This thesis delves into eye tracking, a technique for estimating an individual’s point of gaze and understanding human interactions with the environment. A blossoming area within eye tracking is appearance-based eye tracking, which leverages deep neural networks to predict gaze positions from eye images. Despite its efficacy, the decision-making processes inherent in deep neural networks remain as ’black boxes’ to humans. This lack of transparency challenges the trust human professionals place in the predictions of appearance-based eye tracking models. To address this issue, explainable AI is introduced, aiming to unveil the decision-making processes of deep neural networks and render them comprehensible to humans. This thesis employs various post-hoc explainable AI methods, including saliency maps, gradient-weighted class activation mapping, and guided backpropagation, to generate heat maps of eye images. These heat maps reveal discriminative areas pivotal to the model’s gaze predictions, and glints emerge as of paramount importance. To explore additional features in gaze estimation, a glint-free dataset is derived from the original glint-preserved dataset by employing blob detection to eliminate glints from each eye image. A corresponding glint-free model is trained on this dataset. Cross-evaluations of the two datasets and models discover that the glint-free model extracts complementary features (pupil, iris, and eyelids) to the glint-preserved model (glints), with both feature sets exhibiting comparable intensities in heat maps. To make use of all the features, an augmented dataset is constructed, incorporating selected samples from both glint-preserved and glint-free datasets. An augmented model is then trained on this dataset, demonstrating a superior performance compared to both glint-preserved and glint-free models. The augmented model excels due to its training process on a diverse set of glint-preserved and glint-free samples: it prioritizes glints when of high quality, and adjusts the focus to the entire eye in the presence of poor glint quality. This exploration enhances the understanding of the critical factors influencing gaze prediction and contributes to the development of more robust and interpretable appearance-based eye tracking models. / Denna avhandling handlar om ögonspårning, en teknik för att uppskatta en individs blickpunkt och förstå människors interaktioner med miljön. Ett viktigt område inom ögonspårning är bildbaserad ögonspårning, som utnyttjar djupa neuronnät för att förutsäga blickpositioner från ögonbilder. Trots dess effektivitet förblir beslutsprocesserna i djupa neuronnät som ”svarta lådor” för människor. Denna brist på transparens utmanar det förtroende som yrkesverksamma sätter i förutsägelserna från bildbaserade ögonspårningsmodeller. För att ta itu med detta problem introduceras förklarbar AI, med målet att avslöja beslutsprocesserna hos djupa neuronnät och göra dem begripliga för människor. Denna avhandling använder olika efterhandsmetoder för förklarbar AI, inklusive saliency maps, gradient-weighted class activation mapping och guidad backpropagation, för att generera värmekartor av ögonbilder. Dessa värmekartor avslöjar områden som är avgörande för modellens blickförutsägelser, och ögonblänk framstår som av yttersta vikt. För att utforska ytterligare funktioner i blickuppskattning, härleds ett dataset utan ögonblänk från det ursprungliga datasetet genom att använda blobdetektering för att eliminera blänk från varje ögonbild. En motsvarande blänkfri modell tränas på detta dataset. Korsutvärderingar av de två datamängderna och modellerna visar att den blänkfria modellen tar fasta på kompletterande särdrag (pupill, iris och ögonlock) jämfört med den blänkbevarade modellen, men båda modellerna visar jämförbara intensiteter i värmekartorna. För att utnyttja all information konstrueras ett förstärkt dataset, som inkorporerar utvalda exempel från både blänkbevarade och blänkfria dataset. En förstärkt modell tränas sedan på detta dataset, och visar överlägsen prestanda jämfört med de båda andra modellerna. Den förstärkta modellen utmärker sig på grund av sin träning på en mångfaldig uppsättning av exempel med och utan blänk: den prioriterar blänk när de är av hög kvalitet och justerar fokuset till hela ögat vid dålig blänkkvalitet. Detta arbete förbättrar förståelsen för de kritiska faktorerna som påverkar blickförutsägelse och bidrar till utvecklingen av mer robusta och tolkningsbara modeller för bildbaserad ögonspårning.

The emergence of generative AI has sparked significant hype in recent years, fueled by its ability to perform tasks autonomously and augment human capabilities. This technology has led to excitement in organizations all over the world, with promises of revolutionizing various industries. This study aims to explore the effects of generative AI in workplace environments and how different professional roles may be affected by this. Furthermore, this study seeks to shine light on possible barriers to achieve the positive effects that generative AI enables. More specifically, this study focuses on effects of the generative AI tool Copilot for M365 within a large global organization. By employing qualitative methods including interviews with employees who have been using Copilot in the workplace, this research aspires to uncover how humans interact with AI and what the effects of this might be. The results show that Copilots primarily induces efficiency through automation in the researched organization, while also suggesting signs of augmenting human cognition, where creativity and innovation emerge as effects. Furthermore, the findings indicate that there are no significant profession-specific effects. Lastly, the study has gained insight that there are barriers to achieve the full potential and value of generative AI.

