Satellite remote sensing has revolutionized the acquisition of large amounts of data, employing both active and passive sensors to capture critical information about our planet. These data can be analysed by using deep learning methodologies that demonstrate excellent capabilities in extracting the semantics from the data. However, one of the main challenges in exploiting the power of deep learning for remote sensing applications is the lack of labeled training data. Deep learning architectures, typically demand substantial quantities of training samples to achieve optimal performance. Motivated by the above-mentioned challenges, this thesis focuses on the limited availability of labeled datasets. These challenges include issues such as ambiguous labels in case of large-scale remote sensing datasets, particularly when dealing with the analysis of multi-source satellite remote sensing images. By employing novel deep learning techniques and cutting-edge methodologies, this thesis endeavors to contribute to advancements in the field of remote sensing. In this thesis, the problems related to limited labels are solved in several ways by developing (i) a novel spectral index generative adversarial network to augment real training samples for generating class-specific remote sensing data to provide a large number of labeled samples to train a neural-network classifier; (ii) a mono- and dual-regulated contractive-expansive-contractive convolutional neural network architecture to incorporate spatial-spectral information of multispectral data and minimize the loss in the feature maps and extends this approach to the analysis of hyperspectral images; (iii) a hybrid deep learning architecture with a discrete wavelet transform and attention mechanism to deal with few labeled samples for scene-based classification of multispectral images; and (iv) a weak supervised semantic learning technique that utilises weak or low-resolution labeled samples with multisource remote sensing images for predicting pixel-wise land-use-land-cover maps. The experiments show that the proposed approaches perform better than the state-of-the-art methods on different benchmark datasets and in different conditions.

Breast cancer is one of the deadliest cancers for women. In the US, 1 in 8 women will be diagnosed with breast cancer within their lifetimes. Detection and diagnosis play an important role in saving lives. To this end, many classifiers with varying structures have been designed to classify breast cancer histopathological images. However, randomly partitioning data, like many previous works have done, can lead to artificially inflated accuracies and classifiers that do not generalize. Data leakage occurs when researchers assume that every image in a dataset is independent of each other, which is often not the case for medical datasets, where multiple images are taken of each patient. This work focuses on convolutional neural network binary classifiers using the BreakHis dataset. Previous works are reviewed. Classifiers from previous literature are tested with patient partitioning, where individual patients are placed in the training, testing and validation sets so that there is no overlap. A classifier which previously achieved 93% accuracy consistently, only achieved 79% accuracy with the new patient partition. Robust data augmentation, a Sigmoid output layer and a different form of min-max normalization were utilized to achieve an accuracy of 89.38%. These improvements were shown to be effective with the architectures used. Sigmoid Model 1.1 is shown to perform well compared to much deeper architectures found in literature.

This thesis examines the interplay of visual and textual discourse in Corporate Socia lResponsibility (CSR) reports, offering a systematic framework to analyse a dataset comprising around 66,925 images from 675 CSR reports. By analysing image attributes, colours, and objects in conjunction with textual sentiment and topics, we investigate the similarities, contrast and trends across various sectors and regions, and the impact of company characteristics. The mixed-methods approach, incorporating both qualitative image analysis and quantitative text evaluation, reveals patterns in how CSR initiatives are visually and textually communicated. Image and text extraction were accomplished using PyMuPDF and Tesseract libraries, harnessing the OCR capabilities. The identification of living objects was performed using OpenCV, while image classification was executed with the OpenAI-CLIP model, yielding high accuracy in extracting the visual content of the images. The developed framework achieved accuracy rate of 81% on living object identification using OpenCV model and 76% accuracy in object classification using OpenAI-CLIP model. The study's results indicate that the distinct patterns in how CSR is depicted, varying by sector, geographic location, and company size. These patterns offer key insights for developing more targeted and effective strategies for engaging with stakeholders.

