Precision livestock farming technologies have been widely researched over the last decade. These technologies help in monitoring animal health and welfare parameters in a continuous, automated fashion. Under this umbrella of precision livestock farming, this study focuses on activity classification and body weight prediction in pigs. Activity monitoring is essential for understanding the health and growth of pigs. To automate this task effectively, we propose efficient and accurate sensor-based deep learning (DL) solutions. Among these, the 2D Residual Networks emerged as the best performing model, achieving an accuracy of 95.6%. This accuracy was 15.6% higher than that of other machine learning approaches. Additionally, accurate pig weight estimation is crucial for pork production, as it provides valuable insights into growth rates, disease prevalence, and overall health. Traditional manual methods of estimating pig weights are time-consuming and labor-intensive. To address this issue, we propose a novel approach that utilizes deep learning techniques on depth images for weight prediction. Through a custom image preprocessing pipeline, we train DL models to extract meaningful information from depth images for weight prediction. Our findings show that XceptionNet gives promising results, with a mean absolute error of 2.82 kg and a mean absolute percentage error of 7.42%. In comparison, the best performing statistical model, support vector machine, achieved a mean absolute error of 4.51 kg mean absolute percentage error of 15.56%. / Master of Science / With the increasing demand for food production in recent decades, the livestock farming industry faces significant pressure to modernize its methods. Traditional manual tasks such as activity monitoring and body weight measurement have been time-consuming and labor-intensive. Moreover, manual handling of animals can cause stress, negatively affecting their health. To address these challenges, this study proposes deep learning-based solutions for both activity classification and automated body weight prediction. For activity classification, our solution incorporates strategic data preprocessing techniques. Among various learning techniques, our deep learning model, the 2D Residual Networks, achieved an accuracy of 95.6%, surpassing other approaches by 15.6%. Furthermore, this study also compares statistical models with deep learning models for the body weight prediction task. Our analysis demonstrates that deep learning models outperform statistical models in terms of accuracy and inference time. Specifically, XceptionNet yielded promising results, with a mean absolute error of 2.82 kg and a mean absolute percentage error of 7.42%, outperforming the best statistical model by nearly 8%.

Most medical treatments have been developed aiming at the best-on-average efficacy for large populations, resulting in treatments successful for some patients but not for others. It necessitates the need for precision medicine that tailors medical treatment to individual patients. Omics data holds comprehensive genetic information on individual variability at the molecular level and hence the potential to be translated into personalized therapy. However, the attempts to transform omics data-driven insights into clinically actionable models for individual patients have been limited. Meanwhile, advances in deep learning, one of the most promising branches of artificial intelligence, have produced unprecedented performance in various fields. Although several deep learning-based methods have been proposed to predict individual phenotypes, they have not established the state of the practice, due to instability of selected or learned features derived from extremely high dimensional data with low sample sizes, which often results in overfitted models with high variance. To overcome the limitation of omics data, recent advances in deep learning models, including representation learning models, generative models, and interpretable models, can be considered. The goal of the proposed work is to develop deep learning models that can overcome the limitation of omics data to enhance the prediction of personalized medical decisions. To achieve this, three key challenges should be addressed: 1) effectively reducing dimensions of omics data, 2) systematically augmenting omics data, and 3) improving the interpretability of omics data. / Doctor of Philosophy / Most medical treatments have been developed aiming at the best-on-average efficacy for large populations, resulting in treatments successful for some patients but not for others. It necessitates the need for precision medicine that tailors medical treatment to individual patients. Biological data such as DNA sequences and snapshots of genetic activities hold comprehensive information on individual variability and hence the potential to accelerate personalized therapy. However, the attempts to transform data-driven insights into clinical models for individual patients have been limited. Meanwhile, advances in deep learning, one of the most promising branches of artificial intelligence, have produced unprecedented performance in various fields. Although several deep learning-based methods have been proposed to predict individual treatment or outcome, they have not established the state of the practice, due to the complexity of biological data and limited availability, which often result in overfitted models that may work on training data but not on test data or unseen data. To overcome the limitation of biological data, recent advances in deep learning models, including representation learning models, generative models, and interpretable models, can be considered. The goal of the proposed work is to develop deep learning models that can overcome the limitation of omics data to enhance the prediction of personalized medical decisions. To achieve this, three key challenges should be addressed: 1) effectively reducing the complexity of biological data, 2) generating realistic biological data, and 3) improving the interpretability of biological data.

