7903 research outputs found
Sort by
Hiv3: An Efficient Beehive Monitoring System
Get PDFBeehive monitoring plays a major role in ensuring the health of beehives by checking for overpopulation or underpopulation within a hive. Beehive monitoring provides beekeepers with the opportunity to take action and save the hive before the problem becomes irreversible. Most solutions are too expensive for everyday beekeepers and lack elements of sustainability, making it impractical for small scale beekeepers. In this thesis, we propose a solution to this problem, demonstrating its sustainability and user-friendliness, which enables us to effectively reach a larger consumer market. We support these claims through the use of sustainable systems such as using a solar panel coupled with a rechargeable battery and incorporating deep-sleep capabilities into the system’s low-power embedded system (ESP32) which is connected to a camera. The ESP32 sends images to a Raspberry Pi, which performs image processing using a machine learning model and transmits the processed images to the cloud. We present a system architecture diagram describing how these systems are integrated as well as how other measures, such as security and single sign-on, are implemented to ensure the integrity of the solution. The system tests conducted in the field show that the machine learning model yields a mean average precision (MAP) score of 52.2, compared to the benchmark score of 53.7, ensuring accurate, real-time monitoring utilizing a low-power system
MonteRecycle: An Application for Streamlining Tracking of Recyclables in Uruguay
Get PDFRecycling is a crucial process that has a positive impact on nature, cleans the environment, and saves time and money. Classifying recyclables by material and recycling these items accordingly helps make our world more environmentally-friendly and a better place to live.
In the recycling system in Montevideo, Uruguay, third-party contractors hired by San Vicente transport clients’ recyclables to a central recycling center or to a person that does recycling. In the past, San Vicente has transported recyclables though horse and buggy transportation. However, there has been a push from the city to replace the horses with motorcycles to reduce animal waste. As the first approach, San Vicente used pen and paper for tracking the type, weight, and pickup location of the received materials. Losing the paper on the field due to the nature of the job and rain, lack of a standard template for collecting the data, as well as its lack of error-handling features are some of the downsides of the pen and paper approach. Simplicity of using pen and paper, however, is considered as the pros of this approach.
To improve this process, San Vicente tried using Google Forms for collecting information and storing them into spreadsheets which could then later be used by the admins to verify the data and generate monthly reports for their clients. While this approach alleviates the challenges of losing papers, its usability is more complex for the field workers with limited literature, math, and technological skills.
To solve this problem, we made a full-stack application that tracks recycling data and reports accurate pickup totals and classification details to clients. We designed and developed MonteRecycle using React Native (for the UI), Postgres (for strong data), React JS (for the admin web application), and Google Cloud (for hosting the Postgres). Additionally, we developed the admin website to display all of the data gathered in the mobile application to the administrator of this recycling process. Our project is successfully implemented, tested, and provided as a downloadable Android APK to San Vicente to be used on the field as their main framework for entering and organizing data as well as generating monthly reports for San Vicente’s clients. Through this project, we learned that change is inevitable in the software development process and that user feedback is the most important feedback we can get to improve our application
Performance Evaluation of Databases for Packet Capture and Analysis
No full textToday, the presence of smart devices is constantly on the rise, especially for Internet of Things (IoT) devices. However, their utility-based design means that these devices are limited in computing power. Due to this limited computing power, the devices are more prone to cyber-security threats, and it is vital to construct a method to aid in network traffic analysis, bolstering defense mechanisms to thwart any malicious attacks. To analyze their network data efficiently and externally, we constructed a test-bed infrastructure to measure and evaluate the performance of databases. In this thesis, we have chosen to compare two databases which are both used widely. InfluxDB is a time-series database while DuckDB is an in-process analytical database. The comparison metrics we gathered are CPU utilization and memory utilization. We found that DuckDB’s memory utilization increased linearly whenever a batch of packets was inserted. In contrast, DuckDB’s CPU utilization results indicate spikes during data insertion, but otherwise these results show a steady CPU usage. The results obtained for InfluxDB were more stable, with CPU utilization remaining within a certain range. Likewise, InfluxDB’s memory usage remained constant throughout our testing period.
