22505 research outputs found
Sort by
Investigating spectral rendering techniques to improve colour matching in virtual production
Get PDFAs rendering engines become increasingly important in film and television, with their use in virtual production (VP) to display rendered imagery, some underlying issues become more apparent. This thesis aims to investigate how we can improve asset color matching of VP elements with real-life objects found on sets. Experiments were conducted in which objects were exposed to various types of lighting setups, and digital twins were rendered using traditional computer graphics techniques. The renderings occurred in both classic RGB spaces and the spectral domain. Additionally, data reduction techniques were used for the spectral renderings to determine if any of them provided advantages and improved color reproduction. The rendered images were then filmed using a cinema camera, employing virtual production techniques, alongside their real-life counterparts. The footage was then analyzed, and color difference metrics were used to determine if spectral rendering and the data reduction techniques brought advantages over RGB renderings. Additionally, the time and memory usage of the rendering methods were analyzed to determine their impact. The root mean square error between simulated and real-life spectra was also calculated to predict the quality of the renderings. The conclusion of the research is that spectral rendering offers numerous advantages, such as easier colour matching under different illuminants and higher accuracy in rendering colours of materials. However, the selection of terms and rendering methods has a significant impact on accuracy. Nevertheless, the advantages often outweigh the drawbacks, such as increased time and memory consumption by rendering algorithms, and we hope that the industry will strive to adopt spectral rendering techniques
something’s gonna give
Get PDFThis thesis examines the differences between structures of time and the relationships in which queer people orient themselves to develop an archive of a trans- body navigating the world they inhabit. Through deconstructing the notions of hegemonic time, queer identities and spaces are created, occupied, and left to be re-navigated. This study explores the fluidity of materials such as plaster, metal, and glass, in tandem with a transitioning body and identity, aiming to break down the structures of time and place to build new definitions of existence, identity, and the pursuit of a trans- future
Structure/Property/Processing of a 3D Printed Self-healing Polymer Blend based on a Thermoplastic Healing Agent
Get PDFSelf-healing polymers can regain mechanical performance following damage, offering increased material durability and sustainability. This thesis establishes structure/property/processing relationships for a 3D printable extrinsically self-healing polymer based on a UV polymerizable thermosetting resin system blended with a low temperature thermoplastic healing agent. This work serves as the first example of a vat polymerization 3D printed soft, low Tg low-melt thermoplastic extrinsically self-healing polymer blend. This development enables high resolution fabrication of complex geometries with self-healing functionality. This strategy of imbuing self-healing properties onto vat polymerization resins will enable functionality in many application spaces including aerospace, biomedical, soft robotics, coatings, and military. Successful 3D printing of this type of material was found to have several requirements. The thermoplastic healing agent must first be miscible in the liquid resin system. This enables the light to be able to penetrate through the resin to initiate polymerization. This solubilizing of polymer unfortunately results in an increase of resin viscosity which results in difficulties in 3D printing, as lower resin viscosity are required for vat polymerization techniques. As the thermoset resin undergoes polymerization, the thermoplastic phase separates and subsequently crystallizes, resulting in a two-phase system. Upon heating above the thermoplastic’s melting temperature, it can flow into damaged regions; upon cooling, it recrystallizes to bond the fractured interfaces. Throughout this process, the thermoset phase preserves the overall geometric integrity of the structure. This work explores the intersection of materials chemistry and additive manufacturing by investigating how both the concentration and molecular weight of a thermoplastic healing agent influence 3D printability, as well as the resulting mechanical and self-healing properties of the printed material. In addition, the study explores how key 3D printing process parameters, specifically, print temperature and layer height, affect resin 3D printability and final material performance. This work lays the foundation for uncovering a deeper understanding of polymer behavior, paving the way for the design of highly functional, self-healing printed materials for a wide array of advanced applications
Enhancing Waste Management and Circular Economy Practices in Higher Education Institutions in UAE: A Study on Current Policies and Strategic Framework
Get PDFThis study investigates the integration of Circular Economy (CE) principles within higher education institutions (HEIs) in UAE as a method to address waste challenges in UAE. Despite the announcement of UAE 2021-2031 Circular Economy Policy, and the nationwide continues promotion of sustainability initiatives, HEI’s lack the availability of tailored policies and procedures as well as guiding framework to address their waste challenges and adapt CE within their institutions. To assess this problem, a survey has been distributed to 45 CAA accredited public and private universities in UAE. A qualitative approach has been used to analyze the survey responses. The survey has revealed that there is a low level of awareness of CE principles among the HEI’s employees. HEI’s face the challenge of insufficient policy enforcement, as well as inadequate infrastructure and financial constrains. Although many of the participants showed that their institutional leadership provides positive support towards sustainability initiatives. After analyzing all of these, a CE Framework has been proposed for HEI’s, that would enable the HEI’s to be on the path towards circularity. The framework is a considered a starting point, and could be enhanced further with further research
Police, Prisons, Communities, and Courts: An Examination of Criminal Legal System Responses to Neurodivergence
Get PDFMental health in the United States is largely stigmatized and overlooked as a factor of crime. This capstone project critically analyzes the response of actors in the criminal legal system to neurodivergence and identifies areas for improvement.
