<div>Following a natural disaster, one of the most important facet that influence a persons chances of survival/being found out is the time with which they are rescued. Traditional means of search operations involving dogs, ground robots, humanitarian intervention; are time intensive and can be a major bottleneck in search operations. The main aim of these operations is to rescue victims without critical delay in the shortest time possible which can be realized in real-time by using UAVs. With advancements in computational devices and the ability to learn from complex data, deep learning can be leveraged in real time environment for purpose of search and rescue operations. This research aims to solve the traditional means of search operation using the concept of deep learning for real time object detection and Photogrammetry for precise geo-location mapping of the objects(person,car) in real time. In order to do so, various pre-trained algorithms like Mask-RCNN, SSD300, YOLOv3 and trained algorithms like YOLOv3 have been deployed with their results compared with means of addressing the search operation in</div><div>real time.</div><div><br></div>

The recent boom of Machine Learning Network Architectures like Generative Adversarial Networks (GAN), Deep Convolution Generative Adversarial Networks (DCGAN), Self Attention Generative Adversarial Networks (SAGAN), Context Conditional Generative Adversarial Networks (CCGAN) and the development of high-performance computing for big data analysis has the potential to be highly beneficial in many domains and fittingly in the early detection of chronic diseases. The clinical heterogeneity of one such chronic auto-immune disease like Systemic Lupus Erythematosus (SLE), also known as Lupus, makes it difficult for medical diagnostics. One major concern is a limited dataset that is available for diagnostics. In this research, we demonstrate the application of Generative Adversarial Networks for data augmentation and improving the error rates of Convolution Neural Networks (CNN). Limited Lupus dataset of 30 typical ’butterfly rash’ images is used as a model to decrease the error rates of a widely accepted CNN architecture like Le-Net. For the Lupus dataset, it can be seen that there is a 73.22% decrease in the error rates of Le-Net. Therefore such an approach can be extended to most recent Neural Network classifiers like ResNet. Additionally, a human perceptual study reveals that the artificial images generated from CCGAN are preferred to closely resemble real Lupus images over the artificial images generated from SAGAN and DCGAN by 45 Amazon MTurk participants. These participants are identified as ’healthcare professionals’ in the Amazon MTurk platform. This research aims to help reduce the time in detection and treatment of Lupus which usually takes 6 to 9 months from its onset.

Despite the alarming incidence rate and potential for debilitating outcomes of sports-related concussion, the underlying mechanisms of injury remain to be expounded. Since as early as 1950, researchers have aimed to characterize head impact biomechanics through in-lab and in-game investigations. The ever-growing body of literature within this area has supported the inherent connection between head kinematics during impact and injury outcomes. Even so, traditional metrics of peak acceleration, time window, and HIC have outlived their potential. More sophisticated analysis techniques are required to advance the understanding of concussive vs subconcussive impacts. The work presented within this thesis was motivated by the exploration of advanced approaches to 1) experimental theory and design of impact reconstructions and 2) characterization of kinematic profiles for model building. These two areas of investigation resulted in the presentation of refined, systematic approaches to head impact analysis that should begin to replace outdated standards and metrics.

Since the use of social media has evolved, the ability of its users to bully others has increased. One of the prevalent forms of bullying is Cyberbullying, which occurs on the social media sites such as Facebook©, WhatsApp©, and Twitter©. The past decade has witnessed a growth in cyberbullying – is a form of bullying that occurs virtually by the use of electronic devices, such as messaging, e-mail, online gaming, social media, or through images or mails sent to a mobile. This bullying is not only limited to English language and occurs in other languages. Hence, it is of the utmost importance to detect cyberbullying in multiple languages. Since current approaches to identify cyberbullying are mostly focused on English language texts, this thesis proposes a new approach (called Multilingual Cyberbullying Detection System) for the detection of cyberbullying in multiple languages (English, Hindi, and Marathi). It uses two techniques, namely, Machine Learning-based and Lexicon-based, to classify the input data as bullying or non-bullying. The aim of this research is to not only detect cyberbullying but also provide a distributed infrastructure to detect bullying. We have developed multiple prototypes (standalone, collaborative, and cloud-based) and carried out experiments with them to detect cyberbullying on different datasets from multiple languages. The outcomes of our experiments show that the machine-learning model outperforms the lexicon-based model in all the languages. In addition, the results of our experiments show that collaboration techniques can help to improve the accuracy of a poor-performing node in the system. Finally, we show that the cloud-based configurations performed better than the local configurations.

