Computer Engineering Commons^™

Temporal Dilation In Video Resnet For Sign Language Translation, Xiaoqian Yang

Master's Projects

Sign languages, vital for communication among the deaf and hard-of-hearing (DHH) people, face a significant linguistic diversity challenge with over 200 distinct sign languages worldwide. Bridging this communication gap is a priority. Traditional tools like interpreters and costly translation devices have limitations. This project aims to use deep learning techniques to develop a model capable of recognizing sign language from short videos. Our model not only recognizes the sign from a single video clip, but is also capable of making prediction of consecutive pairs of signs. To achieve zero-short gesture sequence recognition, we propose a novel temporal dilation strategy, converting …

Gesture Recognition With Deep Learning, Chaz Chang

Master's Projects

Gesture recognition is a machine learning and computer vision application where gestures are detected from videos. This project uses pose estimation to find the coordinates of important joints as a preprocessing step before trying to classify the gesture. Machine learning layers such as Convolutional Neural Network and Long Short-Term Memory are used. Various types of machine learning models are trained. The accuracy and f1 score of each model are compared. Feature selection is done by testing with different subsets of features. The results show that pose estimation as a preprocessing step provides good accuracy for gesture recognition. The results also …

Graph Based System For Evidential Reasoning, Divyarajsinh Chauhan

Master's Projects

In the modern data driven world, graph editing tools have become very essential as they provide means to understand, visualize and manipulate complex relationships between various datasets. They have especially played a crucial role in the space of evidential reasoning, where it has made a significant impact in the decision making process by developers, analysts and researchers to understand and represent the connection in the data. Existing tools fail to handle huge amounts of data efficiently and also don’t have the features required to handle tasks related to evidential reasoning.To address these gaps, we developed Pygrapher Web UI tool. We …

Estimating Air Pollution Levels Using Machine Learning, Srujay Rao Devaraneni

Master's Projects

Air pollution has emerged as a substantial concern, especially in developing countries worldwide. An important aspect of this issue is the presence of PM2.5. Air pollutants with a diameter of 2.5 or less micrometers are known as PM2.5. Due to their size, these particles are a serious health risk and can quickly infiltrate the lungs, leading to a variety of health problems. Due to growing concerns about air pollution, technology like automatic air quality measurement can offer beneficial assistance for both personal and business decisions. This research suggests an ensemble machine learning model that can efficiently replace the standard air …

Xai-Driven Cnn For Diabetic Retinopathy Detection, Vikas Shenoy Pete

Master's Projects

Diabetes, a chronic metabolic disorder, poses a significant health threat with potentially severe consequences, including diabetic retinopathy, a leading cause of blindness. In this project, we tackle this threat by developing a Convolutional Neural Network (CNN) to support the diagnosis based on eye images. The aim is early detection and intervention to mitigate the effects of diabetes on eye health. To enhance transparency and interpretability, we incorporate explainable AI techniques. This research not only contributes to the early diagnosis of diabetic eye disease but also advances our understanding of how deep learning models arrive at their decisions, fostering trust and …

Image-Based Malware Classification On Noise Extraction, Venkata Sai Sathwik Nadella

Master's Projects

Any malicious software designed to cause harm or damage to a computer system can be termed as malware. One common form of malware is as executable files. Such files are often used as a delivery mechanism for malware since they can be easily disguised as legitimate software and can be executed without raising suspicion. They are often used to exploit vulnerabilities in software, allowing malware to bypass security measures and gain access to sensitive information.

There are several methods used to detect malware in executable files, including Signaturebased detection, Behavioral-based detection, Heuristic-based detection, Sandboxing, Machine Learning and Artificial Intelligence (AI). …

Blockchain Games: What On And Off-Chain Factors Affect The Volatility, Returns, And Liquidity Of Gaming Crypto Tokens, Sumer Sareen

CMC Senior Theses

Blockchain games took the internet by storm as they offered a new way for users to play video games, own the assets in those games, and benefit monetarily from their efforts. Through Non-Fungible Tokens (NFTs) and cryptocurrencies, new, Web3 games ushered in a unique asset class for retail and institutional investors to diversify into and benefit from. This paper uses cross-sectional data from 30 blockchain gaming companies to identify on and off-chain factors that affect the company’s token volatility, returns, and liquidity. A multiple linear regression found the percentage of tokens dedicated to a company’s private sale and rewarding users, …