Part One focuses on how police respond to those with mental health conditions and community-based supports that can reduce harm during these interactions. This chapter provides an overview of the prevalence of neurodivergence in the United States and how police are, or are not, trained to protect and serve this subset of the population. It explains the Memphis Model of crisis intervention and how this model has been successfully implemented. This chapter explores alternative methods of responding to mental health crises, including co-response models and resources that do not rely on law enforcement.
Part Two examines the treatment of those with neurodivergence in carceral facilities in the United States. It breaks down the biggest problems with how these individuals are cared for, explaining that they are often punished instead of treated, solitary confinement is overused, and care is not accessible. It overviews current federal policies relating to the treatment requirements for those with mental health conditions who are in federal custody. Finally, it identifies solutions to the three problems that are discussed, identifying and discussing therapeutic diversion units and therapy groups, animal-based therapy programs, and telehealth, or virtual therapy appointments.
Part Three explores mental health courts in the United States, starting with a brief history of how neurodivergence has been criminalized and perceived as unnatural. This chapter reviews the development of mental health courts and the differences between adult and juvenile mental health court. It explains various outcomes of this type of treatment court and common measures of success. This chapter also critically examines mental health court through the lenses of therapeutic jurisprudence and labeling theory. It concludes with a discussion of policy implications for mental health court that could lead to increased success and decreased stigmatization of neurodivergence.
Part Four is a content analysis of general orders for police relating to mental health in Monroe County, New York. It starts with a review of existing literature to provide a background and explanation of common outcomes of police responses to those experiencing a mental health crisis, how police make their decisions in these situations, crisis response models, and how stigmatizing labels impact all parties involved in a mental health crisis event. Methodology and results were then explained, and the section concludes with a discussion of the information that was collected and common themes that were identified.
This capstone seeks to critically analyze the criminal legal system responses to neurodivergence and determine areas of improvement. This community has a history of being marginalized and punished, when treatment would be more appropriate. By reviewing our systems and being critical of them, we can identify how they can improve to better serve our vulnerable populations
Cognitive Reserve in the Deaf/Hard of Hearing Community: An Exploration of Contributing Factors
Get PDFCognitive reserve (CR) refers to the individual differences in how people execute tasks, allowing some to cope with cognitive challenges more effectively than others. Additionally, previous studies have established a link between age-related hearing loss and cognitive decline. However, CR and cognitive aging have not been explored extensively in those who might identify as Deaf/hard of hearing (HH). There are four features potentially unique to this community that could shape one’s CR: use of hearing technology (e.g., hearing aid), bimodal bilingualism, and social connectedness and social support. One hundred fifty-nine Deaf/hard of hearing participants completed an online survey study assessing hearing technology usefulness, bilingual status, social connectedness and social support, and CR using the Cognitive Reserve Index questionnaire. Hearing technology usefulness ratings were significantly positively associated with social support. Unexpectedly, social connectedness was significantly correlated with ASL proficiency, suggesting lower social connectedness was related to greater ASL proficiency. CR was significantly correlated with social connectedness scores, such that greater CR was associated with higher social connectedness. All predictors of interest (i.e., hearing technology usefulness, bilingualism, social connectedness, social support), along with age of deafness and SES, did not significantly predict CR. Age accounted for most of the variance in CR, but there was not much change in variance once the predictors were added to the model. Although this study is the first to examine these potential contributors that may be specific to the Deaf/HH community, the study was limited in its measurement of constructs. Therefore, future studies may need to explore more comprehensive measures
Design of a Posit Based 6-Bit Configurable CNN Hardware Accelerator for Audio and Image Classification