The use of AI in the medical field is becoming more widespread, and research on its various applications is very popular. In biomedical image analysis, Convolutional Neural Networks (CNN), which are specialized in image processing, can analyze X-rays and detect signs of different diseases. However, to achieve that, CNNs require vast amounts of X-ray images with labels specifying the disease (labeled training data), which is not always available. One method to overcome this obstacle is the use of data augmentation. Data augmentation is manipulating images through flipping, rotating, or changing the saturation or brightness, among other methods. The purpose is to increase and diversify the training data to make the CNN model more robust. Our study aims to investigate the effects of different data augmentation techniques on the performance of a CNN model in detecting Pneumothorax. After fine-tuning our CNN model’s hyper-parameters, three data augmentation methods (color, geometric, and noise) and their combinations were applied to our model. We then tested and compared the effects of each data augmentation method on the accuracy of our model. Our study concluded that color augmentation performed the best compared to the other augmentation methods, while geometric augmentation had the worst performance. However, none of the augmentation methods significantly improved the original model’s performance, which can be attributed to the model’s configuration of hyper-parameters, leaving no room for improvement. / Användningen av AI inom det medicinska området blir mer utbredd och forskning om dess olika tillämpningar är mycket populär. Inom biomedicinsk bildanalys kan Convolutional Neural Networks (CNN), som är specialiserade på bildbehandling, analysera röntgenstrålar och upptäcka tecken på olika sjukdomar. Men för att uppnå det kräver CNN stora mängder röntgenbilder med etiketter som anger sjukdomen (märkta träningsdata), vilket inte alltid är tillgängligt. En metod för att övervinna detta hinder är användningen av dataförstärkning. Dataförstärkning är att manipulera bilder genom att bläddra, rotera eller ändra mättnad eller ljusstyrka, bland andra metoder. Syftet är att öka och diversifiera träningsdata för att göra CNN-modellen mer robust. Vår studie syftar till att undersöka effekterna av olika dataförstärkningstekniker på prestandan hos en CNN-modell vid detektering av pneumothorax. Efter att ha finjusterat vår CNN-modells hyperparametrar, tillämpades tre dataförstärkningsmetoder (färg, geometrisk och brus) och deras kombinationer på vår modell. Vi testade och jämförde sedan effekterna av varje dataförstärkningsmetod på noggrannheten i vår modell. Vår studie drog slutsatsen att färgförstärkning presterade bäst jämfört med andra förstärkningsmetoder, medan geometrisk förstärkning hade sämst prestanda. Ingen av förstärkningsmetoderna förbättrade dock den ursprungliga modellens prestanda avsevärt, vilket kan tillskrivas modellens konfiguration av hyperparametrar, vilket inte lämnar något utrymme för förbättringar.

This thesis addresses the problem where we want to apply machine learning over a small data set of multivariate time series. A challenge when classifying data is when the data set is small and overfitting is at risk. Augmentation of small data sets might avoid overfitting. The multivariate time series used in this project represent motion data of people with reconstructed ACLs and a control group. The approach was pairing motion data from the training set and using Euclidean Barycentric Averaging to create a new set of synthetic motion data so as to increase the size of the training set. The classifiers used were Dynamic Time Warping -One Nearest neighbour and Time Series Forest. In our example we found this way of increasing the training set a less productive strategy. We also found Time Series Forest to generally perform with higher accuracy on the chosen data sets, but there may be more effective augmentation strategies to avoid overfitting.