The objective of this thesis was to investigate how the addition of a privacy-preserving mechanism to a federated learning model was affecting the performance of the model for an image classification task. Further, it was to get knowledge on how the outlook to use federated learning in the biotech industry is and what possible threats and attacks that could obstruct the utilization of federated learning among competitors. In the project four different aggregation algorithms for federated learning were examined. The methods were weighted fedAvg, unweighted FedAvg, weighted FedProx and unweighted FedProx. The experiment was using tensorflow federated to simulate the different methods. They were evaluated using accuracy, loss, recall, precision and F1 score. The result of this study shows that the performance of the deep neural network model is decreasing as differential privacy is introduced to the process. Out of the four aggregation algorithms used, weighted fedProx was the one that performed the best despite the added noise. It was also concluded that federated learning has potential to be used in the biotechnology industry among competitors, but that there are still security threats and attacks to avoid.

Detta arbete syftar till att undersöka effektiviteten av federated learning (FL) för bildklassificering i decentraliserade databehandlingsmiljöer. Med den ökande mängden av datagenerering från mobil- och ‘edge computing’, särskilt bilddata, så finns ett behov av att förbättra metoderna för bildklassificering. Dessa metoder bör inte bara adressera de utmaningar som ställs av traditionella centraliserade djupinlärningsmodeller, utan även värna om integriteten, minska kommunikationskostnaderna och övervinna skalbarhetshinder. Federated learning erbjuder en lovande lösning som tillhandahåller en ram för modellträning över decentraliserade noder med fokus på datasekretess. Denna studie analyserar FL Förmåga att förbättra bildklassificering med dess distinkta metoder, jämför dess prestanda med konventionella modeller, och granskar dess vidare implikationer och begränsningar i praktiska, verkliga inställningar. Resultatet av denna studie visar att med lämplig hantering av brus kan FL-modeller uppnå jämförbar noggrannhet med traditionella metoder, där datasekretessen förbättras betydelsefull. Vilket demonstrerar en potential balans mellan prestanda och skydd av integritet i decentraliserade miljöer. / This study aims to explore the effectiveness of Federated Learning (FL) in image classification across decentralized computing environments. With the increasing amount of data generated from mobile and edge computing, particularly image data, there is a need to improve image classification methods that not only address the challenges posed by traditional centralized deep learning models but also respect privacy, reduce communication costs, and overcome scalability barriers. Federated Learning is a promising solution that offers a framework for model training across decentralized nodes with a focus on data privacy. This study analyzes FL's capabilities to enhance image classification using its distinct methodologies, compares its performance with conventional models, and examines its wider implications and limitations in practical, real-world settings. The result of the study indicates that with appropriate noise management, FL models can achieve comparable accuracy to traditional approaches while significantly enhancing data privacy. which demonstrates a potential balance between performance and privacy protection in decentralized environments.

<p dir="ltr">This dissertation presents four distinct studies in the fields of image processing and machine learning, focusing on applications ranging from quality assessment for raster images in scanned document and virtual reality facial expression tracking to compression for continual learning and food image classification. First, we shift the traditional focus of image quality assessment (IQA) from natural images to scanned documents, proposing a machine learning-based classification method to evaluate the visual quality of scanned raster images. We enhance the classifier's performance using augmented data generated through noise models simulating scanning degradation. Second, we address the challenges of virtual facial animation in immersive VR, developing a domain adversarial training model to generate domain invariant features and combined it with manifold learning methods for accurate facial action unit (AU) intensity estimation from partially occluded facial images. Third, we explore the use of image compression to increase buffer capacity in continual machine learning systems, thereby enhancing exemplar diversity and mitigating catastrophic forgetting. Our approach includes a new framework that selects compression rate and algorithm, showing significant improvements in image classification accuracy on the CIFAR-100 and ImageNet datasets. Finally, we combine class-activation maps with neural image compression in food image classification systems to adapt to continuously evolving data, extending buffer size and enhancing data diversity, which is validated on food-specific datasets and shows potential for broader applications in continual machine learning systems. Together, these studies demonstrate the versatility of image processing and machine learning techniques in addressing complex and varied challenges across different domains.</p>