Get PDFThe rapid development and growing interest in Artificial Intelligence (AI) have resulted in a trend of models increasing in size and complexity at an accelerating rate. These larger models have demonstrated a greater ability to accomplish increasingly sophisticated tasks compared to smaller models, furthering the trend of making models larger. However, as the models continue to improve and grow larger, this has created a new challenge in deploying them into practical applications. Current models typically run inference on the same systems where they were trained. Large data centers comprising Central Processing Units (CPUs) and Graphics Processing Units (GPUs) are well-suited for running these large models. However, many of the envisioned applications would require a significant reduction in size and power utilization to be effectively implemented. This research will investigate the potential of a new number system called Posits and its ability to optimize models through quantization. The potential will be tested by quantizing two CNN networks from 32-bit floating-point to 6-bit Posits, culminating in the design and verification of a hardware accelerator that implements both models in Posit form
Analytics of Capstone Projects: Understanding Outcomes Through People, Products, And Processes
Get PDFThe capstone project is a bridge from the university to industry. During the project, students not only learn the process of integrating previously learned engineering concepts into an actual product but also practice essential professional skills such as communication, teamwork, and project delivery. How can universities ensure effective capstone student learning and good outcomes? To answer this question, the capstone project is studied as a system comprising of products, people, and processes. People refer to the students and instructors involved. Students develop products by following development processes. Instructors guide students using instruction processes. In educational context there limited control over the students enrolling for the course, instructors available to guide the students, and products that client-sponsors want students to develop. Consequently, development and instructional processes emerge as the key levers through which the capstone system can be improved. Currently, at RIT like many other schools, a single standard development process is used to develop all types of products and a single instructional process is used to teach all students. Early experiments provided evidence of correlations between certain student characteristics and project outcomes, disparities in student experiences, and the potential effectiveness of process interventions in improving outcomes. The quality of outcomes improves when a capstone system can anticipate and address challenges. These findings motivated the central research question: Can poor student learning outcomes be predicted based on student and product characteristics? If so, what specific factors are associated with increased risk? Using machine learning, this research successfully predicts poor student learning outcomes in a capstone course based on product and student characteristics. This study adopts a supervised learning approach using Decision Trees and Random Forest to predict student learning outcomes, which are measured using the standardized rubrics. Input features include product and student characteristics available at the start of each project. Product characteristics include factors such as the type of product being developed and the type of client organization. People characteristics encompass data sets like students’ academic performance, demographic information, personality types, and instructor background. Out of eleven student learning outcomes, six were predicted with a recall of at least 0.75 and an F1-score of at least 0.60. Successfully predicted outcomes were applying engineering design to solve problems, experimentation, written communication, oral communication, teamwork, and independent learning. Risk factors systematically leading to bad outcomes include type of product, type of client organization, student personality, and instructor characteristics. These findings confirm that a single standard capstone process does not work equally well for everyone. In light of these findings, attention turned to identifying process customizations that could prevent poor outcomes. Using qualitative methods, the experiences of expert project guides were elicited. The findings revealed that the standard capstone process is already being implicitly customized. Additionally, certain products possess characteristics that demand greater design effort to resolve ambiguity and uncertainty. During the interviews, project guides emphasized the importance of explicitly confronting problems. Each guide helped student teams overcome emergent technical and non-technical contextual challenges in a distinctive way. This doctoral research advances knowledge at the intersection of integrated product development, capstone education, and applied machine learning. It provides new insights into how process customization can serve as a mechanism to improve student learning outcomes, and introduces methodological innovations that leverage existing student datasets to develop an early warning system for identifying poor outcomes. Through a mixed-method investigation, the study explores how customizing processes can help improve capstone project outcomes. This work lays a strong foundation for future studies and offers potential applications in both research and pedagogy. Ultimately, this work emphasizes the importance of evidence-based approaches in improving systems that involve both people and products, and underscores the importance of interdisciplinary perspectives in addressing the challenges of tomorrow