Conventional concentrating solar power (CSP) plants typically have a very high levelised cost of electricity (LCOE) compared with coal-fired power stations. To generate 1 kWh of electrical energy from a conventional linear Fresnel CSP plant without a storage application, costs the utility approximately R3,08 (Salvatore, 2014), whereas it costs R0,711 to generate the same amount of energy by means of a highly efficient supercritical coal-fired power station, taking carbon tax into consideration. This high LCOE associated with linear Fresnel CSP technology is primarily due to the massive capital investment required per kW installed to construct such a plant along with the relatively low-capacity factors, because of the uncontrollable solar irradiation. It is expected that the LCOE of a hybrid plant in which a concentrating solar thermal (CST) station is integrated with a large-scale supercritical coal-fired power station, will be higher than that of a conventional supercritical coal-fired power station, but much less than that of a conventional CSP plant. The main aim of this study is to calculate and then compare the LCOE of a conventional supercritical coal-fired power station with that of such a station integrated with a linear Fresnel CST field. When the thermal energy generated in the receiver of a CST plant is converted into electrical energy by using the highly efficient regenerative Rankine cycle of a large-scale coal-fired power station, the total capital cost of the solar side of the integrated system will be reduced significantly, compared with the two stations operating independently of one another for common steam turbines, electrical generators and transformers, and transmission lines will be utilised for the integrated plants. The results obtained from the thermodynamic models indicate that if an additional heat exchanger integration option for a 90 MW (peak thermal) fuel-saver solar-augmentation scenario, where an annual average direct normal irradiation limit of 2 141 kWh/m2 is considered, one can expect to produce approximately 4,6 GWh more electricity to the national grid annually than with a normal coal-fired station. This increase in net electricity output is mainly due to the compounded lowered auxiliary power consumption during high solar-irradiation conditions. It is also found that the total annual thermal energy input required from burning pulverised coal is reduced by 110,5 GWh, when approximately 176,5 GWh of solar energy is injected into the coal-fired power station’s regenerative Rankine cycle for the duration of a year. Of the total thermal energy supplied by the solar field, approximately 54,6 GWh is eventually converted into electrical energy. Approximately 22 kT less coal will be required, which will result in 38,7 kT less CO2 emissions and about 7,6 kT less ash production. This electricity generated from the thermal energy supplied by the solar field will produce approximately R8,188m in additional revenue annually from the trade of renewable energy certificates, while the reduced coal consumption will result in an annual fuel saving of about R6,189m. By emitting less CO2 into the atmosphere, the annual carbon tax bill will be reduced by R1,856m, and by supplying additional energy to the national grid, an additional income of approximately R3,037m will be due to the power station. The annual operating and maintenance cost increase resulting from the additional 171 000 m2 solar field, will be in the region of R9,71m. The cost of generating 1 kWh with the solar-augmented coal-fired power plant will only be 0,34 cents more expensive at R0,714/kWh than it would be to generate the same energy with a normal supercritical coal-fired power station. If one considers that a typical conventional linear Fresnel CSP plant (without storage) has an LCOE of R3,08, the conclusion can be drawn that it is much more attractive to generate electricity from thermal power supplied by a solar field, by utilising the highly efficient large-scale components of a supercritical coal-fired power station, rather than to generate electricity from a conventional linear Fresnel CSP plant. / MIng (Mechanical Engineering), North-West University, Potchefstroom Campus, 2015

Conventional concentrating solar power (CSP) plants typically have a very high levelised cost of electricity (LCOE) compared with coal-fired power stations. To generate 1 kWh of electrical energy from a conventional linear Fresnel CSP plant without a storage application, costs the utility approximately R3,08 (Salvatore, 2014), whereas it costs R0,711 to generate the same amount of energy by means of a highly efficient supercritical coal-fired power station, taking carbon tax into consideration. This high LCOE associated with linear Fresnel CSP technology is primarily due to the massive capital investment required per kW installed to construct such a plant along with the relatively low-capacity factors, because of the uncontrollable solar irradiation. It is expected that the LCOE of a hybrid plant in which a concentrating solar thermal (CST) station is integrated with a large-scale supercritical coal-fired power station, will be higher than that of a conventional supercritical coal-fired power station, but much less than that of a conventional CSP plant. The main aim of this study is to calculate and then compare the LCOE of a conventional supercritical coal-fired power station with that of such a station integrated with a linear Fresnel CST field. When the thermal energy generated in the receiver of a CST plant is converted into electrical energy by using the highly efficient regenerative Rankine cycle of a large-scale coal-fired power station, the total capital cost of the solar side of the integrated system will be reduced significantly, compared with the two stations operating independently of one another for common steam turbines, electrical generators and transformers, and transmission lines will be utilised for the integrated plants. The results obtained from the thermodynamic models indicate that if an additional heat exchanger integration option for a 90 MW (peak thermal) fuel-saver solar-augmentation scenario, where an annual average direct normal irradiation limit of 2 141 kWh/m2 is considered, one can expect to produce approximately 4,6 GWh more electricity to the national grid annually than with a normal coal-fired station. This increase in net electricity output is mainly due to the compounded lowered auxiliary power consumption during high solar-irradiation conditions. It is also found that the total annual thermal energy input required from burning pulverised coal is reduced by 110,5 GWh, when approximately 176,5 GWh of solar energy is injected into the coal-fired power station’s regenerative Rankine cycle for the duration of a year. Of the total thermal energy supplied by the solar field, approximately 54,6 GWh is eventually converted into electrical energy. Approximately 22 kT less coal will be required, which will result in 38,7 kT less CO2 emissions and about 7,6 kT less ash production. This electricity generated from the thermal energy supplied by the solar field will produce approximately R8,188m in additional revenue annually from the trade of renewable energy certificates, while the reduced coal consumption will result in an annual fuel saving of about R6,189m. By emitting less CO2 into the atmosphere, the annual carbon tax bill will be reduced by R1,856m, and by supplying additional energy to the national grid, an additional income of approximately R3,037m will be due to the power station. The annual operating and maintenance cost increase resulting from the additional 171 000 m2 solar field, will be in the region of R9,71m. The cost of generating 1 kWh with the solar-augmented coal-fired power plant will only be 0,34 cents more expensive at R0,714/kWh than it would be to generate the same energy with a normal supercritical coal-fired power station. If one considers that a typical conventional linear Fresnel CSP plant (without storage) has an LCOE of R3,08, the conclusion can be drawn that it is much more attractive to generate electricity from thermal power supplied by a solar field, by utilising the highly efficient large-scale components of a supercritical coal-fired power station, rather than to generate electricity from a conventional linear Fresnel CSP plant. / MIng (Mechanical Engineering), North-West University, Potchefstroom Campus, 2015