Evaluation Of Different Machine Learning, Deep Learning And Text Processing Techniques For Hate Speech Detection, Nabil Shawkat

Graduate Theses/Dissertations

Social media has become a domain that involves a lot of hate speech. Some users feel entitled to engage in abusive conversations by sending abusive messages, tweets, or photos to other users. It is critical to detect hate speech and prevent innocent users from becoming victims. In this study, I explore the effectiveness and performance of various machine learning methods employing text processing techniques to create a robust system for hate speech identification. I assess the performance of Naïve Bayes, Support Vector Machines, Decision Trees, Random Forests, Logistic Regression, and K Nearest Neighbors using three distinct datasets sourced from social …

Incorporating Novel Sensors For Reading Human Health State And Motion Intent Into Real-Time Computing Systems, Adam Sawyer

Masters Theses

"Integrating sensors that read states of the human body into everyday life is an increasing desire, especially with the rise of deep learning which requires vast stores of data to make predictions. This work explores integrating these sensors into the human experience through two methods and recording the results. The first of these methods integrates a MXene based field-effect transistor sensor for the 2019-nCov spike protein with a mobile app. This allows the user to read how saturated their breath is with Covid-19. The second method integrates 3D-printed pressure sensors, and a motion capture system, into a glove to read …

A Hybrid Framework For Critical Infrastructures Interdependency Modeling, Simulation, And Analysis, David Corder Hinton

Masters Theses

"Flow system models, also known as flow network models, encompass vastly complex, ever-expanding problem sets which comprise the foundation for maintenance, operation, and improvement of critical infrastructures around the world. The stable operation of these vast critical infrastructures is fundamental to the continued advancement of modern society. These infrastructures are tightly interdependent and vulnerable to interruption by both natural circumstance and malicious targeting. This necessitates representation of such critical infrastructures and their multi-domain interdependencies in defense focused constructive and virtual simulation environments as a matter of national interest and security. By breadth exploration of the problem space, this work body …

Optimizing The Placement Of Multiple Uav--Lidar Units Under Road Priority And Resolution Requirements, Zachary Michael Osterwisch

Masters Theses

"Real-time road traffic information is crucial for intelligent transportation systems (ITS) applications, like traffic navigation or emergency response management, but acquiring such data is tremendously challenging in practice because of the high costs and inefficient placement of sensors. Some modern ITS applications contribute to this problem by equipping vehicles with multiple light detection and ranging (LiDAR) sensors, which are expensive and gather data inefficiently; one solution that avoids vehicle-mounted LiDAR acquisition has been to install elevated LiDAR instruments along roadways, but this approach remains unrefined. The eventual development of sixth-generation (6G) wireless communication will enable new, creative solutions to solve …

การตรวจจับมะม่วงและนับผลผลิตด้วยการเรียนรู้เชิงลึก, อภิศักดิ์ จันทร์ดา

Chulalongkorn University Theses and Dissertations (Chula ETD)