Master of Arts / Department of Geography / Kevin P. Price / Due primarily to changes in land management practices, eastern redcedar (Juniperus virginiana L.), a native Kansas conifer, is rapidly invading onto valuable rangelands. The suppression of fire and increase of intensive grazing, combined with the rapid growth rate, high reproductive output, and dispersal ability of the species have allowed it to dramatically expand beyond its original range. There is a growing interest in harvesting this species for use as a biofuel. For economic planning purposes, density and biomass quantities for the trees are needed. Three methods are explored for mapping eastern redcedar and quantifying its biomass in Riley County, Kansas. First, a land cover classification of redcedar cover is performed using a method that utilizes a support vector machine classifier applied to a multi-temporal stack of Landsat TM satellite images. Second, a Small Unmanned Aircraft System (sUAS) is used to measure individual redcedar trees in an area where they are encroaching into a pasture. Finally, a hybrid approach is used to estimate redcedar biomass using high resolution multispectral and light detection and ranging (LiDAR) imagery. These methods showed promise in the forestry, range management, and bioenergy industries for better understanding of an invasive species that shows great potential for use as a biofuel resource.

Corneal diseases are one of the major causes of visual impairment and blindness worldwide. Used for diagnoses, a laser confocal microscope provides a sequence of images, at incremental depths, of the various corneal layers and structures. From these, ophthalmologists can extract clinical information on the state of health of a patient’s cornea. However, many factors impede ophthalmologists in forming diagnoses starting with the large number and variable quality of the individual images (blurring, non-uniform illumination within images, variable illumination between images and noise), and there are also difficulties posed for automatic processing caused by eye movements in both lateral and axial directions during the scanning process. Aiding ophthalmologists working with long sequences of corneal image requires the development of new algorithms which enhance, correctly order and register the corneal images within a sequence. The novel algorithms devised for this purpose and presented in this thesis are divided into four main categories. The first is enhancement to reduce the problems within individual images. The second is automatic image classification to identify which part of the cornea each image belongs to, when they may not be in the correct sequence. The third is automatic reordering of the images to place the images in the right sequence. The fourth is automatic registration of the images with each other. A flexible application called CORNEASYS has been developed and implemented using MATLAB and the C language to provide and run all the algorithms and methods presented in this thesis. CORNEASYS offers users a collection of all the proposed approaches and algorithms in this thesis in one platform package. CORNEASYS also provides a facility to help the research team and Ophthalmologists, who are in discussions to determine future system requirements which meet clinicians’ needs.

Durant cette thèse, nous abordons le problème de la fouille exhaustive de motifs pour un cas particulier de graphes : les grilles. Ces grilles peuvent être utilisées pour modéliser des objets ayant une structure régulière. Ces structures sont naturellement présentes dans de nombreux jeux de plateaux (les dames, les échecs ou le go par exemple) ou encore dans les modélisations d’écosystèmes utilisant des automates cellulaires. On les retrouve également à un plus bas niveau dans les images, qui sont des grilles 2D de pixels ou encore les vidéos, qui sont des grilles spatio-temporelles 2D+t de pixels. Au cours de cette thèse, nous avons proposé un nouvel algorithme de fouille de motifs fréquents dédié aux grilles spatio-temporelles, GriMA. L’usage des grilles régulières permet à notre algorithme de réduire la complexité des tests d’isomorphismes. Ces tests sont souvent utilisés par les algorithmes génériques de fouilles de graphes mais ayant une complexité importante, cela limite leur usage sur des données réelles. Deux applications ont été proposées pour évaluer notre algorithme : la classification d’images pour la fouille de grilles 2D et la prédiction d’automates cellulaires pour la fouille de grilles 2D+t. / During this thesis, we consider the exhaustive graph mining problem for a special kind of graphs : the grids. Theses grids can be used to model objects that present a regular structure. These structures are naturally present in multiple board games (checkers, chess or go for instance) or in ecosystems models using cellular automata. It is also possible to find this structure in a lower level in images, which are 2D grids of pixels, or even in videos, which are 2D+t spatio-temporal grids of pixels. In this thesis, we proposed a new algorithm to find frequent patterns dedicated to spatio-temporal grids, GriMA. Use of regular grids allow our algorithm to reduce the complexity of the isomorphisms test. These tests are often use by generic graph mining algorithm but because of their complexity, they are rarely used on real data. Two applications were proposed to evaluate our algorithm: image classification for 2D grids mining and prediction of cellular automata for 2D+t grids mining.