International Telemetering Conference Proceedings / October 23-26, 2000 / Town & Country Hotel and Conference Center, San Diego, California / White Sands Missile Range is supporting Ballistic Missile Defense Organization (BMDO) target firings from Ft. Wingate, NM. This two Off-Range Corridor allows BMDO to conduct long range testing within the continental U.S. The Transportable Range Augmentation and Control System (TRACS), consisting of a control van and one of two Mobile Telemetry Systems (MTS), provide the necessary on-site telemetry support. The Dual Remote Interferometer System (DRDAS) that tracks the telemetry RF carrier in support of Missile Flight Safety (MFS) is also included in this paper. This paper describes the telemetry support scenario in terms of preliminary simulations followed by real-time support. Real-time support consists of data distribution from the MTS to the Telemetry Distribution Center, TRACS Control van, Missile Flight Safety display van, Project Support vans, on-site data processing, as well as relaying raw data to the main WSMR Telemetry Data Center (TDC) for real-time analysis. As soon as telemetry data arrives at the TDC, it is converted into information. This information is used by MFS during real-time monitoring of vehicle performance. This paper includes the methods used for the conversion of data into information on-site and at TDC. Real-time data processing involves multiple independent systems performing their respective tasks on a particular segment of data.

Pulse wave velocity (PWV) and augmentation index (AI) are independent predictors of cardiovascular health. However, the comparability of multiple oscillometric modalities currently available for their assessment was not studied in detail. In the present study, we aimed to evaluate the relationship between indices of arterial stiffness assessed by diastolic and suprasystolic oscillometry. In total, 56 volunteers from the general population (23 males; median age 70 years [interquartile range: 65–72 years]) were recruited into observational feasibility study to evaluate the carotid-femoral/aortic PWV (cf/aoPWV), brachial-ankle PWV (baPWV), and AI assessed by 2 devices: Vicorder (VI) applying diastolic, right-sided oscillometry for the determination of all 3 indices, and Vascular explorer (VE) implementing single-point, suprasystolic brachial oscillometry (SSBO) pulse wave analysis for the assessment of cfPWV and AI. Within- and between-device correlations of measured parameters were analyzed. Furthermore, agreement of repeated measurements, intra- and inter-observer concordances were determined and compared for both devices. In VI, both baPWVand cfPWVinter-correlatedwell and showed good level of agreement with bilateral baPWVmeasured byVE (baPWV[VI]– baPWV[VE]R: overall concordance correlation coefficient [OCCC]¼0.484, mean difference¼1.94 m/s; cfPWV[VI]–baPWV[- VE]R: OCCC¼0.493, mean difference¼1.0m/s). In contrast, SSBO derived aortic PWA (cf/aoPWA[VE]) displayed only weak correlation with cfPWV(VI) (r¼0.196; P¼0.04) and ipsilateral baPWV (cf/ aoPWV[VE]R–baPWV[VE]R: r¼0.166; P¼0.08). cf/aoPWA(VE) correlated strongly with AI(VE) (right-sided: r¼0.725, P<0.001). AI exhibited marginal between-device agreement (right-sided: OCCC¼ 0.298, mean difference: 6.12%). All considered parameters showed good-to-excellent repeatability giving OCCC > 0.9 for 2-point-PWV modes and right-sided AI(VE). Intra- and inter-observer concordances were similarly high except for AI yielding a trend toward better reproducibility in VE (interobserver–OCCC[VI] vs [VE]¼0.774 vs 0.844; intraobserver OCCC[VI] vs [VE]¼0.613 vs 0.769). Both diastolic oscillometry-derived PWV modes, and AI measured either with VI or VE, are comparable and reliable alternatives for the assessment of arterial stiffness. Aortic PWV assessed by SSBO in VE is not related to the corresponding indices determined by traditional diastolic oscillometry.

