Creative Components
Permanent URI for this collection
Browse
Search Collection
1 - 5 of 1835
-
Creative componentCoralAI: A RAG model to answer Coral-related queries( 2025-05)The expansive volume of coral reef literature presents significant challenges for researchers seeking timely and accurate answers to domain-specific questions. Traditional reliance on general-purpose large language models (LLMs) often leads to incomplete or inaccurate responses, which can undermine scientific rigor. In response, we introduce CoralAI, a Retrieval-Augmented Generation (RAG) model designed specifically for coral reef research. CoralAI integrates a carefully curated database of high-quality research papers on coral ecosystems, enabling it to generate precise and contextually accurate answers to user queries that are grounded in verified scientific literature. The model begins by splitting research papers into smaller text chunks. Through a retrieval mechanism, the model then selects and ranks text chunks most relevant to each query, ensuring high degrees of relevance and similarity. These selected chunks are then synthesized into concise summaries of relevant chunks based on the given query, accompanied by citations that allow researchers to trace information back to the original sources. CoralAI is evaluated on metrics such as Context Recall, Context Precision, Faithfulness, Answer Relevance, and Answer Semantic Similarity to demonstrate its performance. We evaluate these metrics using queries generated both by humans and LLMs. Additionally, a user-friendly interface has been developed to provide seamless access to the model, allowing researchers to benefit from its insights without needing to engage with complex backend processes.
-
Creative componentPredicting Cross Architecture Performance of Source Codes using Graph Neural Networks( 2024-12)In this creative component, we explore the application of Graph Neural Networks (GNNs) to predict the cross-architecture performance of source code, focusing specifically on intermediate representation-based features. The goal of this work is to identify patterns that can help optimize the performance of programs when executed on different architectures, such as CPUs and GPUs. By leveraging GNNs, we aim to capture the relationships and structures inherent in source code, which are often challenging for traditional performance prediction models. Our experiments show promising results in predicting performance variations and highlight potential opportunities for program parallelization based on GNN-derived insights.
-
Creative componentInfrared Deep Learning System for Non-Invasive Drunk Detection and Alcohol Level Classification( 2024-12)This project proposes a novel approach to alcohol detection using an infrared (IR) camera integrated with deep learning technology. Designed to address delays and contamination issues associated with conventional methods such as breath analyzers, the system utilizes an FLIR ONE thermal camera to capture both infrared and visual images, which are processed via a deep learning-based classification system. The system operates on an iPhone, offering portability and convenience. The dataset comprises images from 50 volunteers, each contributing data across four levels of alcohol consumption: sober, one glass, two glasses, and three glasses. Data augmentation techniques expanded the dataset to over 1,200 images, improving the robustness and generalization of the classification models. Two classification tasks were performed: binary classification (sober vs. drunk) and four-level classification (sober, 1 glass, 2 glasses, 3 glasses). The NasNetMobile model achieved the highest accuracy for the four-level classification task at 85.10%, while MobileNet performed best in the binary classification task, achieving an accuracy of 74.07%. The use of Grad-CAM interpretability methods demonstrated that the models focused effectively on relevant features, such as facial regions affected by alcohol consumption. Despite these promising results, challenges such as distinguishing minor differences between consumption levels and accounting for external factors like body temperature variations remain. This research highlights the potential for non-invasive, real-time alcohol detection systems and their broader implications for public safety and health monitoring. Future work will focus on addressing limitations and improving system scalability for real-world applications.
-
Creative componentInfrared Deep Learning System for Non-Invasive Drunk Detection and Alcohol Level Classification( 2024-12)This project proposes a novel approach to alcohol detection using an infrared (IR) camera integrated with deep learning technology. Designed to address delays and contamination issues associated with conventional methods such as breath analyzers, the system utilizes an FLIR ONE thermal camera to capture both infrared and visual images, which are processed via a deep learning-based classification system. The system operates on an iPhone, offering portability and convenience. The dataset comprises images from 50 volunteers, each contributing data across four levels of alcohol consumption: sober, one glass, two glasses, and three glasses. Data augmentation techniques expanded the dataset to over 1,200 images, improving the robustness and generalization of the classification models. Two classification tasks were performed: binary classification (sober vs. drunk) and four-level classification (sober, 1 glass, 2 glasses, 3 glasses). The NasNetMobile model achieved the highest accuracy for the four-level classification task at 85.10%, while MobileNet performed best in the binary classification task, achieving an accuracy of 74.07%. The use of Grad-CAM interpretability methods demonstrated that the models focused effectively on relevant features, such as facial regions affected by alcohol consumption. Despite these promising results, challenges such as distinguishing minor differences between consumption levels and accounting for external factors like body temperature variations remain. This research highlights the potential for non-invasive, real-time alcohol detection systems and their broader implications for public safety and health monitoring. Future work will focus on addressing limitations and improving system scalability for real-world applications.
-
Creative componentPCB Design for IC Testing & Lab Measurement: J00267 Voltage Controlled Current Sources’ Current Mirrors( 2024-12)Throughout my Masters Program, I worked on a couple projects for my creative component. This Report will summarize my work and the result obtained on the J00267 Lab Measurement. The J00267 is a Chip containing Integrated Circuit (IC) design by Dr Degang Chen’s PhD student. This report will contain the general steps used for designing a PCB for IC testing. It will focus on the design I made to test the Voltage Controlled Current Source’s Current Mirrors designed by Michael Sekyere. The details of the Circuits contained in the Chip will not be discussed here. After that, I will discuss my work on the preliminary lab measurement obtained using a breadboard.