and different fusions between these. The Convolutional Neural Network approach proved to be the better of the two in terms of precision and recall but the clustering approach was not far behind. This work was done using a relatively small amount of data which potentially could have impacted the results of the Machine Learning models in a negative way.

Ce travail de thèse concerne l'analyse structurelle des maillages triangulaires surfaciques, ainsi que leur traitement en vue de l'amélioration de leur qualité (remaillage) ou de leur simplification. Dans la littérature, le repositionnement des sommets d'un maillage est soit traité de manière locale, soit de manière globale mais sans un contrôle local de l'erreur géométrique introduite, i.e. les solutions actuelles ne sont pas globales ou introduisent de l'erreur géométrique non-contrôlée. Les techniques d'approximation de maillage les plus prometteuses se basent sur une décomposition en primitives géométriques simples (plans, cylindres, sphères etc.), mais elles n'arrivent généralement pas à trouver la décomposition optimale, celle qui optimise à la fois l'erreur géométrique de l'approximation par les primitives choisies, et le nombre et le type de ces primitives simples. Pour traiter les défauts des approches de remaillage existantes, nous proposons une méthode basée sur un modèle global, à savoir une modélisation graphique probabiliste, intégrant des contraintes souples basées sur la géométrie (l'erreur de l'approximation), la qualité du maillage et le nombre de sommets du maillage. De même, pour améliorer la décomposition en primitives simples, une modélisation graphique probabiliste a été choisie. Les modèles graphiques de cette thèse sont des champs aléatoires de Markov, ces derniers permettant de trouver une configuration optimale à l'aide de la minimisation globale d'une fonction objectif. Nous avons proposé trois contributions dans cette thèse autour des maillages triangulaires 2-variétés : (i) une méthode d'extraction statistiquement robuste des arêtes caractéristiques applicable aux objets mécaniques, (ii) un algorithme de segmentation en régions approximables par des primitives géométriques simples qui est robuste à la présence de données aberrantes et au bruit dans la position des sommets, (iii) et finalement un algorithme d'optimisation de maillages qui cherche le meilleur compromis entre l'amélioration de la qualité des triangles, la qualité de la valence des sommets, le nombre de sommets et la fidélité géométrique à la surface initiale. / The work in this thesis concerns structural analysis of 2-manifold triangular meshes, and their processing towards quality enhancement (remeshing) or simplification. In existing work, the repositioning of mesh vertices necessary for remeshing is either done locally or globally, but in the latter case without local control on the introduced geometrical error. Therefore, current results are either not globally optimal or introduce unwanted geometrical error. Other promising remeshing and approximation techniques are based on a decomposition into simple geometrical primitives (planes, cylinders, spheres etc.), but they generally fail to find the best decomposition, i.e. the one which jointly optimizes the residual geometrical error as well as the number and type of selected simple primitives. To tackle the weaknesses of existing remeshing approaches, we propose a method based on a global model, namely a probabilistic graphical model integrating soft constraints based on geometry (approximation error), mesh quality and the number of mesh vertices. In the same manner, for segmentation purposes and in order to improve algorithms delivering decompositions into simple primitives, a probabilistic graphical modeling has been chosen. The graphical models used in this work are Markov Random Fields, which allow to find an optimal configuration by a global minimization of an objective function. We have proposed three contributions in this thesis about 2-manifold triangular meshes : (i) a statistically robust method for feature edge extraction for mechanical objects, (ii) an algorithm for the segmentation into regions which are approximated by simple primitives, which is robust to outliers and to the presence of noise in the vertex positions, (iii) and lastly an algorithm for mesh optimization which jointly optimizes triangle quality, the quality of vertex valences, the number of vertices, as well as the geometrical fidelity to the initial surface.