The differences in the results allude to the structural variances between the two databases. DuckDB, being an in-process database, shares resources with the interacting application. In contrast, InfluxDB, being a time series database, utilizes it’s own set of resources, distinct from those used by the interacting application. As a result, we observe that the memory utilization of DuckDB is higher when we insert data. Future work for our project breaks down into two components: changing the design of our test-bed, and improving the accuracy of measuring performance. The former would involve incorporating additional databases (i.e., PostgreSQL) and also obtaining alternative metrics for comparing them
Personal Trip Planner
No full textGenerating routes with multiple destinations considering optimal route factors as well as users’ previous preferences poses a challenge in existing navigation applications. Without specific destinations in mind, users’ often struggle to find optimal routes that accommodate different stops for each one of their needs along with their primary destination. We aim to develop user-centric navigation and planning software that can process natural language and generate multiple efficient route options tailored to individual preferences and needs. This project aims to address the aforementioned problem by developing a software that utilizes natural language processing to understand user requests for flexible route planning. The software will have functionalities such as location identification, route generation, personalized recommendation systems, and secure user data storage. We will employ technologies such as AWS SAM backend, React Native frontend, Google Maps API, and OpenAI’s GPT API. We plan to thoroughly test our software to ensure the software achieves its objectives and meets our standards
Jibu Customer Carbon Savings in Rwanda
Get PDFJibu Inc. is a social enterprise in Rwanda that addresses the essential service need for clean drinking water. The following report aims to understand Jibu’s potential for acquiring a carbon credit certification by investigating customers’ water treatment methods before becoming Jibu customers. Data was collected on Jibu customers by way of a quantitative survey that uncovered customer profiles and carbon savings. Our analysis reveals that before purchasing Jibu water, many current customers treated water at home using carbon-emitting practices such as boiling water with charcoal or gas. Jibu customers are now reducing carbon emissions since they are no longer boiling water, thus no longer emitting as much carbon into the atmosphere. Based on customers’ carbon savings, this report asserts that it is worthwhile for Jibu to pursue the carbon credit certification process further. In tandem with pursuing certification, our research confirmed that Jibu should continue to provide an affordable and accessible drinking water source for customers across Rwanda
The Entanglement of Education and Employment in an AI Economy
Get PDFArtificial intelligence (AI) has been ushering in a new wave of innovation. Yet the implementation of AI in the workplace has sparked controversy. Although robots taking over the workplace, or even society, has been a recurring trope since the industrial revolution, it resounds even more profoundly in today’s society. Much like the anxiety that emerged when factory workers’ jobs were being replaced by automation in the 1900s, today individuals\u27 loss of value and economic security has taken center stage. Their anxiety is not without reason. AI has the potential to and has replaced human positions in many fields. However, much like the automation seen in factories, AI does not function without human support or collaboration
A Step Towards Automated Ethical Analysis in Journalism: Measuring LLMs’ Performance in Extracting Sourcing Information
Get PDFThis thesis explores the potential of Large Language Models (LLMs) in automating the extraction of sourcing information from news articles, a crucial step towards enhancing transparency and ethical analysis in journalism. We evaluate the performance of two state-of-the-art LLMs, GPT-4 and Claude 3, in identifying and categorizing various source types across four diverse news articles. The thesis employs a zero-shot learning approach with two different prompt designs, assessing the models’ ability to adapt to varying source structures and prompt instructions.
Our findings reveal that while LLMs show promise in extracting sourcing information, their performance varies significantly across different article types and source structures. The research highlights the complex interplay between prompt design, source types, and model performance, with both LLMs demonstrating strengths and limitations in handling diverse journalistic contexts. This thesis contributes to the growing body of work on AI in journalism by providing initial insights into the current capabilities of LLMs in sourcing analysis and outlining key areas for future research and development in automated ethical analysis of news content
Moving Beyond Human Capital: Constructivism, Critical Pedagogy, and Interdisciplinary Collaboration in a Building and Construction Pathway
Get PDFThe silo structure and neoliberal rationality of Career Technical Education (CTE) pathways limit student learning and propagate the purpose of education as the creation of human capital, not a socially aware citizenry (Brown, 2017; Jacobs, 2010). Despite its history of racial tracking, CTE pathways still neglect critical thinking/dialogue around social/environmental justice issues, drastically hurting workers’ ability to confront these inequities (Darder, 2017; Oakes & Saunders, 2011; Shor & Freire, 1987). Using a qualitative research design that employs ethnographic elements, this study found that constructivism, critical pedagogy, and interdisciplinary collaboration can positively impact the experience, participation, and critical consciousness of students and educators in a CTE building and construction pathway; however, more research is required to fully understand how these pedagogical shifts can be more effectively integrated into diverse CTE career clusters
Blended Work/Life Relationships: Organizational Communication Involving Workplace Peers, Friends, and Lovers
Get PDFFostering Success: An Examination of the Influence of First-Generation Scholars Programs on Latinx College Students Throughout Their Educational Journey
Get PDFThis study investigates the influence of participation in first-generation scholars programs (FGSPs) on the educational journey of first-generation Latinx college students. The research question guiding this inquiry is: How does participation in FGSPs affect Latinx college students throughout their educational journey? Qualitative research methods were employed to address this question, including in-depth interviews with six Latinx first-generation college students who participated in FGSPs. Thematic analysis was utilized to identify patterns and themes within the data. The findings reveal that participation in FGSPs significantly contributes to Latinx students’ academic success, retention, and educational experiences. Key themes emerging from the data included the role of financial assistance in alleviating barriers to education, the importance of mentorship and support networks, the impact of culturally responsive programming, and the long-term benefits of program participation beyond graduation. These findings provide valuable insights into the effectiveness of FGSPs in supporting Latinx college students and highlight areas for program improvement and expansion. The study concludes with implications for practice and recommendations for future research to further enhance support for Latinx first-generation college students in higher education