Obtaining accurate and precise position information is critical in precision and autonomous agriculture. Systems accurate to the centimeter-level are available, but may be prohibitively expensive for relatively small farms and tasks that involve multiple vehicles. Optical resection is proposed as a potentially more cost-effective and scalable positioning system for such cases. The proposed system involves the placement of optical beacons at known locations throughout the environment and the use of cameras on the vehicle to detect the apparent angles between beacons. The position of the vehicle can be calculated with resection when three or four beacons are identified. In addition, the system provides precise orientation information, so a separate inertial measurement unit is not required. The system is seen as potentially cost-effective by taking advantage of the precision and low cost of digital image sensors. Whereas the components in other positioning systems tend to be more specialized, the widespread consumer demand for inexpensive and high quality cameras has allowed for billions of dollars of research and development to be spread across billions of image sensors. / Master of Science

Impact of precision feeding with feed management software was assessed for whole farm nutrient balance (WFNB) and feeding management from January through December 2006. Nine treatment and six control farms were selected in four regions of the Chesapeake Bay Watershed of Virginia. Herd sizes averaged 271 and 390 lactating cows for treatment and control farms while milk yield averaged 30 and 27 kg/d per lactating cow, respectively. Crop hectares grown averaged 309 and 310 ha for treatment and control farms, respectively. Treatment farms purchased and installed feed management software (TMR Tracker, Digi-Star LLC, Fort Atkinson WI) between May and October 2006 and received more frequent feed analysis and feedback. Data were collected for calendar year 2005 and 2006 to compute WFNB using software from the University of Nebraska. On treatment farms, up to five feed samples were obtained monthly from individual feedstuffs and each total mixed ration (TMR) fed to lactating cows. Control farms submitted TMR samples every 2 mo. Standard wet chemistry analysis of samples was performed. Data stored in the software were collected monthly from each treatment farm concurrent with feed sampling. Producers from each treatment farm participated in a 24-question personal interview in December 2006 addressing installation, operation, and satisfaction with the software. Daily feeding deviation of all ingredients across treatment farms averaged 173 ± 163 kg/d. This corresponded to average daily overfeeding of CP and P of 17.6 ± 17 and 0.4 ± 0.3 kg/d, respectively. Feeding deviation did not differ between feeders. Milk production was negatively associated with kg total deviation and kg CP deviation, but positively related to P deviation. Whole farm nutrient balance did not differ between treatment and control farms. All producers indicated TMR Tracker met expectations. Change made to the feeding program due to TMR Tracker was correlated (r=0.80) with perceived improvement in ration consistency. In conclusion, producers perceived feed management software as beneficial, but WFNB was not reduced after 3 to 6 mo of using feed management software; however, the large variation in daily over or under feeding indicates potential for future reductions in WFNB through reduced feeding variability. / Master of Science

<p dir="ltr">The main goal of my study was to use a novel precision medicine approach to optimize assignment to short-term abstinence from alcohol based on a variety of individual characteristics. The sample consisted of 97 moderate-to-heavy drinkers aged 21-35. A within-subjects design was employed where each participant completed two counter-balanced intravenous alcohol sessions (one following abstinence and one during usual drinking). For the primary aim of this study (N = 47), crossover generalized outcome weighted learning was used to estimate an optimal individualized assignment rule to short-term abstinence based on prescriptive factors, including HiTOP-relevant dimensions and other characteristics. For a secondary aim (N = 50), logistic regression was used to test whether the subgroups estimated by the optimal rule were associated with a set of genetic and behavioral factors related to AUD, and subjective perceptions to alcohol intoxication. Findings showed that an estimated rule with higher granularity – higher-specificity traits and demographics – led to lower alcohol consumption overall compared with one-size-fits-all rules (i.e., assigning everyone to abstinence or assigning no one to abstinence). The effect sizes of the difference were small-to-medium and fell short of statistical significance. Family history of AUD had a positive trend association with benefit from abstinence, with one standard deviation increase in family history of AUD being associated with twice as high odds of being assigned to abstinence. Due to the limited sample size, the results should be interpreted with caution. Study results provided preliminary evidence that an individualized assignment rule based on relatively simple and easily accessible individual characteristics can lead to lower alcohol consumption than that observed if everyone or no one was assigned to abstinence (i.e., one-size-fits-all approach). Genetic predispositions reflected in family history of AUD may be a potential mechanism linking the assessed prescriptive factors with abstinence response, which is worth further exploration.</p>

Effective weed control is crucial, especially for herbicide-resistant species. Machine vision technology, through weed detection and localization, can facilitate precise, species-specific treatments. Despite the challenges posed by unstructured field conditions and weed variability, deep learning (DL) algorithms show promise. This study evaluated thirteen DL-based weed detection models, including YOLOv5, RetinaNet, EfficientDet, Fast RCNN, and Faster RCNN, using pre-trained object detectors. RetinaNet (R101-FPN) achieved the highest accuracy with a mean average precision (mAP@0.50) of 79.98%, though it had longer inference times. YOLOv5n, with the fastest inference (17 ms on Google Colab) and only 1.8 million parameters, achieved a comparable 76.58% mAP@0.50, making it suitable for real-time use in resource-limited devices. A prototype using YOLOv5 was tested on two datasets, showing good real-time accuracy on In-season data and comparable results on Cross-season data, despite some accuracy challenges due to dataset distribution shifts.