Online and Offline Multi-Variate Time Series Forecasting with NeuroEvolution Based Neural Architecture Search
Get PDFTime series forecasting plays a crucial role in various fields, ranging from financial markets and predictive maintenance in industrial settings to healthcare analytics. Traditionally, statistical approaches were employed primarily for univariate forecasting tasks, but modern applications often require robust solutions capable of handling complex, multivariate, noisy, and non-stationary data. Current state-of-the-art solutions predominantly utilize transformer-based models, which demand significant computational resources, limiting their practicality, especially in resource-constrained environments. This dissertation introduces novel and efficient approaches to multivariate time series forecasting, leveraging NeuroEvolution-based Neural Architecture Search (NAS) methodologies. Specifically, two robust algorithms---Evolutionary eXploration of Augmenting Memory Models (EXAMM) for offline forecasting and Online NeuroEvolution-based Neural Architecture Search (ONE-NAS) for online forecasting---were developed. EXAMM significantly advances offline forecasting capabilities through innovations including distributed island repopulation, Xavier and Kaiming weight initialization, and Lamarckian weight inheritance, achieving enhanced forecasting accuracy with lightweight recurrent neural networks (RNNs) that are substantially smaller than transformer-based architectures. Real-world validations of EXAMM, including stock market forecasting and predictive maintenance for coal-fired power plants, demonstrated its practicality and financial benefits, including a notable \$7.3 million cost reduction. ONE-NAS is introduced as the first dedicated online NAS method tailored specifically for time series forecasting. It continuously evolves RNN architectures without prior training, dynamically adjusting through mutations and crossover to mitigate data drift. Comparative experiments validate ONE-NAS’s superior performance over traditional statistical methods and other contemporary online forecasting approaches, underscoring its robust adaptability to evolving real-time data. Overall, both neuroevolution-based NAS frameworks EXAMM and ONE-NAS provide scalable, efficient, and high-performing alternatives for multivariate time series forecasting, particularly suitable for deployment in computationally limited settings
Analyzing Heat Transfer Mechanisms in Tapered Microgaps and Film Boiling With Computational Modeling Incorporating Nucleation Sites and Subcooled Fluids
No full textApplications such as nuclear reactors, data centers, and microelectronics require an understanding of the hydrodynamic and thermal behavior of liquid and vapor phases on heated surfaces. Phase change and interfacial dynamics enhance heat transfer in evaporation and condensation systems. This study simulates film evaporation, subcooled nucleate boiling, and confined boiling in tapered microgaps using customized Ansys Fluent with user-defined functions (UDFs) for sharp interface tracking and temperature-gradient-driven mass transfer. In film evaporation, the peak local heat transfer coefficient (HTC) reached 2170 W/(m²·K) at 0.5 mm film thickness. The maximum Nusselt number was 5.69 at 0.5 mm and the minimum was 1.20 at 12 mm, reflecting the thermal resistance of the vapor layer. In subcooled boiling, increasing subcooling from 1 to 5 K reduced departure diameter (from 2.3 to 1.6 mm), lengthened thermal film thinning (from 0.8 to 1.2 mm), and increased HTC from 90,000 to 115,000 W/(m²·K). Tapered microgap evaporation revealed self-propelling vapor structures that rewet the surface, suppress the thermal layer, and raise local HTC above 30,000 W/(m²·K) within 1 mm of the interface. Interface shape and motion agreed with high-speed imaging results. In microgravity, reduced buoyancy promotes vapor film buildup and degrades heat transfer. Simulations with modified surfaces and condensation effects, along with tapered microgaps, generated continuous self-propelled bubble departure, disturbing the thermal layer and enhancing mixing. Peak HTC in microgravity exceeded 30,000 W/(m²·K) for single and multiple nucleating bubbles. This work advances understanding of interfacial transport, bubble dynamics, and passive vapor removal, offering design guidance for next-generation cooling systems in terrestrial electronics and spacecraft