<p>Powder bed fusion (PBF) process is a metal additive manufacturing process, which can build parts with any complexity from a wide range of metallic materials. Research in the PBF process predominantly focuses on the impact of a few parameters on the ultimate properties of the printed part. The lack of a systematic approach to optimizing the process parameters for a better performance of given material results in a sub-optimal process limiting the potentialof the application. This process needs a comprehensive study of all the influential parameters and their impact on the mechanical and microstructural properties of a fabricated part. Furthermore, there is a need to develop a quantitative system for mapping the material properties and process parameters with the ultimate quality of the fabricated part to achieve improvement in the manufacturing cycle as well as the quality of the final part produced by the PBF process. To address the aforementioned challenges, this research proposes a framework to optimize the process for 316L stainless steel material. This framework characterizes the influence of process parameters on the microstructure and mechanical properties of the fabricated part using a series of experiments. These experiments study the significance of process parameters and their variance as well as study the microstructure and mechanical properties of fabricated parts by conducting tensile, impact, hardness, surface roughness, and densification tests, and ultimately obtain the optimum range of parameters. This would result in a more complete understanding of the correlation between process parameters and part quality. Furthermore, the data acquired from the experimentsare employed to develop an intelligent parameter suggestion multi-layer feedforward (FF) backpropagation (BP) artificial neural network (ANN). This network estimates the fabrication time and suggests the parameter setting accordingly to the user/manufacturers desired characteristics of the end-product. Further, research is in progress to evaluate the framework for assemblies and complex part designs and incorporate the results in the network for achieving process repeatability and consistency.</p><br>

Au XXIe siècle, la prévision de l'aléa hydrométéorologique et des impacts associés aux crues rapides demeurent un défi pour les prévisionnistes et les services de secours. Les mesures structurelles et / ou les avancées des systèmes de prévision hydrologique ne garantissent pas, à elles seules, la réduction des décès lors de ces phénomènes d'inondation rapide. La littérature souligne la nécessité d'intégrer d'autres facteurs, liés aux processus de vulnérabilité sociaux et comportementaux, afin de mieux prendre en compte les risques encourus par les populations lors de ces épisodes extrêmes. Cette dissertation conduit une analyse théorique couplés à ceux de une analyse des accidents historiques mortels afin d'expliquer les interactions qui existent entre les processus hydrométéorologiques et sociaux responsables de l'apparition de vulnérabilités humaines lors de crues rapides aux États-Unis. Des données d'enquêtes liées aux crues rapides sont examinées afin d'élaborer un système de classification des circonstances du décès (en voiture, à l'extérieur, à proximité d'un cours d'eau, dans un camping, dans un bâtiment ou en mobile-home). L'objectif est d'établir un lien entre la conception des vulnérabilités et l'estimation des pertes humaines liées à ces catastrophes naturelles. "Random forest" est utilisé et est basé sur un arbre de décision, qui permet d'évaluer la probabilité d'occurrence de décès pour une circonstance donnée en fonction d'indicateurs spatio-temporels. Un système de prévision des décès liés à l'usage de la voiture lors des crues rapides, circonstance la plus répandue, est donc proposé en s'appuyant sur les indicateurs initialement identifiés lors de l'étude théorique. Les résultats confirment que la vulnérabilité humaine et le risque associé varient de façon dynamique et infra journalière, et en fonction de la résonance spatio-temporelle entre la dynamique sociale et la dynamique d'exposition aux dangers. Par exemple, on constate que les jeunes et les personnes d'âge moyen sont plus susceptibles de se retrouver pris au piège des crues rapides particulièrement soudaines(par exemple, une durée de près de 5 heures) pendant les horaires de travail ou de loisirs en extérieur. Les personnes âgées sont quant à elles plus susceptibles de périr à l'intérieur des bâtiments, lors d'inondations plus longues, et surtout pendant la nuit lorsque les opérations de sauvetage et / ou d'évacuation sont rendues difficiles. Ces résultats mettent en évidence l'importance d'examiner la situation d'exposition aux risques en tenant compte de la vulnérabilité dynamique, plutôt que de se concentrer sur les conceptualisations génériques et statiques. Ce concept de vulnérabilité dynamique est l'objectif de modélisation développée dans cette thèse pour des vulnérabilités liés aux véhicules. À partir de l'étude de cas sur les crues rapides survenues en mai 2015, et en analysant principalement les états du Texas et de l'Oklahoma, principaux états infectés par ces évènements,le modèle montre des résultats prometteurs en termes d'identification spatio-temporelle des circonstances dangereuses. Cependant, des seuils critiques pour la prédiction des incidents liés aux véhicules doivent être étudiés plus en profondeur en intégrant des sensibilités locales non encore résolues par le modèle. Le modèle établi peut être appliqué, à une résolution journalière ou horaire, pour chaque comté du continent américain. Nous envisageons cette approche comme une première étape afin de fournir un système de prévision des crues rapides et des risques associés sur le continent américain. Il est important que la communauté scientifique spécialisée dans l'étude des crues éclairs récoltent des données à plus haute résolution lorsque ces épisodes entrainement des risques mortels, et ce afin d'appuyer la modélisation des complexités temporelles et spatiales associées aux pertes humaines causées par les futures inondations soudaines. / In the 21st century the prediction of and subsequent response to impacts due to sudden onset and localized flash flooding events remain a challenge for forecasters and emergency managers. Structural measures and/or advances in hydrological forecasting systems alone do not guarantee reduction of fatalities during short-fuse flood events. The literature highlights the need for the integration of additional factors related to social and behavioral vulnerability processes to better capture risk of people during flash floods. This dissertation conducts a theoretical analysis as well as an analysis of flash flood-specific historic fatalities to explain complex and dynamic interactions between hydrometeorological, spatial and social processes responsible for the occurrence of human life-threatening situations during the "event" phase of flash floods in the United States (U.S.). Individual-by-individual fatality records are examined in order to develop a classification system of circumstances (i.e., vehicle-related, outside/close to streams, campsite, permanent buildings, and mobile homes). The ultimate goal is to link human vulnerability conceptualizations with realistic forecasts of prominent human losses from flash flood hazards. Random forest, a well-known decision-tree based ensemble machine learning algorithm for classification is adopted to assess the likelihood of fatality occurrence for a given circumstance as a function of representative indicators at the county-level and daily or hourly time steps. Starting from the most prevalent circumstance of fatalities raised from both the literature review and the impact-based analysis, flash flood events with lethal vehicle-related accidents are the subject to predict. The findings confirm that human vulnerability and the subsequent risk to flash flooding, vary dynamically depending on the space-time resonance between that social and hazard dynamics. For example, it is found that younger and middle-aged people are more probable to get trapped from very fast flash floods (e.g., duration close to 5 hours) while participating in daytime outdoor activities (e.g., vehicle-related, recreational). In contrary, older people are more likely to perish from longer flooding inside buildings, and especially in twilight and darkness hours when rescue and/or evacuation operations are hindered. This reasoning places the importance of situational examination of dynamic vulnerability over generic and static conceptualizations, and guides the development of flash flood-specific modeling of vehicle-related human risk in this thesis. Based on the case study of May 2015 flash floods with a focus in Texas and Oklahoma, the model shows promising results in terms of identifying dangerous circumstances in space and time. Though, critical thresholds for the prediction of vehicle-related incidents need to be further investigated integrating local sensitivities, not yet captured by the model. The developed model can be applied on a daily or hourly basis for every U.S. county. We vision this approach as a first effort to provide a prediction system to support emergency preparedness and response to flash flood disasters over the conterminous U.S. It is recommended that the flash flood disaster science community and practitioners conduct data collection with more details for the life-threatening scene, and at finer resolutions to support modeling of local temporal and spatial complexities associated with human losses from flash flooding in the future.