Syftet med arbetet var att utvärdera mobilapplikationen KATAM avseende noggrannhet, tidsåtgång, precision och praktisk användning i jämförelse med volymuppskattning med dataklave och skördarrapport. Resultatet vid diameterjämförelsen visar på snarlika uppskattningar från KATAM respektive dataklaven. KATAM hade högre medelgrundyta, 7 % och grövre medeldiameter, 3,7 %, i jämförelse med dataklaven. KATAM hade även överskattningar av medelstammen volym i jämförelse med skördarrapporten och dataklaven, från 2,5 % till 17,6 % beroende på vilket urval av provytor och vilken programversion av KATAM som användes. Underlaget från volymuppskattningar var litet och hade felkällor vilket gjorde resultaten från mätningarna osäkra. Trots att studien visade på en överskattning av diametern talar den inbördes precisionen för att KATAM skulle kunna bli ett alternativ till Dataklaven vid uppskattning av medeldiametern. / The purpose of this essay was to evaluate the mobile application KATAM of accuracy, time, precision and practical use in comparison to volume estimation with data Digital Caliper and harvester report. The result of the diameter comparison showed similar estimates from KATAM and the Digital Caliper respectively. KATAM had a higher mean basel area, 7% and coarser mean diameter, 3.7%, compared to the Digital Caliper. KATAM also had overestimations in volume as compared to the harvesting report and the Digital Caliper concerning the mean stem, from 2.5% to 17.6%, depending on which sample areas were included and which version of KATAM was used. However, the basis of volume estimates was small and had error sources, which made the results of the measurements uncertain. Although the study shows an overestimation of the diameter, the mutual precision indicates that KATAM could be an alternative to the Digital Caliper when estimating the mean diameter.

Este trabalho tem por finalidade mostrar a aplicação e a utilização de um aeromodelo elétrico de asa fixa, também conhecido como veículo aéreo não tripulado (VANT), com controle manual ou automático, para coleta de dados e imagens em propriedades rurais, com a premissa de auxiliar os gestores no processo de gestão e tomada de decisão. A metodologia utilizada para a realização das coletas foi feita por meio de voos programados em dias e condições diferentes, para verificação e análise de desempenho do aeromodelo. Os resultados obtidos com os voos foram acima do esperado, gerando excelentes imagens e dados confiáveis. Sendo assim, pôde-se concluir que a utilização de VANTs, em coletas de dados e imagens em propriedades rurais foi satisfatória e auxiliou os gestores no processo de gerenciamento e rotacionamento de animais no pasto, uma vez que as imagens permitiram uma boa visualização e o aeromodelo desenvolvido cumpriu o seu objetivo com bom desempenho e agilidade. / This study aims to show the application and the use of a fixed-wing electric model aircraft, also known as unmanned aerial vehicle (UAV), with manual or automatic control, for collecting data and images in rural properties, with the premise of assisting managers in the management process and decision making. The methodology used to carry out the collection was made through scheduled flights on different days and conditions, for verification and performance analysis of the model aircraft. The results obtained with the flight were higher than expected, generating excellent images and reliable data. Thus, we can conclude that the use of UAV, in data and images collection in rural properties has been satisfactory and has assisted managers in the process of management and rotation of animals on pasture, since the images have allowed a preview of the terrain, with sharp images and the model aircraft has fulfilled its goal with a good performance and agility.

In this thesis we consider the Lanczos algoritm and its behaviour in finite precision. Having summarized theoretical properties of the algorithm and its connection to orthogonal polynomials, we recall the idea of the Lanczos method for approximating the matrix eigenvalues. As the behaviour of the algorithm is strongly influenced by finite precision arithmetic, the linear independence of the Lanczos vectors is usually lost after a few iterations. We use the most im- portant results from analysis of the finite precision Lanczos algorithm according to Paige, Greenbaum, Strakos and others. Based on that, we study formulation and properties of the mathematical model of finite presicion Lanczos computati- ons suggested by Greenbaum. We carry out numerical experiments in Matlab, which support the theoretical results.

Precision dairy farming technologies provide a variety of functions to dairy farmers. Little is known about dairy producer perception of these technologies. A study was performed to understand dairy producer perception of parameters monitored by precision dairy farming technologies. Calving has potential to be predicted using these same parameters and technologies. A second study was performed using two commercially marketed technologies in calving prediction. In order for these technologies to generate accurate and useful information for dairy farm use, they must accurately quantify these parameters. The final study evaluated the accuracy of five commercially marketed technologies in monitoring feeding, rumination, and lying behaviors.