มะม่วงเป็นสินค้าเกษตรส่งออกที่มีความสำคัญทางเศรษฐกิจของประเทศไทย จึงมีความต้องการคาดการณ์ผลผลิตเพื่อวางแผนขั้นตอนการเก็บเกี่ยวอย่างมีประสิทธิภาพ งานวิจัยนี้ ได้นำเสนอการตรวจจับผลมะม่วงบนลำต้นด้วยโมเดลการเรียนรู้เชิงลึก Mask R-CNN โดยใช้โครงข่ายแบบ RestNext-101-32x8d สำหรับสกัดคุณลักษณะ ทั้งได้พัฒนาการเพิ่มความหลากหลายของข้อมูลในการเรียนรู้โมเดลการลบวัตถุด้วยการบดบังบางส่วน Random Erasing with Partial Occlusion เพื่อเพิ่มประสิทธิภาพแก้ไขปัญหาการบดบังของวัตถุ โดยมีผลลัพธ์การตรวจจับ [email protected] ที่ 0.879, Precision ที่ 0.944, Recall ที่ 0.790 และ F1-score ที่ 0.860 โดยมีค่า Precision, Recall, F1-score เพิ่มขึ้น 0.022, 0.004 และ 0.011 ตามลำดับ เนื่องจากโมเดลตรวจจับไม่สามารถที่จะคาดการณ์ผลผลิตได้ทั้งหมด พบว่าผลมะม่วงไม่ได้แสดงออกภายนอกต้นทั้งหมด มีจำนวนผลมะม่วงที่อยู่ภายในลำต้นจำนวนมาก งานวิจัยนี้ได้นำเสนอ การคาดการณ์ผลผลิตที่ได้ผลลัพธ์จากการตรวจจับผลมะม่วง ด้วยโมเดลการถดถอย ซึ่งวิเคราะห์การถดถอยของข้อมูลที่บันทึกได้จากการนับจริงบนลำต้น เทียบผลลัพธ์การนับจากโมเดลการตรวจจับ จากผลการทดลองดีที่สุด ผลลัพธ์การนับจากโมเดลตรวจจับ ของภาพถ่ายลำต้นมะม่วงในระดับสายตา ลักษณะเห็นทั้งต้นแบบ 2 ด้าน โดยการใช้ฟังก์ชันการถดถอยแบบพหุนามดีกรีสี่ และการประมาณค่าพารามิเตอร์แบบแกร่งด้วยฟังก์ชัน Huber ได้ผลลัพธ์ค่าเฉลี่ยของค่าสัมบูรณ์เปอร์เซ็นต์ความคลาดเคลื่อน (MAPE) ที่ 25.04% ต่อต้น ค่าเฉลี่ยผลต่างสัมบูรณ์ (MAE) ที่ 5.52 ผลต่อต้น จากการพัฒนา สามารถนำระบบปรับใช้งานกับสถานการณ์จริง เพื่อคาดการณ์ผลผลิตมะม่วงบนลำต้นได้

Accelerating Precision Station Keeping For Automated Aircraft, James D. Anderson

Browse all Theses and Dissertations

Automated vehicles pose challenges in various research domains, including robotics, machine learning, computer vision, public safety, system certification, and beyond. These vehicles autonomously handle navigation and locomotion, often requiring minimal user interaction, and can operate on land, in water, or in the air. In the context of aircraft, one specific application is Automated Aerial Refueling (AAR). Traditional aerial refueling involves a "tanker" aircraft using a mechanism, such as a rigid boom arm or a flexible hose, to transfer fuel to another aircraft designated as the "receiver". For AAR, the boom arm may be maneuvered automatically, or in certain instances the …

Leveraging Signal Transfer Characteristics And Parasitics Of Spintronic Circuits For Area And Energy-Optimized Hybrid Digital And Analog Arithmetic, Adrian Tatulian

Electronic Theses and Dissertations, 2020-2023

While Internet of Things (IoT) sensors offer numerous benefits in diverse applications, they are limited by stringent constraints in energy, processing area and memory. These constraints are especially challenging within applications such as Compressive Sensing (CS) and Machine Learning (ML) via Deep Neural Networks (DNNs), which require dot product computations on large data sets. A solution to these challenges has been offered by the development of crossbar array architectures, enabled by recent advances in spintronic devices such as Magnetic Tunnel Junctions (MTJs). Crossbar arrays offer a compact, low-energy and in-memory approach to dot product computation in the analog domain by …

Parallel Real Time Rrt*: An Rrt* Based Path Planning Process, David Yackzan

Theses and Dissertations--Mechanical and Aerospace Engineering

This thesis presents a new parallelized real-time path planning process. This process is an extension of the Real-Time Rapidly Exploring Random Trees* (RT-RRT*) algorithm developed by Naderi et al in 2015 [1]. The RT-RRT* algorithm was demonstrated on a simulated two-dimensional dynamic environment while finding paths to a varying target state. We demonstrate that the original algorithm is incapable of running at a sufficient rate for control of a 7-degree-of-freedom (7-DoF) robotic arm while maintaining a path planning tree in 7 dimensions. This limitation is due to the complexity of maintaining a tree in a high-dimensional space and the network …