This research project presents a study of the fixation of screws in augmented and non-augmented cancellous bone at a microscopic scale. It is estimated that somewhere close to one million screws are failing each year. Therefore, the aim is to identify the key parameters affecting screw pull-out in order to improve screw fixation in cancellous bone, and hence screw design. The background for this study comes from work by Stryker, comparing screw pull-out from augmented and non-augmented cancellous bone, where a few cases of screw pull-out gave better results without bone augmentation. This is contrary to most evidence and the hypothesis to explain these results is that the screw pull-out from cancellous bone could be strongly affected by the cancellous bone micro architecture. The effect of the influence of the screw’s initial position was first verified with 2D finite element (FE) models of screw pull-out from simplified cancellous bone models. The results showed a force reaction variation up to 28% with small change in position. The hypothesis was then tested with 3D FE models of screw pull-out from more complex cancellous bone models with different volume fractions. Three volume fractions were tested and again the effects were confirmed, but only in models with the lower volume fraction. A variation up to 30% of the force reaction was observed. The 3D simplified cancellous bone models with 5.3% volume fraction were also used to study the influence of augmentation using calcium phosphate cement. A significant improvement of the screw holding power (almost 2 times) as well as an important diminution of the variability of the pull-out force due to the screw initial position was found. Other augmentation geometries were used to model cement. They all showed an increase of the screw pull-out force reaction with an increase of the cement volume. Validation of FE results was achieved by comparing screw pull-out from a cadaver cancellous bone and the FE model constructed from the same bone sample. New studies were then carried out from the cadaver cancellous bone model. The first study examined the screw initial position influence with cancellous and cortical screws and again showed that there is a strong correlation between screw pull-out stiffness and bone volume fraction. The cortical screw showed improved performance over the cancellous screw. Augmentation cases were explored using three bone samples with a range of volume fractions obtained from different sites within the cadaver bone sample. The cancellous screw was tested with 3 types of augmentation and the cortical screw was tested with one augmentation in these 3 samples. The results showed each time a significant improvement of stiffness with augmentation but when compared with the effect of volume variation inside the bone sample, it appeared that the improvement of stiffness from augmentation might not cover the loss in stiffness from a small change in bone structure. Finally, screw design parameters were investigated, as cortical screws seemed to give as good or better stiffness results than cancellous screw. The thread pitch, the thread angle and the core diameter were analysed independently and it appeared that the most important parameter was the thread pitch with an improvement of the stiffness of +46% for cancellous screws with a smaller thread pitch. The two other factors studied (core diameter and thread angle) showed somewhat stiffer results but with a relatively small influence (less than 10%). From this study, the best screw for use in cancellous bone could be a cortical screw (diameter and pitch) with thread angles similar to a cancellous screw.

Durant cette thèse, nous avons introduit les problèmes de Bin Covering avec Contrainte de Distance (BCCD) et de Bin Packing avec Contrainte de Distance (BPCD), qui trouvent leur application dans les réseaux de grande échelle, tel Internet. L'étude de ces problèmes que nous effectuons dans des espaces métriques quelconques montre qu'il est impossible de travailler dans un tel cadre sans avoir recours à de l'augmentation de ressources, un procédé qui permet d'élaborer des algorithmes construisant des solutions moins contraintes que la solution optimale à laquelle elles sont comparées. En plus de résultats d'approximation intéressants, nous prouvons la difficulté de ces problèmes si ce procédé n'est pas utilisé. Par ailleurs, de nombreux outils ont pour objectif de plonger les grands réseaux qui nous intéressent dans des espaces métriques bien décrits. Nous avons alors étudié nos problèmes dans les espaces métriques générés par certains de ces outils, comme Vivaldi et Sequoia.