<p>  </p> <p>Hyperspectral Imaging is one of the most popular technologies in plant phenotyping due to its ability to predict the plant physiological features such as yield biomass, leaf moisture, and nitrogen content accurately, non-destructively, and efficiently. Various kinds of hyperspectral imaging systems have been developed in the past years for both greenhouse and field phenotyping activities. Developing the plant physiological prediction model such as relative water content (RWC) using hyperspectral imaging data requires the adoption of machine learning-based calibration techniques. Convolutional neural networks (CNNs) have been known to automatically extract the features from the raw data which can lead to highly accurate physiological prediction models. Once a reliable prediction model has been developed, sharing that model across multiple hyperspectral imaging systems is very desirable since collecting the large number of ground truth labels for predictive model development is expensive and tedious. However, there are always significant differences in imaging sensors, imaging, and environmental conditions between different hyperspectral imaging facilities, which makes it difficult to share plant features prediction models. Calibration transfer between the imaging systems is critically important. In this thesis, two approaches were taken to address the calibration transfer from the greenhouse to the field. First, direct standardization (DS), piecewise direct standardization (PDS), double window piecewise direct standardization (DPDS) and spectral space transfer (SST) were used for standardizing the spectral reflectance to minimize the artifacts and spectral differences between different greenhouse imaging systems. A linear transformation matrix estimated using SST based on a small set of plant samples imaged in two facilities reduced the root mean square error (RMSE) for maize physiological feature prediction significantly, i.e., from 10.64% to 2.42% for RWC and from 1.84% to 0.11% for nitrogen content. Second, common latent space features between two greenhouses or a greenhouse and field imaging system were extracted in an unsupervised fashion. Two different models based on deep adversarial domain adaptation are trained, evaluated, and tested. In contrast to linear standardization approaches developed using the same plant samples imaged in two greenhouse facilities, the domain adaptation extracted non-linear features common between spectra of different imaging systems. Results showed that transferred RWC models reduced the RMSE by up to 45.9% for the greenhouse calibration transfer and 12.4% for a greenhouse to field transfer. The plot scale evaluation of the transferred RWC model showed no significant difference between the measurements and predictions. The methods developed and reported in this study can be used to recover the performance plummeted due to the spectral differences caused by the new phenotyping system and to share the knowledge among plant phenotyping researchers and scientists.</p>