Learned Compressive Representations For Single-Photon 3d Imaging, Felipe Gutierrez-Barragan, Fangzhou Mu, Andrei Ardelean, Atul Ingle, Claudio Bruschini, Edoardo Charbon, Yin Li, Mohit Gupta, Andreas Velten

Computer Science Faculty Publications and Presentations

Single-photon 3D cameras can record the time-of-arrival of billions of photons per second with picosecond accuracy. One common approach to summarize the photon data stream is to build a per-pixel timestamp histogram, resulting in a 3D histogram tensor that encodes distances along the time axis. As the spatio-temporal resolution of the histogram tensor increases, the in-pixel memory requirements and output data rates can quickly become impractical. To overcome this limitation, we propose a family of linear compressive representations of histogram tensors that can be computed efficiently, in an online fashion, as a matrix operation. We design practical lightweight compressive representations …

Panoramas From Photons, Sacha Jungerman, Atul Ingle, Mohit Gupta

Computer Science Faculty Publications and Presentations

Scene reconstruction in the presence of high-speed motion and low illumination is important in many applications such as augmented and virtual reality, drone navigation, and autonomous robotics. Traditional motion estimation techniques fail in such conditions, suffering from too much blur in the presence of high-speed motion and strong noise in low-light conditions. Single-photon cameras have recently emerged as a promising technology capable of capturing hundreds of thousands of photon frames per second thanks to their high speed and extreme sensitivity. Unfortunately, traditional computer vision techniques are not well suited for dealing with the binary-valued photon data captured by these cameras …

Understanding And Predicting Cognitive Improvement Of Young Adults In Ischemic Stroke Rehabilitation Therapy, Helard Becerra Martinez, Katryna Cisek, Alejandro Garcia-Rudolph, John Kelleher, Andrew Hines

Articles

Accurate early predictions of a patient's likely cognitive improvement as a result of a stroke rehabilitation programme can assist clinicians in assembling more effective therapeutic programs. In addition, sufficient levels of explainability, which can justify these predictions, are a crucial requirement, as reported by clinicians. This article presents a machine learning (ML) prediction model targeting cognitive improvement after therapy for stroke surviving patients. The prediction model relies on electronic health records from 201 ischemic stroke surviving patients containing demographic information, cognitive assessments at admission from 24 different standardized neuropsychology tests (e.g., TMT, WAIS-III, Stroop, RAVLT, etc.), and therapy information collected …

Learnings From A National Cyberattack Digital Disaster During The Sars-Cov-2 Pandemic In A Pediatric Emergency Medicine Department, Fiona Leonard, Hugh O'Reilly, Carol Blackburn, Laura Melody, Dani Hall, Eleanor Ryan, Kate Bruton, Pamela Doyle, Bridget Conway, Michael Barrett

Articles

Objective: The primary objective was to analyze the impact of the national cyberattack in May 2021 on patient flow and data quality in the Paediatric Emergency Department (ED), amid the SARS-CoV-2 (COVID-19) pandemic. Methods: A single site retrospective time series analysis was conducted of three 6-week periods: before, during, and after the cyberattack outage. Initial emergent workflows are described. Analysis includes diagnoses, demographic context, key performance indicators, and the gradual return of information technology capability on ED performance. Data quality was compared using 10 data quality dimensions. Results: Patient visits totaled 13 390. During the system outage, patient experience times …

New Fxlmat-Based Algorithms For Active Control Of Impulsive Noise, Alina Mirza, Farkhanda Afzal, Ayesha Zeb, Abdul Wakeel, Waqar Shahid Qureshi, Ali Akgul

Articles

In the presence of non-Gaussian impulsive noise (IN) with a heavy tail, active noise control (ANC) algorithms often encounter stability problems. While adaptive filters based on the higher-order error power principle have shown improved filtering capability compared to the least mean square family algorithms for IN, however, the performance of the filtered-x least mean absolute third (FxLMAT) algorithm tends to degrade under high impulses. To address this issue, this paper proposes three modifications to enhance the performance of the FxLMAT algorithm for IN. To improve stability, the first alteration i.e. variable step size FxLMAT (VSSFxLMAT)algorithm is suggested that incorporates the …