Plâncton são organismos microscópicos que constituem a base da cadeia alimentar de ecossistemas aquáticos. Eles têm importante papel no ciclo do carbono pois são os responsáveis pela absorção do carbono na superfície dos oceanos. Detectar, estimar e monitorar a distribuição das diferentes espécies são atividades importantes para se compreender o papel do plâncton e as consequências decorrentes de alterações em seu ambiente. Parte dos estudos deste tipo é baseada no uso de técnicas de imageamento de volumes de água. Devido à grande quantidade de imagens que são geradas, métodos computacionais para auxiliar no processo de análise das imagens estão sob demanda. Neste trabalho abordamos o problema de identificação da espécie. Adotamos o pipeline convencional que consiste dos passos de detecção de alvo, segmentação (delineação de contorno), extração de características, e classificação. Na primeira parte deste trabalho abordamos o problema de escolha de um algoritmo de segmentação adequado. Uma vez que a avaliação de resultados de segmentação é subjetiva e demorada, propomos um método para avaliar algoritmos de segmentação por meio da avaliação da classificação no final do pipeline. Experimentos com esse método mostraram que algoritmos de segmentação distintos podem ser adequados para a identificação de espécies de classes distintas. Portanto, na segunda parte do trabalho propomos um método de classificação que leva em consideração múltiplas segmentações. Especificamente, múltiplas segmentações são calculadas e classificadores são treinados individualmente para cada segmentação, os quais são então combinados para construir o classificador final. Resultados experimentais mostram que a acurácia obtida com a combinação de classificadores é superior em mais de 2% à acurácia obtida com classificadores usando uma segmentação fixa. Os métodos propostos podem ser úteis para a construção de sistemas de identificação de plâncton que sejam capazes de se ajustar rapidamente às mudanças nas características das imagens. / Plankton are microscopic organisms that constitute the basis of the food chain of aquatic ecosystems. They have an important role in the carbon cycle as they are responsible for the absorption of carbon in the ocean surfaces. Detecting, estimating and monitoring the distribution of plankton species are important activities for understanding the role of plankton and the consequences of changes in their environment. Part of these type of studies is based on the analysis of water volumes by means of imaging techniques. Due to the large quantity of generated images, computational methods for helping the process of image analysis are in demand. In this work we address the problem of species identification. We follow the conventional pipeline consisting of target detection, segmentation (contour delineation), feature extraction, and classification steps. In the first part of this work we address the problem of choosing an appropriate segmentation algorithm. Since evaluating segmentation results is a subjective and time consuming task, we propose a method to evaluate segmentation algorithms by evaluating the classification results at the end of the pipeline. Experiments with this method showed that distinct segmentation algorithms might be appropriate for identifying species of distinct classes. Therefore, in the second part of this work we propose a classification method that takes into consideration multiple segmentations. Specifically, multiple segmentations are computed and classifiers are trained individually for each segmentation, which are then combined to build the final classifier. Experimental results show that the accuracy obtained with the combined classifier is superior in more than 2% to the accuracy obtained with classifiers using a fixed segmentation. The proposed methods can be useful to build plankton identification systems that are able to quickly adjust to changes in the characteristics of the images.